aiLL 6/27/23 - The Democratization of AI and What It Means for You

Description

Chapters

Transcript

0:16 hey hey it is Kyle Shannon with the AI
0:19 learning lab welcome welcome welcome
0:21 it is
0:24 Tuesday Tuesday night
0:27 about 7 20 here in Denver Colorado
0:30 welcome everybody this is the AI
0:32 learning lab we talked about generative
0:34 AI in here and
0:36 all things Ai and what's happening and
0:38 what's not happening and
0:40 how much it hallucinates and
0:43 how cool it is
0:45 if you have questions about generative
0:47 AI pop them down in the questions below
0:49 sorry my glasses are dirty I can't see a
0:51 thing and uh I'll do my best to answer
0:55 things and uh we'll wait for some more
0:57 folks to get in here um if you're new to
1:00 generative AI if you haven't played with
1:01 chat GPT
1:03 check out the the URLs behind my head
1:06 here so chat.openai.com that is the
1:09 official chat GPT website if you've not
1:11 played with that go play with that it's
1:14 worth playing with and then Bing that's
1:16 uh Microsoft search engine that if you
1:18 click the chat button that's also chat
1:20 GPT it's not quite as good as the uh the
1:24 official one just because of some things
1:27 that happened there they kind of had to
1:29 detune it a bit and then you got poe.com
1:31 where you have access to six different
1:33 chat GPT models and a bunch of bots and
1:36 cool stuff there so go play there
1:38 a little background on me while while
1:41 we're waiting for folks to come in I'm
1:43 an entrepreneur of many decades I've
1:45 started I don't know 12 or 15 companies
1:47 something like that I started one of the
1:49 early digital agencies in the mid 90s
1:51 and the uh the parallels between the
1:55 early days of the web and what's going
1:57 on with this generative AI stuff are
1:59 insane
2:00 and so I had a big Epiphany about the
2:02 web back in 94 and I had a similar
2:04 Epiphany about generative AI in December
2:07 of 22. so
2:10 um so welcome and that's a little bit
2:12 about me
2:14 um I don't want to be oh I want to be
2:17 rich oh and there's comedy we do comedy
2:21 here uh I do like to swear so if you
2:23 don't like swearing motor on uh I don't
2:26 know that I like this where I get
2:27 passionate about stuff when I get
2:29 talking about stuff I get passionate
2:30 about it I'll drop an F-bomb I'll call
2:33 something shitty you know it happens
2:34 things happen I'm a next New Yorker I
2:36 got a little Edge but I've been living
2:38 in Denver now for 14 years so I I think
2:41 I think all those sharp edges have been
2:43 rounded
2:44 Denver's a polite City it's too polite
2:48 it's too polite I need some [ __ ] in
2:50 my life all right and by the way
2:53 speaking [ __ ] if you're gonna troll
2:55 here at least be clever with it at least
2:58 be clever
3:00 um you can tell me which d-list
3:02 celebrity you think I look like or
3:05 I had a couple of a-listers but they
3:07 they were all there was always sort of a
3:10 it was somewhere between a compliment
3:12 and a troll which was good good that's
3:14 solid that's that's good work uh if you
3:16 have questions about generative AI pop
3:18 them in the comments below
3:20 um
3:21 let me see it looks like there's a bunch
3:22 more people here so um
3:24 that top URL that's the official chat
3:27 GPT website I think you owe it to
3:29 yourself if you haven't played with chat
3:30 gbt
3:31 go play there long enough that you have
3:33 a Kevin McAllister moment the Home Alone
3:35 moment
3:37 it will happen
3:39 um if you if you don't have a clue where
3:40 to start with this stuff go to this URL
3:42 right here prompts.chat so this is just
3:44 like a Word document that tells you
3:47 about prompting and how to use chat GPT
3:49 but then it's a whole list of prompts
3:51 that that teach you how to turn Chat GPT
3:54 into into different kind of personas
3:56 like act as an accountant act as a
3:58 screenwriter act as a lawyer
4:01 and it'll give you a real sense of the
4:03 depth and power of that tool and then
4:06 Bing and Poe are both just variations
4:09 you can use uh chat GPT in both of those
4:12 as well that's what those URLs are there
4:15 for and then futurepedia.io that is just
4:18 a big old directory of like I don't know
4:20 3 700 AI apps so if you wanna
4:25 Go Geek out you can do that but quite
4:28 frankly
4:29 I feel like I could spend the next
4:31 decade just at chat GPT
4:33 not even scratching the surface of what
4:35 it makes possible I mean it's it's it's
4:37 Insanity it's Insanity it's Insanity I
4:40 tell you it's crazy crazy I tell you
4:44 um so anyway if you have um oh a lot of
4:46 people following me I appreciate that
4:47 Dan Aykroyd from Walmart that's solid
4:50 that's solid uh John good John Goodman
4:54 I've gotten John Goodman before Dan
4:56 Aykroyd from Walmart haven't gotten that
4:59 one I've gotten
5:00 um
5:01 um Alec Baldwin but fat I've gotten a
5:04 lot of Rosie O'Donnell
5:05 which again
5:08 compliment no I don't think so troll nah
5:12 hard to know gotten Philip Seymour
5:14 Hoffman
5:15 some polite people said Nicholson but
5:17 you know I think I you know I don't know
5:20 not with these glasses but
5:22 but the uh Dan Aykroyd from Walmart
5:24 that's solid that's solid
5:28 all right uh I appreciate y'all
5:31 following me pop some questions down
5:33 there is there wait is there a prompt
5:36 there you go I do wait what did I say
5:38 meta victory
5:40 uh
5:44 I don't know what your question is is
5:45 there a prompt I don't know what Dollar
5:48 Tree I was Baldwin
5:53 Dan Aykroyd from Walmart Dollar Tree
5:56 Alec Baldwin
5:57 that you you're on tonight y'all are on
6:01 tonight that's solid that's solid comp
6:06 ly trolls
6:08 troll lament compliment I don't know
6:11 it's good they're solid AI is the new
6:14 grift completely disagree I assume that
6:16 you have not played with it and that's
6:18 why this is here if you haven't played
6:20 with it you can make snarky ass comments
6:22 about AI but only if you have used it
6:26 if you haven't go there right now make
6:29 an account and start playing with it
6:30 don't just listen to people that are
6:32 like it lies to you it's not perfect
6:36 I told her to write me a poem and the
6:38 poem sucked
6:40 don't be that guy
6:41 go there dig deep start pushing it
6:45 um I'm happy to show how it works I'm
6:46 happy to show plugins
6:48 um I wish I could get it to generate a
6:51 response that makes sense then I'd be
6:53 sold well makes sense in what context so
6:56 if tell me what you're trying to solve
6:59 and maybe we can go play I'll go I'll go
7:00 hop on the chat GPT and we'll see what
7:02 we can we can do it can program it's
7:05 really good at writing gpt4 in
7:07 particular is really good at writing
7:08 it's good at programming
7:10 um cousin cousin Rosie that's nice
7:13 cousin Rosie again solid solid Troll
7:16 game here kids uh prompt was below when
7:20 you were talking
7:21 uh I don't know I don't know what it was
7:24 okay
7:25 I was afraid I would miss you and not
7:26 find you tonight I'm here sniper thank
7:29 you welcome
7:31 hello someone said hey Kyle who was that
7:33 my shirt okay here's the deal I my my
7:37 wife buys me lovely shirts this is a
7:39 lovely shirt but they are horrific for
7:42 reading stuff I might go change my shirt
7:44 here
7:46 um what are vector databases okay
7:48 so
7:50 the way these tools work it's absolutely
7:53 fascinating the more you learn about how
7:55 these work the more mind-blowing it is
7:57 that they do what they do so what they
7:59 do is they
8:01 they sort of Point these large language
8:03 models using a thing called a
8:05 Transformer they they pointed it kind of
8:07 all of the internet in in the case of
8:09 open Ai and these large language models
8:12 they basically find as much data on the
8:15 public internet that they can Wikipedia
8:17 and blogs and websites and research
8:19 facilities and whatever just all the
8:21 data they go find all the data and then
8:23 they do this process called embeddings
8:25 and what the embeddings do is they
8:28 convert words into numbers a series of
8:32 numbers that are vectors so you know how
8:34 like in uh in junior high we learn sort
8:38 of XY coordinates right you plot things
8:41 on an XY graph right so those are
8:43 vectors but instead of it being two
8:45 Dimensions it's like thousands of
8:47 dimensions and what happens is all of
8:49 the words on all of the data like all of
8:52 the internet get turned into these
8:53 coordinates and they they effectively
8:56 get grouped semantically so so you might
8:58 have the word dog could live in lots of
9:00 different places semantically in one
9:02 area dog is the thing that we have in
9:05 our house and we love and it's furry and
9:07 in another area a dog is the guy that
9:09 did his woman wrong right so so those
9:12 live in semantic different places
9:14 so a vector database is where you store
9:17 all of those numerical figures and so
9:20 what happens with a large language model
9:21 is you type in your prompt and then
9:23 based on your prompt it converts that
9:25 into some sort of math
9:27 and then it it basically goes into the
9:29 vector database and says what is the
9:31 next most probable word in all of the
9:34 words you've got in that Vector space
9:36 they call it in that Vector database
9:38 which one is the next most probable word
9:40 and then after that one which one is the
9:42 next most probable word so it's
9:44 literally just doing sort of
9:46 sentence autocomplete but like on
9:49 steroids right so when it writes you a
9:51 whole marketing plan it's doing it one
9:53 word at a time in a vector database
9:56 Vector databases in in in kind of
9:59 today's technology they're obviously
10:01 used for things like like chat gbt and
10:04 GPT three and four but there's a thing
10:06 called Lang chain which allows you to
10:09 basically fine-tune the large language
10:12 models on your own data so basically you
10:15 take your own data you embed it you use
10:18 open ai's embedding technology or
10:20 whatever model you're using and then
10:22 those vectors those number sets get
10:25 stored in a vector database so that's
10:27 that's what the vector database is hope
10:29 that wasn't too geeky
10:31 um just finished the course on prompt
10:33 engineering from Vanderbilt University
10:35 awesome that's that's awesome how was it
10:37 I'm curious how it was because there's
10:40 this is still relatively early to be
10:42 having prompt engineering courses but
10:44 that's great
10:45 um AI drone attacking operator real or
10:48 not um well who who knows it's military
10:50 so
10:52 at a maximum it was a simulation and
10:56 then
10:57 um and then that that same person came
10:59 out and said well it was just a thought
11:01 experiment so either in a simulation
11:04 that happened or and and he was told
11:07 just go say it was a thought experiment
11:09 or it was just a thought experiment and
11:11 he kind of just misstated it so but it
11:13 didn't happen for real but but what the
11:15 thing was was for those of you that
11:18 don't know there was an Air Force drone
11:20 operator in either a simulation or
11:22 thinking about a simulation
11:25 that basically had an AI powered drone
11:28 that that was set to go kill the enemy
11:30 and at whatever point the operator said
11:33 no don't go kill the enemy and and kind
11:35 of withdrew the the
11:38 um the command to go get the enemy and
11:40 the Drone didn't like that so apparently
11:42 it turned around and fired on the
11:43 operator I don't know that it's true I
11:45 think that's like
11:47 like it certainly wasn't true with a
11:49 physical thing it's the kind of thing
11:51 that could happen like I'm not saying it
11:53 couldn't and I'm not saying it didn't
11:54 but you know the official line from the
11:57 Air Force is well that was just a
11:59 thought experiment so
12:01 so it's as real as that it's it's as
12:03 real as the communications uh team from
12:06 the air force behind you in your images
12:09 I was asking about where a prompt
12:12 bulletin board was but you got behind
12:15 oh oh behind you on your oh oh oh oh
12:19 I see behind me these things sorry about
12:22 that totally unclear on my part
12:24 so the purpose of all these is really
12:27 this is really just kind of Public
12:29 Service stuff so the top three are
12:32 basically three different places you can
12:34 play with chat GPT
12:37 chat.openai.com is the official chatgpt
12:39 website bing.com is Microsoft search
12:41 engine that if you click the chat button
12:44 or if you use chat in the edge browser
12:46 that's gpt4 connected to the internet
12:49 and it's free it's not as good as this
12:51 one but it's good and it's gpt4 and then
12:54 poe.com gives you access to gpt4 GPT 3.5
12:58 Claude plus Claude instant Claude
13:00 instant 100K and Sage so there's six
13:03 different models you can play with there
13:05 prompts.chat is is it basically just a
13:08 document it's called awesome chat GPT
13:10 prompts
13:12 and what it is is it will teach you
13:14 about prompting and then it's a whole
13:15 long list of personas that you can
13:18 assign chat GPT to act like so act like
13:21 a screenwriter act like a business
13:22 consultant act like my therapist act
13:25 like a companion act like a
13:27 translator act like a programmer right
13:29 and so that's what that is and then
13:31 futurepedia.io is just a big old list of
13:34 uh
13:37 AI apps and services and [ __ ]
13:42 all right and Court in wait encourage
13:45 more viewers to share oh someone shared
13:48 life oh yeah if you would share this
13:49 live or tap the screen to let people
13:51 know we're here I'm just getting started
13:53 although my my voice is already going
13:55 interesting
13:57 um I can I can also show how the chat
13:59 CPT stuff works if you want
14:01 um I can also demo plugins show you how
14:04 they work and what they are uh because
14:06 they are not intuitive
14:09 um are you as patient with your
14:10 employees as you are with us asking
14:13 silly questions I sure hope so I you
14:14 know listen I I think
14:17 it is my deep belief that everyone's
14:19 just trying to [ __ ] figure life out
14:21 and so none of us should make it harder
14:24 like I feel I feel like we've just
14:27 gotten so [ __ ] nasty with each other
14:29 because of all the political [ __ ]
14:30 that I I just I want to get back to
14:33 empathy
14:34 like can we just not be [ __ ] to one
14:37 another like if you're gonna troll in
14:38 here say I look like Rosie O'Donnell
14:41 sure it's mean but it's funny
14:44 you know there's love behind that
14:51 I'm new to this so it was in so it was
14:54 enlightening great place to oh great
14:56 place to start great awesome terrific
14:59 how do you save your prompts for later
15:00 use so uh well okay that great question
15:04 someone needs to [ __ ] solve that one
15:07 and I think they will
15:09 um one of the things about these tools
15:11 is that they are very new
15:14 um chat EBT the initial version of it
15:17 was written in a two-week Sprint uh by
15:20 the people at open AI don't know if you
15:21 knew that
15:22 it was also the instructions for how to
15:25 build chat GPT we're in the
15:28 um documentation for gpt3 and nobody
15:31 built it and so Sam Alton finally said
15:33 [ __ ] it Go build it and they built it
15:35 and I got to 100 million users in six
15:38 weeks
15:39 fastest adoption of uh technology in
15:42 history okay
15:44 in chat GPT so let me let me go to a new
15:47 chat here
15:48 so in chat GPT this is the window if you
15:50 haven't been there
15:52 um I've got chat GPT plus which gives me
15:54 access to gpt4 you just type your
15:57 message down your prompt down here so
16:00 um uh let's see let's see what's going
16:03 to be a good one
16:04 um
16:06 make me a marketing plan for my aloe
16:13 Vera
16:15 smoothie company
16:20 and sure there it is okay so now
16:23 over here on the lift this is your
16:24 prompt history so good news is that chat
16:27 CPT saves these and it will eventually
16:30 rename it to it'll probably rename it to
16:33 something like aloe vera yeah aloe
16:35 smoothie Mark marketing plan right
16:38 so so the good news is it saves your
16:41 prompts here a couple of bad things
16:42 about it you can't search them which is
16:44 a real drag also if you're concerned
16:47 about privacy and security
16:49 if you go to your settings there's a
16:51 thing called Data controls and one of
16:54 the things that you can do is you can
16:55 turn off your chat history
16:57 and training so that means giving your
17:00 data to open AI to train these models so
17:03 if you want to be secure with your data
17:04 and in fact the the house U.S House of
17:07 Representatives just put out a document
17:09 today saying that chat GPT is is the
17:12 only
17:13 um
17:14 generative AI that that the house is
17:17 allowed to use and they have to turn
17:19 this off
17:20 the problem with that is it deletes all
17:22 of your history and it won't store it
17:24 anymore because that's a more secure
17:25 thing to do so if you're going to do
17:27 that make sure that you do this thing
17:29 export data so you export your your chat
17:31 history so then the question becomes
17:35 um if these things aren't searchable and
17:37 depending on your settings they aren't
17:38 there at all like how else would you do
17:40 it
17:40 um
17:41 I've got ADHD and I'm I'm really good at
17:45 creating things like spreadsheets for
17:47 how to organize stuff but then I'm
17:49 really horrible at actually using them
17:51 so I don't have a good way right now I
17:53 think you could use notion I think you
17:55 could use spreadsheets I think that
17:57 there will be probably there probably
18:00 are right now Chrome plugins to manage
18:02 prompts
18:04 um I know that there are whole sites
18:05 dedicated to prompts
18:07 um
18:08 we're so early right now that that
18:11 I'm not super concerned if I can't like
18:15 like if I don't have a perfect system
18:16 for it because I think how prom how
18:19 prompting happens is Gonna Change you
18:22 know pretty dramatically
18:24 um over the over the next six months to
18:25 a year like I don't think
18:28 right now everyone essentially has to be
18:30 a prompt engineer this is this is very
18:32 similar to the early days of the World
18:34 Wide Web where if you wanted to make a
18:37 web page you had to be a little geeky
18:38 you had to learn stuff like there wasn't
18:40 all this point-and-click stuff that that
18:42 we have today it's very similar with the
18:44 generative AI all these apps are like
18:46 you know duct tape and spit holding them
18:49 together right now
18:50 so not a not a good answer not really an
18:55 answer at all about how do I manage
18:56 prompts or or you know how do I how do I
18:59 save them
19:00 uh drone Pilots are safe for now that's
19:02 correct
19:03 um it will get real I agree with that
19:05 the Vanderbilt course taught me how to
19:08 have chat EPT teach me more about prompt
19:10 engineering oh that's super cool yeah so
19:12 one of the big one of the big big
19:14 lessons for me personally that that made
19:17 the biggest difference for me with with
19:19 prompting
19:20 was a mindset shift and I've talked
19:23 about this on these lives before but
19:27 if you think about how Google has taught
19:30 us to use Google
19:33 what they've effectively taught us is
19:35 write in short discrete
19:39 searches right
19:41 short efficient searches without any
19:43 extraneous stuff because the extraneous
19:45 stuff will confuse it and won't know
19:47 what to find it
19:48 and they've also kind of taught us like
19:50 you put in one prompt and you get back a
19:52 pile of answers and you're done
19:54 Chachi PT is almost the opposite it's
19:56 it's a completely different mindset so
19:58 rather than thinking like oh it's a
20:00 question and answer kind of thing it's
20:03 more like a conversation it's more like
20:06 a like a collaborative back and forth
20:09 with an intern who needs a lot of
20:12 coaching
20:13 but who's smart as [ __ ]
20:15 knows everything it's like chat EBT is
20:18 like having an intern that knows
20:20 everything but is totally clueless
20:22 and so so doing things like hey chat EBT
20:27 I want you to act like a business coach
20:29 I'm just starting this business and I
20:31 don't know what to do and here's what my
20:34 business is and here's what I'm thinking
20:35 about is that something you could help
20:37 with sure I can help with that and it'll
20:39 give you all sorts of answers and then
20:40 you could even say you know what I want
20:42 you to do is I want you to ask me 12 or
20:44 15 questions about my business so you
20:46 understand it and then together we can
20:48 start working on absolutely and it'll
20:50 probably blast out 12 or 15 questions
20:53 and then you can say no no just ask ask
20:55 me the questions one at a time and don't
20:57 move on to the next question until I've
20:59 answered the previous one oh sure here's
21:01 the first question and then you answer
21:03 that and you'll see
21:05 it gets better and better and better as
21:07 you give it more context and then
21:09 there's a point at which it sort of
21:11 shits the bed after that and I can talk
21:12 about that that's got to do with the
21:14 What's called the context window so
21:15 basically basically it's little memory
21:17 it's got a relatively short memory right
21:19 now so it it kind of frips out if you
21:21 talk too long but
21:23 that idea of chat GPT more more like a
21:28 very very competent intern or a creative
21:31 collaborator versus kind of a search
21:33 engine Dynamic was the biggest thing
21:35 that made a difference to me and yeah
21:37 you can do things like teach me how to
21:39 prompt you better you can say things
21:41 like hey you gave me an answer there can
21:44 you evaluate your own answer and tell me
21:46 how you'd improve it and then it'll get
21:48 much better
21:50 um that's that's a hack
21:51 um rather than writing one massive long
21:54 prompt right like lots of little pieces
21:56 and and it just gets smarter as you as
21:58 you sort of give it more context it's
22:00 pretty pretty fun
22:02 um okay so welcome everybody my name is
22:04 Kyle Shannon this is the AI learning lab
22:06 I'm happy to answer any questions uh pop
22:08 them in the comments below I try to
22:10 answer them all I do get very far behind
22:13 I'm probably not doing Tick Tock right
22:15 but I'm a gen xer so I don't give a [ __ ]
22:17 and you know you know it's free I'm on
22:20 here free I I got a pretty picture of an
22:22 eye a creepy eyeball and a hairball with
22:26 a sort of like creepy chin down here and
22:29 then I got URLs like I put these URLs on
22:32 screen for free it's not like you can
22:34 just go find those on the internet well
22:37 you can but that's not the point at all
22:39 all right uh ask questions my my
22:43 background by the way is I'm an
22:45 entrepreneur I've started a lot of
22:47 companies
22:48 um I was an entrepreneur in the early
22:50 days of the web and had a big Epiphany
22:53 about that [ __ ] like holy [ __ ] this is
22:55 going to change the world and it did uh
22:57 and and this generative AI stuff is like
23:00 that but way bigger and way faster so so
23:03 I'm very excited about it I'm excited
23:05 about it enough that I come here every
23:07 night rather than doing Sim racing which
23:10 I love Sim racing
23:12 um but I just I've been coming here so
23:14 anyway that's what's up with me ask
23:17 questions below and I will do my best to
23:20 answer them I'm not computer savvy
23:22 but I log in a lot for day-to-day stuff
23:25 and it saves me so much time it's
23:26 amazing I know Laura like that's the
23:28 thing like even if you're not computer
23:29 savvy the the thing about this is like
23:32 you can use it for anything right you
23:34 can have it create meal plans for you if
23:36 you're a gardener you can tell it when
23:38 to plant you know ask it when to plant
23:39 stuff
23:41 um
23:42 and
23:46 the other the other thing is this we are
23:48 not that far away probably within a year
23:50 we'll start to see these where you won't
23:53 have to go to a website like
23:55 chat.openai.com these tools are now
23:58 starting to get integrated in all the
24:00 other tools we already use right like
24:01 within Photoshop there's now this thing
24:04 called generative fill
24:05 Microsoft has said they're going to roll
24:07 chat gbt across all of office 365.
24:10 Google has said they're going to roll
24:11 Bart across all of their office suite
24:14 um
24:15 it's in GitHub there's co-pilot where
24:19 where Engineers are pair programming
24:21 with this thing so it's going to be
24:22 everywhere the other thing that it's
24:24 going to be is right now you have to do
24:26 prompt engineering and you have to type
24:28 a lot and and it doesn't always give you
24:29 good answers and it hallucinates and it
24:31 lies and there's all that sort of happy
24:33 horseshit you got to deal with
24:35 where we're going in the future is
24:37 you'll be able to just you'll basically
24:39 have like an agent that you can just ask
24:43 to go do [ __ ] right so so all the stuff
24:46 you kind of have to like you know code
24:48 code around with with whatever your
24:50 native language is
24:51 um you'll just be able to talk to your
24:53 agent and say hey go book me a trip or
24:55 hey go make me a meal plan or do this or
24:57 do that it will sort of figure out all
24:59 the stuff it needs to do to go execute
25:00 that so that's kind of where we're
25:01 headed is a a much more seamless
25:03 experience but right now
25:06 the the the cost of getting early access
25:08 to the most powerful technology ever put
25:11 in the hands of mere mortals is that
25:13 it's a little janky gotta deal with the
25:15 Jank
25:16 and it you know it's that that why not
25:19 why not why not all right
25:23 did you sign up as a developer waiting
25:25 for your approval
25:28 YouTube collaborator did you sign up as
25:30 a developer waiting for your approval
25:32 YouTube
25:33 oh calibrator collaborator
25:37 um I don't quite know what you mean I
25:39 signed up as a developer for openai yes
25:43 and
25:44 um they haven't given me access to gpt4
25:46 yet and I'm cranky about it and they
25:48 haven't also given me access to code
25:50 interpreter which if you if anyone on
25:52 here is into data analytics or is it
25:55 data analyst and you have not heard of
25:57 code interpreter go to YouTube right now
26:00 and search for code interpreter chat GPT
26:03 just go do it watch the video and watch
26:06 a bunch of data analysts go like this
26:10 it does what
26:12 it's insane or go to Twitter and look up
26:15 code interpreter so people that have
26:17 access to it their minds are absolutely
26:19 melting basically what you can do is you
26:20 can upload a chunk of data to chat GPT
26:23 and then it just learns it and you can
26:25 just say what's interesting about this
26:27 data where are any anomalies in this
26:29 data make me a graph that compares this
26:32 trend to that Trend like insane it's one
26:36 of the things chat CPT is really good at
26:37 is analyzing other documents so for
26:40 example if you're a writer
26:44 you could give it a Blog that you wrote
26:47 and say hey here's a Blog that I wrote
26:49 paste it in there and then say
26:51 one you could have it analyze it and
26:53 tell tell you interesting things about
26:54 it but you could also say write me a new
26:57 blog on this topic that includes these
27:00 four bullet points bullet bullet bullet
27:02 bullet uh and then write it in the style
27:04 of the thing that I gave you above and
27:06 it'll do it brilliantly it's really good
27:08 at that sort of
27:09 analysis of documents and reconstituting
27:12 of them it's brilliant at that stuff
27:16 agree 100 I don't know what I said but
27:20 awesome unless I said something shitty
27:22 and then I'm not glad with you agreeing
27:24 with me but I don't think I did okay I'm
27:27 just did my girlfriend is looking into
27:29 prompt engineering I'm a security
27:31 engineer so she seems to like
27:33 engineering that's cool I think you're a
27:36 security engineer second
27:38 security engineer
27:39 um
27:40 yeah I listen why this channel exists is
27:43 if you haven't played with chat GPT
27:45 that's what these URLs are that's the
27:47 official one those are two Alternatives
27:50 um
27:53 you owe it to yourself to go there and
27:55 start playing with this stuff it whether
27:57 or not this is going to be called prompt
27:59 engineering for long
28:01 um remains to be seen right now it's
28:03 kind of like prompt engineering like
28:04 like working with chat GPT
28:07 the the thought process is is very much
28:10 like programming but you're kind of
28:11 programming with in my case the English
28:13 language whatever your native language
28:15 is
28:16 so you're thinking kind of logically
28:18 about how you're going to ask this thing
28:20 that kinda knows everything but is
28:22 really stupid how to give you the
28:24 results that you want so that's kind of
28:26 the art of it and the science of it
28:29 um but yeah I think it's I think anyone
28:31 that spends time learning this stuff
28:34 right now it's going to pay big
28:36 dividends over the next two or three
28:37 years
28:39 because this stuff's this this stuff is
28:42 coming way faster than I think anyone's
28:44 prepared for great attitude a little bit
28:46 I don't know what lobit said but I
28:49 assume that's snarky uh how does a
28:52 complete Noob use chat GPT
28:55 um okay let's uh let me go show you
28:57 right now we'll go through some complete
28:59 Noob stuff
29:01 um but what you should do as a complete
29:03 Noob is go to chat.openai.com sign up
29:05 for an account
29:07 they will ask for your phone number if
29:09 you're if you're like an nft nerd and
29:11 you're like I won't dox myself to the
29:14 man then go to bing.com that's just
29:16 Microsoft search engine and you can use
29:18 their chat GPT for free there I don't
29:21 think you need to sign up for that so
29:22 just click on the chat button there
29:23 that's also chat gbt but that one's
29:25 better
29:27 um just because they invented it and
29:28 there's reasons why Bing isn't as good
29:30 right now okay
29:32 complete new
29:34 so step one go there also in another tab
29:38 go to this URL go to prompts.chat this
29:41 is a document that will that will teach
29:43 you about prompting and give you a bunch
29:45 of ideas and a bunch of examples
29:47 but let's
29:52 flip over
29:56 let's go live to the internet all right
29:59 so here we are chat GPT you can make my
30:02 name
30:05 all of those Tick Tock channels that
30:07 sound like that
30:08 ignore those
30:10 those are people trying to rip you up
30:15 you can buy my course by 19.95. you can
30:18 make money right okay
30:22 so if you're complete Noob so when you
30:23 come to chat gbt like I said this is
30:26 your chat history
30:27 if you pay for the plus uh thing that
30:31 gives you access to gpt4 to look like
30:32 this if you don't it'll look slightly
30:34 different I think they call it GPT turbo
30:37 or something like that if if uh if it's
30:39 the free version but it's really good
30:41 and then you can do things like
30:45 um these are what I call parlor tricks
30:47 so so you can just start asking it
30:50 things like I'll write a really bad
30:52 prompt here right
30:54 me a
30:57 short story about
31:01 um a tree
31:07 bad prompt
31:09 but then it'll write you a little short
31:11 story and it's Once Upon a Time
31:14 nestled within a lush sprawling Forest
31:16 at a magnificent tree named Eldred it's
31:18 trunk with solid and Tall baba baba
31:20 boring boring poem but whatever that's
31:22 fine you can say
31:24 um
31:25 can you
31:27 tell me who some
31:31 famous short story writers
31:35 are question mark
31:38 Edgar Allan Poe check off
31:42 guide him up Asante Oh Henry
31:44 Kate Chopin or Chopin or chopping
31:50 Chopin
31:52 Ernest Hemingway so let's say
31:55 um
31:55 [Applause]
31:58 how would Poe
32:01 write that story
32:05 and this should go much darker
32:09 because Poe is dark in the Deep recesses
32:11 of a mysterious forest cloaked in the
32:13 Ethereal Mist loomeda colossal and
32:16 ancient tree it's gnarled branches reach
32:18 toward the heavens twisting and turning
32:20 like the fingers of ghostly specter blah
32:22 blah blah okay
32:23 um
32:25 awesome and then now we can say turn
32:28 this into an m
32:32 rap
32:35 and
32:37 boom
32:38 there's a bunch of verses that all rhyme
32:41 so and our dark like Poe like somewhere
32:44 between Edgar Allan Poe and Eminem
32:45 there's your there's your poem and then
32:48 you could say
32:49 um
32:50 I want
32:52 um
32:53 I want that uh
32:57 poem
33:00 to look
33:03 pretty can you make it in HTML
33:09 and look nice
33:13 and boom there's your HTML it knows code
33:17 and you can go use that code or you can
33:20 modify the code or you can tell it I
33:21 don't like your code give me better code
33:23 let's see did it do any CSS stuff here
33:27 um yeah
33:29 there's your CSS stuff there
33:31 so you can do that you can save this
33:34 into a web browser and open the styled
33:36 version of the poem
33:38 can you
33:40 make a python
33:43 application
33:46 that writes
33:49 poems
33:51 in the style
33:53 of famous authors
33:57 question mark
33:59 certainly here's a python script that
34:02 does all that [ __ ]
34:03 and it just writes that and then you
34:05 know and it goes on and on and on um I
34:07 could also say uh wait
34:11 can you make a table
34:19 comparing the author's you suggested
34:24 above
34:28 certainly here's a table so it gives you
34:30 Poe
34:32 their nationality their notable works
34:33 and their writing style right and
34:35 there's and it and it does this thing
34:37 called markdown where it actually makes
34:40 the you know the formatting so I can
34:42 literally just copy and paste this and
34:44 dump it over to excel
34:46 so all of what I just did there are kind
34:49 of what I call
34:50 parlor tricks right like they're the
34:52 kinds of things that if you haven't seen
34:54 this before the fact that it can write
34:55 python code or that it can reformat a
34:59 poem into HTML complete with CSS and the
35:02 fact that it can write the poem in the
35:03 first place right or the story or know
35:05 all that sort of stuff
35:08 um if you haven't seen this before it's
35:09 this is clearly doing something
35:11 different than a Google Search right
35:13 this is why chat GPT is such an
35:16 existential threat to Google because
35:19 why would you so Google's entire
35:23 business model
35:24 is not about finding you what you need
35:29 their entire business model is to get
35:31 you to click on as many links as
35:33 possible because the more you click the
35:36 more they get paid
35:37 and there's a you know the entire SEO
35:41 industry is built around
35:43 gaming that system
35:45 so if you have something like chat EPT
35:48 that just gives you the answers
35:50 as an end user I'm like [ __ ] yes I don't
35:54 like right now
35:56 I will wait because chatgpt4 is slow the
35:59 one that's connected to the internet is
36:00 pretty slow so it sometimes will take 20
36:03 seconds 30 seconds to like give me an
36:05 answer
36:06 but I will go to chat GPT to get the
36:10 answer for something rather than going
36:12 to Google even though I know Google
36:14 would give me the answer more quickly
36:15 just because I'm like I don't want to
36:17 see all the crap I don't want to see all
36:18 the links I just want the answer so I
36:20 just fire off a prompt and then I go do
36:22 something else and I come back and there
36:24 it is
36:25 um so so it's it's just wild stuff so
36:28 anyway hopefully that that gives you a
36:29 sense if you're brand new to this stuff
36:32 um jump in there the other thing let me
36:33 show you one other thing
36:35 um
36:36 about uh when did I start this December
36:39 I started a community called the AI
36:41 salon and we meet every other week
36:44 our next meeting is July 5th the day
36:46 after the Fourth of July obviously huh
36:49 I'm good with math
36:52 um we meet every other week we also have
36:54 a Discord server where you can geek out
36:56 so if you go there the it's a link tree
36:58 to to that Community it doesn't cost
37:00 anything this is not I'm not doing this
37:02 to make money
37:04 um
37:06 anyway
37:08 I saved my prompts in notepad yeah I've
37:10 been using my notes app as well for
37:12 prompts in uh on my Mac
37:14 you don't look like Rosie thank you I
37:16 think
37:19 I guess it depends on the hair tonight
37:20 the hair is looking it's I think the
37:22 hairs looking like you know someone
37:24 quaffed it normally there is an abject
37:26 [ __ ] nightmare but the glue's holding
37:28 apparently all right I have a
37:31 spreadsheet I've been looking for
37:32 something to organize then Google drive
37:34 with a doc four
37:36 something for each prompt oh oh I have a
37:40 spreadsheet for prompts I've been
37:42 looking for something to organize then
37:43 Google drive with a doc for prompts yeah
37:45 a Dockers or spreadsheets you know some
37:48 sort of database notion I think notion
37:50 is probably the way to do it because
37:51 notion also has generative AI built into
37:54 it
37:57 I just came up with a thing the other
37:58 day that
38:00 it it it's it's the first time that I'm
38:03 actually confronting the fact that I
38:04 don't have a good system for this [ __ ]
38:06 where
38:07 I came up with a a sort of whole series
38:11 nested prompts that just generate
38:14 increasingly sophisticated output and
38:17 and I'm really excited about the whole
38:19 project but it's all over the place
38:22 right now and I've got prompts
38:23 everywhere and I kind of have to dig
38:24 them out of
38:26 um chat GPT and the playground and it's
38:28 a mess so anyway it's prompt management
38:31 is is a big deal because here's one of
38:34 the things that that happens for me with
38:36 with Chachi BT is
38:40 where I used to have writer's block
38:42 where I'd have an idea and then I'd
38:44 think oh but then I gotta write an
38:45 outline and
38:47 maybe I'll go get an outline piece of
38:49 software or a flowchart software and
38:51 then I'll spend like 45 minutes trying
38:53 to find flow charting software like all
38:55 of that's gone now like I I have an idea
38:57 go to chat GPT and bring you know like
39:00 the idea is manifested it's it's [ __ ]
39:02 Magic
39:04 um
39:06 the the problem is it
39:08 it
39:10 it helps you realize ideas so quickly
39:13 that
39:15 all of the energy that what you would
39:17 have spent creating the 80 first draft
39:20 that the hour that it would have taken
39:21 you to make the first draft historically
39:23 is now made for you in 30 seconds
39:27 and so then you're kind of in this place
39:29 of like well I have another idea and
39:31 then you realize that when I have
39:33 another one and so like I am more
39:35 creatively alive than I've been in
39:37 decades
39:39 but it's a challenge because it's just
39:41 like all the like all these ideas come
39:43 up and you know as an ADHD guy
39:46 um it's a blessing and a curse to be
39:48 able to to realize ideas this quickly
39:52 um so yeah prompt management is a thing
39:55 25 messages per three hours is such a
39:57 rip-off it's not a rip-off listen uh so
40:00 for one thing you can use GPT 3.5 as
40:04 much as you want and it's it's really
40:05 good I don't think I've ever hit the 25
40:10 um prompts in a session limit so I don't
40:14 know how like I think they probably put
40:16 that there just to
40:18 um
40:20 uh you know just to get people to to
40:22 maybe not use it as much
40:24 um and maybe in times when it's really
40:26 busy they enforce that but I've never
40:28 hit it I've never been told okay you got
40:30 to stop for three hours now
40:32 um it's also not for nothing it's the
40:35 most powerful piece of [ __ ] software
40:37 in the history of humanity
40:40 um so so
40:41 um the fact that we get to use it even
40:42 for 20 bucks is cheap but it's free also
40:45 if you want gpt4 for free go to bing.com
40:48 that's gpt4 it's connected to the
40:51 internet and it's free
40:52 free free as in no dollars so that not a
40:56 rip-off and I also don't think that's a
40:57 rip-off at all I think they're just
40:59 they're trying to figure out like they
41:02 got to 100 million users in six weeks I
41:05 don't I can't even imagine how many
41:06 users they're up to now probably 300
41:08 million is my guess because it's been
41:10 six months
41:11 um so so they've got you know luckily
41:14 they have Microsoft behind them scaling
41:17 up all the iron so they can handle that
41:19 many users but anyway you know I don't
41:22 think it's a rip-off I think it's a it's
41:24 a brand new technology that was way more
41:27 popular way more fast than anyone
41:29 including Google or Microsoft or open AI
41:33 or meta or apple or anyone expected
41:37 listen to Sam Waltman when he talks
41:39 about how quickly chat GPT got to a
41:42 million years it got to a million users
41:43 in five days it got to 100 million users
41:46 in six weeks the world wide web in the
41:49 90s it took um six years to get to 100
41:52 million users for the World Wide Web six
41:54 weeks for that
41:56 so the fact that they're throttling it
41:58 as they try to figure out how to scale
42:00 it and and launch all the new features
42:02 is I think it's a reasonable part of
42:05 where we are right now I don't think
42:06 it's a rip-off at all but you know maybe
42:08 I just drank the Kool-Aid and by the way
42:10 I don't work for them I don't give a
42:11 [ __ ] if you go there or not quite
42:12 honestly
42:14 um if they start paying me commissions
42:16 then I don't give a [ __ ] but I don't
42:17 think they're gonna do that
42:20 um looking at the downside sorry
42:21 couldn't this
42:23 be used to learn how to make things we
42:25 shouldn't yes so
42:27 if you go to there and you try to do
42:29 things like hey give me instructions for
42:31 how to make something I shouldn't it's
42:33 got safety guard rails in it right so it
42:35 says Hey as a large language model that
42:38 doesn't want you to hurt yourself or
42:39 others I'm not going to do that for now
42:42 and then if you go to Reddit there's a
42:43 thing called jailbreaking uh where where
42:46 people are trying to jailbreak uh chat
42:48 GPT to get around those guard rails and
42:50 there's also now a whole open source
42:52 Community where there's open source
42:54 large language models that don't have
42:55 those safety guard rails on them so this
42:58 is one of the dangers of of these large
43:01 language models and it's one of the
43:02 reason that Sam Altman from chat gbt and
43:06 some of the other leaders in the space
43:07 are saying hey government you should
43:10 regulate us like once once a company has
43:13 developed large language models of a
43:15 certain power like gpt4 and up they
43:20 should be regulated and Sam Altman is
43:22 actually saying once we get to a thing
43:23 called AGI or artificial general
43:25 intelligence then they should actually
43:27 have to be licensed just like you
43:29 license I want to go build a nuclear
43:31 power plant you actually have to have
43:33 that license by the government so so the
43:36 leaders in the industry are actually
43:37 saying when we're dealing with the super
43:39 powerful versions of this those should
43:42 actually be licensed and then the the
43:44 chat gbt level of them should be
43:47 regulated and then below that
43:50 you know they're saying they should not
43:51 be regulated like small companies should
43:54 be able to innovate on this stuff
43:55 without over overly burdening them with
43:57 regulation but that's why everyone's
44:00 talking about the regulation and safety
44:02 guard rails right now is for exactly
44:03 what what your concern is and and many
44:05 others right there's there's lots of
44:07 stuff this could do bad
44:09 um but but he here's where I said I
44:13 think that
44:15 um
44:16 if we just live in the place of oh these
44:19 things are going to kill us then we're
44:21 never going to learn what they are and
44:23 what they're actually good at and the
44:25 fact that that they might do
44:27 profoundly good things to humanity as
44:30 much as there's a risk of them doing
44:32 profoundly bad things to humanity and
44:34 I'm just choosing to live on the side of
44:36 I want to learn as much about this [ __ ]
44:38 as I can
44:40 and get as good at it as I can and and
44:42 Inspire as many people to do the same as
44:44 I can because I think that any positive
44:47 contribution in this world right now in
44:49 in all this stuff is going to have a
44:51 disproportionately positive effect on
44:53 the future trajectory of where this
44:56 technology goes
44:57 um so that's that's where my head is
44:59 right now
45:00 every prompt is like a new world it
45:02 absolutely is this honest to God like
45:05 this the depth of that little search box
45:09 is it's mind-blowing
45:12 uh uh uh uh welcome everybody looks like
45:16 there's some new folks here my name is
45:17 Kyle Shannon I'm an entrepreneur uh my
45:20 current gig is I'm the co-founder and
45:21 CEO of a company called storyvine which
45:24 is a video storytelling platform we
45:26 figured out a way 11 years ago to
45:28 automate video production uh we don't
45:30 use AI oddly enough to to do that
45:33 automation but we have Incorporated uh
45:35 open ai's uh whisper technology to
45:38 generate transcripts and I'm completely
45:40 rethinking the product and value
45:42 proposition of my 11 year old company
45:44 because AI is going to [ __ ] everything
45:46 up yay
45:47 [Laughter]
45:50 but
45:52 I'm really excited about it like I'm you
45:54 know
45:55 I'm I'm as excited as I've ever been
45:57 about a technology and as scared as I've
45:59 ever been about a technology and that to
46:01 me feels like the appropriate emotional
46:04 response to this stuff right now
46:06 um yes me too wrote several several new
46:08 business plans in three months now now
46:10 got to decide which to do yeah you can
46:13 that's a thing man with with these tools
46:15 like
46:16 what these really are is these These are
46:19 amplification tools so whatever are your
46:22 skills this will amplify them and
46:25 mid-journey for images and some of the
46:27 audio things and the animation things
46:30 amplify who you are these are Humanity
46:33 amplifiers they're incredible
46:37 um
46:38 but then you're kind of stuck with okay
46:39 so now what do I do like if I can just
46:41 have an idea and realize it like that
46:45 you gotta pick one or do you got to pick
46:47 two
46:49 like I started
46:51 I started the AI Salon the everyday AI
46:54 newsletter this channel
46:57 um and re-architecting my business I did
46:59 all of that within about a month
47:01 and now I'm like tired and then I
47:04 started doing these lives like like a
47:05 month ago
47:07 um will it audit my taco truck web page
47:09 and tell me how to SEO absolutely in
47:12 fact let me show you I'll show you I'll
47:14 show you some plugins right now
47:17 um just to show in general how plugins
47:19 work because they're they're not
47:21 intuitive and they're kind of a pain in
47:23 the ass
47:24 um so so let me just walk through
47:25 through how this works in general so if
47:28 you haven't been to chat gbt before this
47:29 is chat gbt
47:31 um I've got the Plus subscription so
47:34 this is the 20 Buck a month subscription
47:36 that gives me access to 3.5 which is
47:40 super fast and gpt4 which is magical and
47:42 good and then within that because I pay
47:46 they they've also given me access to to
47:50 gpt4 connected to the internet and
47:52 plugins now if you pay your 20 bucks a
47:55 month and you don't see these two
47:56 options you have to actually turn them
47:58 on so how you do that
48:00 is you come over to your settings
48:03 and then within your settings there's
48:05 this thing called beta features and you
48:07 turn those on
48:08 and there's browse with big and plugins
48:11 and so you turn those both on and then
48:13 you'll see these
48:15 and then the way plugins work is it's
48:18 it's just like normal chat gbt but you
48:21 now have this drop down menu that's got
48:23 here's all my installed plugins
48:25 and
48:27 um
48:29 I can in any given chat session I can
48:32 turn on up to three of these
48:34 so here's one called website performance
48:36 measure key metrics about your website
48:38 so we'll turn that on
48:41 um there's another one called Web pilot
48:43 which I think does this do that no
48:45 that's just oh it says QA web page data
48:49 browse in QA we'll turn that on
48:53 um
48:56 I'm loving it all right I'm gonna turn
48:58 off prompt perfect and let's go we'll go
49:00 to the plugin store
49:02 so the plug-in stores where you can
49:04 install new plugins and this is growing
49:06 daily
49:08 and as of today
49:12 there are 147 times four plug-ins so you
49:17 know almost 600 yeah almost 600 plugins
49:21 um and I can go and I can search so I'm
49:23 going to search for SEO
49:25 and there's probably going to be a bunch
49:26 that show up so let's see SEO core use
49:31 AI to analyze and improve SEO of a
49:33 website
49:34 I don't know which of these are any good
49:36 the the problem with the plugins right
49:38 now is a lot of them are [ __ ]
49:41 horrible
49:42 effortless effortless
49:45 effort Leslie buffer and post-social
49:50 that looks like a service I need to sign
49:52 up for some of these are just sort of
49:54 cheap ways to get you to go to their
49:56 website
49:57 Speedy marketing we'll install that one
49:59 and we'll go turn that on all right so
50:01 I'm going to install Speedy marketing so
50:04 now I've got Speedy marketing website
50:07 performance and web pilot turned on and
50:10 I'm going to go
50:11 I'd
50:13 like to know
50:16 how I can improve the performance
50:23 of story
50:26 vine.com which is my website
50:29 and so what it should do now is it will
50:31 figure out which of those plugins is
50:32 most appropriate so it pulled up website
50:35 performance it might do multiple
50:38 seems to be an issue with the
50:40 Performance Tool so it looks like that
50:42 thing broke did it break
50:45 oh please sign up for it right so like I
50:48 told you before
50:49 um
50:51 yeah so so we're gonna go turn that one
50:53 off let's go turn that one off
50:56 so this is this is one of the problems
50:58 with plugins right now is that there's
50:59 not really any sort of vetting for them
51:02 so I'll turn that off
51:04 so now we've got these two okay so let's
51:06 do uh let me just go back and
51:10 steal that from before
51:13 gonna go to the new chat
51:15 I'm gonna go to plugins
51:18 and
51:20 please check out storyline.com now let's
51:23 hopefully it grabs a plug-in that
51:25 actually works this time
51:27 dang you yeah crappy Developers
51:32 okay this is just giving me generic ones
51:35 um can you
51:39 oh my God can you analyze
51:44 uh
51:45 analyze
51:47 this site
51:51 and wreck omend specifically
51:57 and by the way I just did a bunch of
51:59 typos there it doesn't care about typos
52:01 that's another thing about chat GPT it
52:04 can understand other languages so your
52:06 little typos it does not concern it at
52:09 all
52:10 using
52:11 web pilot let's see what did it do the
52:14 website could benefit from more targeted
52:16 keywords for example things like video
52:18 storytelling automated video editing
52:20 user generated content could be used
52:23 more
52:24 ensure the website is fully optimized
52:27 I'm going to stop this I'm going to say
52:29 can you give me a report
52:34 that scores the site
52:38 in key areas
52:41 we'll see if it can actually go do some
52:43 analysis
52:45 so it's using web pilot let's see what
52:47 it's actually doing behind the scenes so
52:48 I can open this up so it says
52:51 go to go to storybine.com so it return
52:55 this Json so description
52:59 the content of storyvine
53:02 it looks like it got oh an error
53:04 occurred when it looked at something
53:09 so website content 7 out of 10 SEO
53:12 optimization 6 out of ten mobile
53:13 optimization 8 out of 10 user experience
53:16 7 out of ten call to action 7 out of ten
53:18 so
53:20 performance optimization 8 out of ten so
53:23 it's not bad so these scores are based
53:25 on the general analysis and may not
53:27 reflect the actual performance of the
53:28 website within areas for more detailed
53:30 and accurate thing so let me turn off
53:33 I'm gonna I'm gonna do a new chat now
53:36 and I'm going to turn off
53:39 web pilot I'm just going to leave on
53:41 that one that said SEO
53:44 so I'm going to turn off web pilot so
53:46 now what's this one called Speedy SEO
53:50 Speedy marketing
53:51 marketing tool oh for your Shopify store
53:54 oh it's not going to do that all right
53:56 anyway so that's how you use plugins so
53:58 yes you can do SEO analysis and
54:01 optimization you probably can also just
54:03 use browse with Bing
54:05 so this is just gpt4 connected to the
54:07 internet
54:09 um act as an SEO expert
54:15 and please go
54:19 analyze
54:22 uh and Report
54:25 on
54:29 how I can improve
54:33 storyline
54:36 .com
54:38 and now this should say yeah browsing
54:40 the web let's go look at what it's doing
54:41 clicked on the link to clicked failed
54:47 yeah it didn't it didn't like that
54:49 I'm unable to do it directly huh all
54:52 right
54:53 so I guess that's not a good way to do
54:55 it so find plugins that can read
54:57 websites and then have it go do that so
54:59 that's how I would do that
55:02 all right back to questions back to
55:04 questions
55:06 happy Tuesday night welcome everybody my
55:08 name is Kyle Shannon this the AI
55:10 learning lab I'm trying to turn my
55:11 little background back on Boom if you're
55:13 new here if you're new to chat GPT that
55:16 top URL is the official chat GPT website
55:19 go there right now just do it go there
55:22 I'm not affiliated with it thank you for
55:24 the lovely Swan Swan I think that's a
55:26 swan
55:28 um if you don't like that and you love
55:31 Microsoft go to bing.com and click on
55:33 the chat button that's chatgpt4 and if
55:35 you hate both of those go to there and
55:37 you can play with six different chat GPT
55:39 models
55:41 um so those are all just variations on
55:42 chat GPT then prompts.chat will teach
55:45 you about prompting a little bit and
55:47 give you a bunch of examples of things
55:50 that you can do which I find inspiring
55:52 and gives me lots of ideas of how
55:55 [ __ ] deep that stuff is how powerful
55:57 it is and then futurepedia.io that's
55:59 just like just under 4 000 AI tools that
56:03 you can go play with and geek out on
56:05 okay
56:07 somebody said wow to something I showed
56:09 so awesome how would you use this to
56:12 find prospects in real estate in your
56:14 neighborhood There's real estate plugins
56:15 so if you sign up for the plugins you
56:18 can basically just turn on the real
56:20 estate
56:22 um plugins that work some are some are
56:25 local and some are National and then you
56:27 just start asking it questions and it
56:29 will fire off those plugins and go out
56:31 and do that or go to bing.com in fact
56:33 let's go to bing.com
56:35 uh because because wait what do you want
56:37 to do you want to find prospects in real
56:40 estate in your neighborhood I don't
56:42 quite know what I don't know know enough
56:44 about real estate what prospects means
56:46 but I assume prospective properties
56:48 but let's go to Bing and see how it is
56:53 I tend to find myself a little
56:55 disappointed with Bing because they it's
56:57 so lobotomized but let's see all right
57:00 so when you go to Bing it looks like
57:02 this looks like just like you know it
57:04 used to although it says ask me anything
57:06 now because it's now a chat bot but if
57:08 you go to the chat button
57:11 you get a full-on
57:14 um chat GPT window and I can I can say
57:17 more precise more balanced or more
57:18 creative I'm going to leave it on more
57:20 precise because we're doing
57:23 um analysis in my neighborhood so I'm
57:25 gonna say
57:27 um find me
57:30 um any
57:32 houses
57:34 um
57:37 for sale
57:39 and
57:42 80211 which is my zip code
57:46 any houses for sale in 80211 under
57:51 um one million dollars
57:56 and
57:57 um listed within the last week
58:02 let's see what this does
58:06 so there's my prompt so it's searching
58:08 for houses for sale in 80211 I found
58:11 some houses for sale
58:14 a three bed two bath house
58:17 3 3277 Julian for 9.49 with a link
58:22 let's go to the link and see if it's
58:23 actually a functional link sometimes
58:25 they're not
58:26 yep
58:27 let's see where is it it didn't go to
58:30 that specific one
58:32 but there's a bunch of things in my
58:34 neighborhood so that's kind of fun
58:40 um
58:42 and here's zillow.com so if I click on
58:45 that link is that going to take me yeah
58:47 so that's actually a search in my
58:48 neighborhood so there you go
58:51 um so yes
58:52 go to bing.com
58:56 I love doing I love those doing those
58:58 things live where I'm like I have no
58:59 [ __ ] idea if this is going to work or
59:00 not it worked
59:04 and again I mean the the difference for
59:06 me like if you think about like if you
59:08 know nothing about real estate and which
59:10 sites to go to and things like that
59:13 and if you went to Google and said I
59:15 want to find all this stuff like you get
59:17 like Google search results are so
59:19 polluted with SEO optimized websites and
59:24 just you know
59:26 stuff to get you to click that it's it's
59:28 actually really hard to find
59:31 what you want like you wanted a very
59:32 specific house in a very specific
59:34 neighborhood that just gave it to me
59:36 right it automatically jumped me to the
59:38 sites like it gave me the answer first
59:40 and then I can dig deeper and find the
59:42 sites rather than the other way around
59:43 here's 20 sites you go [ __ ] figure it
59:46 out right that's Google's attitude
59:48 here's 40 things we hope you click on
59:51 all of them because it makes us more
59:52 money
59:53 chat GPT and Bing are completely the
59:55 opposite right
59:57 you asked me a question I'll give you
59:59 the answer if you want to dig deeper
1:00:00 feel free
1:00:02 um what do you think will happen when
1:00:04 they host it on a quantum computer I
1:00:06 think I think the I think two things
1:00:08 will happen I think
1:00:10 they'll probably be able to update these
1:00:12 models more quickly right now they take
1:00:14 a long time
1:00:17 um
1:00:17 so that could be a thing and then you
1:00:19 know I would assume that they're gonna
1:00:21 they're gonna start doing some crazy ass
1:00:24 AI training Ai and AI developing more AI
1:00:28 so some nested
1:00:30 exponential AI
1:00:33 evolving AI [ __ ] where they'll probably
1:00:36 come up with with new technologies that
1:00:39 are significantly more sophisticated
1:00:41 than what we have right now like all of
1:00:43 the stuff that we're playing with right
1:00:45 now is remarkable as it is it's all kind
1:00:47 of based on a technology that Google
1:00:48 invented in 2017 this thing called the
1:00:51 Transformer which basically said you can
1:00:53 take this autocomplete feature and train
1:00:55 it on massive data sets and and and you
1:00:59 know on on massive amounts of gpus and
1:01:03 and make these incredibly powerful tools
1:01:05 right but it's all the stuff that we're
1:01:09 playing with right now is is some
1:01:10 derivative of that all this sort of
1:01:12 public stuff that we're playing with
1:01:13 right now so I would think that what the
1:01:16 quantum computers are going to do at
1:01:18 some point is is change the nature of
1:01:21 how the thank you very much how the
1:01:23 machine learning uh is actually done in
1:01:26 the first place so but we'll see I don't
1:01:27 know I I have no like this stuff is
1:01:30 already so beyond like I I if you've
1:01:33 been playing with this stuff tell me
1:01:35 this isn't like we're entering a post
1:01:38 science fiction world
1:01:39 like I kind of feel like all the movies
1:01:42 of my childhood I'm like going check
1:01:44 yeah we can do that now oh yeah yeah
1:01:46 hal2000 that that exists right it just
1:01:49 does doesn't have the voice yet and
1:01:51 we're not in space but yeah check that
1:01:52 Technology's there like I just feel like
1:01:54 we're knocking off TV shows and and
1:01:57 science fiction movies you know one item
1:02:00 after another you know here's the
1:02:02 Holodeck oh okay we're close to that
1:02:06 it's just crazy it's crazy I tell you
1:02:09 it's crazy
1:02:11 I have a degree in acting if you
1:02:13 couldn't tell
1:02:16 free pizza for life
1:02:18 or your favorite haircut lasts three
1:02:21 months at a time what
1:02:23 I don't know what you're talking about
1:02:25 but I like pizza and I have bad hair all
1:02:28 right there's nothing wrong with knowing
1:02:31 how to do something totally agree
1:02:33 good evening everyone good evening Rick
1:02:36 thank you for that gummy bear
1:02:40 is that like a CBD gummy bear
1:02:44 most dangerous people that have it is
1:02:46 the government well listen the guy like
1:02:49 like here's the thing about this AI [ __ ]
1:02:51 this stuff's been around of decades the
1:02:54 government said this for decades you
1:02:56 don't think they've been using this ship
1:02:57 for decades
1:02:59 what's different now is that Chachi PT
1:03:02 gave the rest of us access to these
1:03:05 tools right
1:03:07 machine learning isn't new artificial
1:03:09 intelligence isn't new what's new is
1:03:12 popular culture having access to it
1:03:15 small businesses individuals having
1:03:18 access to it
1:03:19 you and me being able to be on a tick
1:03:22 tock live on a Tuesday night and go hey
1:03:24 could I have something actually not just
1:03:27 teach me how to
1:03:28 program in Python but do the programming
1:03:30 for me yep that exists
1:03:33 how long does it take 30 seconds
1:03:36 what if there's a bug it'll fix it
1:03:39 what if I don't know how to run it it'll
1:03:41 tell you how thank you for the Roses
1:03:44 I I mean
1:03:45 holy [ __ ] like that's what's new but
1:03:48 yeah sure the the government's dangerous
1:03:50 and they've got it but they've had it
1:03:52 for a while
1:03:54 um
1:03:55 a document came out today from the U.S
1:03:57 House of Representatives that said
1:03:59 Chachi PT this this website
1:04:02 is the official cat GPT of the U.S House
1:04:06 and they have to turn off
1:04:09 um prompt tracking you know tracking and
1:04:12 uh prompt history
1:04:13 but they but they can now use it it's
1:04:16 official they're not blocking it
1:04:19 uh there's plenty of reasons to want to
1:04:21 know how yeah a boom boom is built
1:04:23 that's fine
1:04:25 um and listen go use the open source
1:04:27 large language models they don't have
1:04:29 any safety restrictions on them if you
1:04:31 want to do that
1:04:33 um don't they want a reg to slow down
1:04:35 smaller businesses no and in fact if you
1:04:37 listen to what Sam Altman is saying to
1:04:40 to the governments out there is he's
1:04:42 saying that
1:04:44 um
1:04:45 he doesn't want to regulate below a
1:04:47 certain threshold so he's what basically
1:04:49 what he's saying is if you're a company
1:04:51 the size of Microsoft open AI Google
1:04:56 where you're building these you know
1:04:58 hundreds of billion parameter models
1:05:01 that that the core models those
1:05:04 companies should be regulated any one of
1:05:06 those companies that develops technology
1:05:08 significantly more powerful than that
1:05:11 should be licensed and then anyone under
1:05:13 that he's saying don't regulate them at
1:05:15 all right now and there's so much open
1:05:18 source stuff out there right now it's
1:05:20 not like this is this technology is
1:05:22 living within a handful of companies it
1:05:24 was until meta released their their
1:05:27 foundational models and the developer
1:05:29 Community was like [ __ ] it let's go
1:05:31 right so we're in a completely different
1:05:34 place right now than we were even six
1:05:35 months ago
1:05:37 um but but no they don't they don't want
1:05:40 to slow down small businesses trying to
1:05:42 figure this out
1:05:44 um and and you know Sam Walton and uh
1:05:47 the Google CEO and the Microsoft CEO and
1:05:50 the anthropic CEO are all pushing for
1:05:52 the same thing regulate when they're at
1:05:54 this size don't regulate below that
1:05:57 license if they get significantly more
1:05:59 powerful is is where they are right now
1:06:01 now listen
1:06:02 is that maybe what they're saying for TV
1:06:04 and then they're cutting different deals
1:06:06 in the back rooms possibly but you know
1:06:07 that like you know welcome to American
1:06:09 capitalism that's probably going on but
1:06:12 but
1:06:13 you know
1:06:15 I don't know that I trust Sam Altman but
1:06:17 I do feel like he's at least aware of
1:06:19 what the stakes are and he's trying to
1:06:21 do the right thing that's my sense of
1:06:22 him
1:06:23 so big corpse will dominate it I don't
1:06:25 know that that's true in fact if you go
1:06:28 to again I think it's so easy to get
1:06:30 cynical about this stuff but I wouldn't
1:06:32 do it there's
1:06:34 there is a chance here that this doesn't
1:06:36 just go down the [ __ ] within a week
1:06:39 within a year oh [ __ ] I just I lost my
1:06:42 place
1:06:43 um
1:06:47 someone just said make a bot I'll make a
1:06:49 bot I'll show you that in a second I'm
1:06:51 just trying to find my
1:06:53 spot where I was before Tick Tock keeps
1:06:55 jumping me to the bottom of the comments
1:06:57 and I try not to skip any comments so
1:07:00 part of your experience with me is
1:07:02 watching an old man trying to find text
1:07:04 in tick tock
1:07:05 hahaha
1:07:09 um I can beat up AI I bet you could [ __ ]
1:07:12 AI man slap it in the [ __ ] face
1:07:15 needs to be put in its place
1:07:20 I don't know where I was click bait I
1:07:23 don't know what click bait was
1:07:25 what does docs mean so docs means so the
1:07:28 the blockchain and crypto Bros
1:07:31 when they talk about doxing it's when
1:07:32 you when you reveal your real identity
1:07:34 or when you sign up for open AI they ask
1:07:37 for your phone number
1:07:38 and like the the nft guys are like I'm
1:07:41 not gonna give them my my real
1:07:43 information because
1:07:45 even if I'm not nefarious I want the
1:07:47 freedom to be nefarious if I want to so
1:07:49 they they don't want anyone knowing
1:07:51 their info so when you sign up there you
1:07:54 have to give them your phone number but
1:07:55 you can go to bing.com and not have to
1:07:57 do that
1:07:58 um anyway the the comment I was trying
1:08:00 to go back to it the comment about so
1:08:02 the big companies are going to control
1:08:03 this I don't think they're going to
1:08:04 because
1:08:07 um
1:08:10 because meta released their their
1:08:12 foundational models like they released
1:08:14 the court What's called the weights of
1:08:16 the llm so it's the core models open AI
1:08:20 hasn't done that it's called open AI but
1:08:22 it's not open they're core models they
1:08:24 still have behind the firewall Google
1:08:26 still has their core stuff behind the
1:08:28 firewall meta put some of their weaker
1:08:31 models out there the development
1:08:32 Community has pounced on it so there's
1:08:34 now this thriving open source movement
1:08:38 for these large language models that are
1:08:40 getting very very quickly as powerful as
1:08:43 open Ai and Google
1:08:46 you can go look up there's there's a a
1:08:49 leak from Google from about a month ago
1:08:53 um it was an engineer at Google and the
1:08:55 the comment or the the title of the
1:08:57 piece was we have no moat
1:09:00 and moat being the defensive barrier
1:09:03 between competition what the engineer
1:09:05 was basically saying is that the open
1:09:06 source Community for a hundred dollars
1:09:09 is getting these smaller large language
1:09:12 models to the same level of performance
1:09:14 as what took Google six months and 10
1:09:17 million dollars to build
1:09:19 so for 100 bucks over here they're
1:09:20 getting to almost the same performance
1:09:22 as what Google does for 10 million
1:09:23 dollars so I don't think it's a foregone
1:09:26 conclusion that the big companies are
1:09:28 going to
1:09:29 dominate now
1:09:31 if the quality of the machine learning
1:09:35 models gets dramatically higher and it
1:09:37 requires you know tens of thousands of
1:09:40 gpus and you know years to do and that
1:09:44 becomes a competitive barrier than
1:09:46 possibly but open source right now is
1:09:49 very very powerful
1:09:51 can I do some AI image if Dan Aykroyd in
1:09:53 ghost Ghost in the Shell Busters that's
1:09:56 pretty good uh I'll do that in a sec
1:09:58 um what are some practical ways to
1:10:00 integrate AI into specific content like
1:10:03 customer support I think there's all
1:10:04 sorts of things customer support can do
1:10:08 one is every customer support
1:10:11 every customer support agent could have
1:10:14 access to all the data right so so
1:10:17 rather than having to do kind of logic
1:10:19 trees
1:10:21 um
1:10:22 you could be dynamically
1:10:24 transcribing what the person on the
1:10:27 phone is saying
1:10:28 dynamically searching all of the content
1:10:31 and bringing up the most appropriate
1:10:33 answer kind of in real time
1:10:36 um again
1:10:37 ai's been around for a while that I I
1:10:40 would be surprised if systems like that
1:10:41 didn't already exist
1:10:44 um
1:10:45 but that's that's one way
1:10:48 um
1:10:49 you know you you could also do things
1:10:51 like
1:10:52 you could take the transcription of
1:10:54 someone talking on the phone and
1:10:56 actually
1:10:57 instead of having an agent type in the
1:11:00 notes from that call
1:11:02 AI could be transcribing and and doing
1:11:04 notes for that call in real time so that
1:11:06 everyone that talks to that person knows
1:11:08 exactly what's been going on in that
1:11:10 conversation like wouldn't that be nice
1:11:12 if after spending you know 45 minutes on
1:11:16 hold the second time the person that got
1:11:18 on the phone actually knew what the [ __ ]
1:11:20 you talked about for the last you know
1:11:21 20 minutes with the previous clueless
1:11:23 agent
1:11:25 um where I go with customer support
1:11:27 ideally is you get rid of call centers
1:11:29 all together
1:11:31 hold
1:11:33 you replace low-level call centers like
1:11:36 the the phone tree call centers that are
1:11:38 you know out of India and places like
1:11:39 that
1:11:40 with a virtual customer support agent
1:11:44 that knows everything
1:11:47 and so can talk to you it could even do
1:11:50 things like understand that oh you're
1:11:52 calling from the south of the United
1:11:54 States so I'm going to have an agent
1:11:55 that answers the phone that talks like
1:11:57 you talk
1:11:59 um that's totally possible that that
1:12:02 virtual agent can understand what the
1:12:04 hell the person's talking about and and
1:12:05 have the authority to help them in all
1:12:07 sorts of ways but it would also be smart
1:12:09 enough to say if this person actually
1:12:11 needs to talk to a real human being I
1:12:13 can escalate that like that so you
1:12:15 retrain your low-level customer support
1:12:17 people to be high level customer support
1:12:20 people that only deal with the anomalies
1:12:22 and you let
1:12:24 um the AI system deal with 95 of the
1:12:27 calls no one has to wait online it all
1:12:29 just goes immediately into a system
1:12:31 there's no press one for this and two
1:12:33 for that and it just you call someone
1:12:36 picks up the phone they sound like you
1:12:38 sound
1:12:39 they know what the [ __ ] they're talking
1:12:41 about and if you indeed have a problem
1:12:43 that they can't solve they'll escalate
1:12:45 it to a human that's what I'd like to
1:12:47 see I think that's a business to build
1:12:49 by the way
1:12:50 if if you want to make money with chat
1:12:52 GPT
1:12:53 that's a way you could do it
1:12:55 um I'm grabbing that info tonight great
1:12:57 I don't know what that is code
1:12:59 interpreter would be great for
1:13:00 predicting the spread of infectious
1:13:02 diseases absolutely any sort of data set
1:13:06 uh I'm the one last night who just tried
1:13:08 for the first time last night it came up
1:13:10 with great stuff for something
1:13:13 um it's encouraging I may be able to get
1:13:16 my head around this since as you say
1:13:18 it's still so early yet as as as over
1:13:21 hyped as this [ __ ] is I know it feels
1:13:24 like and if
1:13:26 it just I feel it every day like I feel
1:13:30 like I'm way behind on this stuff I've
1:13:32 only been dealing you know into this
1:13:34 stuff for about a year
1:13:35 I've known about it for a while but it's
1:13:37 always been just like it's been kind of
1:13:39 the purview of researchers and
1:13:41 scientists and engineers
1:13:43 and then chat GPT and and you know to a
1:13:46 large degree mid-journey and stable
1:13:48 diffusion the image generation tools
1:13:50 all of a sudden made these powerful
1:13:52 tools you didn't have to be an engineer
1:13:55 to get it right and that's that that's
1:13:57 the inflection point that we're at
1:13:59 um
1:14:00 so so the the the the kind of date that
1:14:04 that sort of where I put my flag in the
1:14:06 ground is November 30th 2022. that's the
1:14:08 day that chat GPT launches that thing
1:14:11 right there and that for me marks the
1:14:13 thing where now these tools are
1:14:15 available to the rest of us
1:14:17 and 100 million people in six weeks went
1:14:20 there and signed up and many more since
1:14:23 then all right Microsoft Bing can't even
1:14:25 find stuff on their own servers I know
1:14:27 don't get me wrong I still hate
1:14:29 Microsoft I'm a I'm an apple Fanboy from
1:14:31 way back so I've been a Microsoft Basher
1:14:33 for 35 years
1:14:35 um so I'm right there with you but what
1:14:38 they've done with open Ai and chat GPT
1:14:40 and how they [ __ ] with Google is
1:14:42 impressive and fun so I'm giving them a
1:14:46 pass for this year
1:14:49 I'll go back to hating them when Apple
1:14:51 gets their [ __ ] together
1:14:53 I'll tell you what
1:14:54 when Apple jumps into the AI game
1:14:56 they're gonna [ __ ] murder it they're
1:14:59 gonna they're gonna
1:15:01 they're gonna be it's gonna be
1:15:03 insane
1:15:05 they're gonna be amazing they're also
1:15:07 what what Apple's gonna have that I
1:15:09 don't think anyone else can touch is the
1:15:11 privacy and safety piece they're going
1:15:13 to be up because they're building all of
1:15:15 this this machine learning
1:15:17 capability into their chips into their
1:15:19 native silicon so they're going to be
1:15:21 able to do all the AI processing locally
1:15:23 on your machine with without having to
1:15:25 send [ __ ] out
1:15:27 um and and they've already got all the
1:15:28 privacy and security stuff that their
1:15:30 whole ecosystem is is tightly locked
1:15:32 down
1:15:33 um so Apple's gonna do some amazing
1:15:36 stuff with this I think
1:15:38 I'm a copywriter does this mean I'll be
1:15:40 out of a job if you're a copywriter and
1:15:42 you're not jumping on this
1:15:46 your your job is absolutely a threat but
1:15:48 here's the deal
1:15:50 there's you should also as a copywriter
1:15:54 um
1:15:55 you should you should listen there's a
1:15:57 guy named Paul rateser he started this
1:15:59 thing called the marketing AI Institute
1:16:01 he's got a weekly podcast that you
1:16:03 should listen to
1:16:04 he's got this famous to me saying that
1:16:07 he says AI will not replace your job but
1:16:10 someone using AI will
1:16:12 as a copywriter if you're AI literate
1:16:15 and a copywriter you're going to be able
1:16:17 to use these tools better than non-copy
1:16:19 writers but if you're a copywriter
1:16:21 that's not using these tools you're
1:16:23 going to get blindsided it like it's
1:16:25 going to happen it's going to happen
1:16:26 fast for copywriters so yeah so I think
1:16:29 that right now tonight if you're still
1:16:31 here valley floor art you should be
1:16:33 going there right now and if if you've
1:16:37 never been there and haven't played with
1:16:38 it go to prompts.chat understand what
1:16:40 prompting's all about as a writer you'll
1:16:42 get it
1:16:44 you'll be good at it you'll have fun at
1:16:46 it and it's going to blow your [ __ ]
1:16:48 mind
1:16:52 it is
1:16:54 like an experiment would be
1:16:58 take take some copy that you've written
1:17:00 this week
1:17:01 and especially if it's something like
1:17:03 maybe you did a blog post or maybe you
1:17:05 did something that's like a decently
1:17:07 researched well-written piece that you
1:17:09 put a lot of time into
1:17:10 take that
1:17:12 go to chatgpt go here say based on this
1:17:17 piece of writing and then paste in your
1:17:20 piece of writing
1:17:21 write a new piece on this subject with
1:17:25 these four bullet points and do it in
1:17:28 the style and form of the thing that I
1:17:30 just pasted in there and just watch what
1:17:32 it does in 30 seconds in eight seconds
1:17:36 and it should blow your mind you should
1:17:37 so so here's the thing I feel like
1:17:39 everyone right now owes it to themselves
1:17:41 to go here and play long enough to have
1:17:44 this moment
1:17:47 you should go have your Kevin McAllister
1:17:49 moment from Home Alone
1:17:52 and
1:17:55 if it doesn't both excite you and scare
1:17:58 the [ __ ] out of you at the same time
1:18:00 then then you're not paying attention
1:18:02 then you haven't at this moment
1:18:05 um
1:18:07 so so is it going to take your job
1:18:10 I think you you have a hell of a lot
1:18:12 better chance of not being displaced if
1:18:14 you're AI literate
1:18:16 and it's going to absolutely transform
1:18:18 your job
1:18:20 but it's also going to absolutely
1:18:21 transform every job every knowledge
1:18:23 worker every creative professional all
1:18:27 of the all of the stuff we do on
1:18:29 computers right now
1:18:32 the the historical wisdom was that AI
1:18:37 would start with the lower level jobs
1:18:39 and the last ones to go would be the
1:18:41 knowledge workers and the creative
1:18:42 creative class right
1:18:45 they're the first ones that are going to
1:18:46 go because it does the it does tactical
1:18:49 execution really really well and
1:18:52 especially if you have someone that
1:18:54 knows what they're doing prompting these
1:18:55 machines
1:18:59 there's a
1:19:01 I was talking to a guy yesterday A buddy
1:19:03 of his just started an agency with five
1:19:05 other people so a six person person
1:19:07 agency
1:19:09 where they are turning around fully
1:19:12 executed Creative Concepts within eight
1:19:15 hours
1:19:16 so they're meeting with a client like in
1:19:18 the morning
1:19:20 and delivering that afternoon
1:19:22 fully executed Creative Concepts with
1:19:25 writing Graphics you know like like
1:19:30 stuff that would take an agency three
1:19:31 weeks to
1:19:33 six weeks depending on how big the
1:19:35 agency is they're turning it around in
1:19:36 the day same day
1:19:38 and they're starting an agency to do
1:19:40 that
1:19:41 now they don't have a bunch of
1:19:42 copywriters writing really fast
1:19:45 no they've got
1:19:47 creative directors that are
1:19:50 acting like conductors of a symphony
1:19:53 playing with all these tools
1:19:55 and they have enough life experience to
1:19:57 pull that [ __ ] together that quickly so
1:19:59 yeah
1:20:00 Love Your Entertainment thank you
1:20:02 I did story time for preschoolers in the
1:20:05 urban park with detailed activities it
1:20:07 was awesome oh that's super cool yeah
1:20:09 you can you use chat gbd to come up with
1:20:11 ideas for events it's great at that [ __ ]
1:20:13 because it's I mean why is it good at
1:20:16 that stuff because it was trained on all
1:20:17 of the internet so every time anyone has
1:20:20 done a blog post or or done a thing
1:20:22 about you know 10 activities for
1:20:24 preschoolers you can do in the park like
1:20:26 all that content's already out there
1:20:28 right so when you ask it to do stuff
1:20:30 like that it's like oh let me go look at
1:20:32 all of the internet how about these five
1:20:34 you're like [ __ ]
1:20:37 laughs
1:20:39 please y'all clap for the man all right
1:20:41 clap uh double tap the screen and do
1:20:44 stuff like that let people know we're
1:20:45 here by the way it looks like we got
1:20:46 some new folks my name is Kyle Shannon
1:20:48 this is the AI learning lab I can't
1:20:50 believe this many people come and hang
1:20:51 out here but I appreciate it
1:20:54 um
1:20:55 I'm an entrepreneur I'm the founder and
1:20:58 CEO of a company called storyvine we're
1:20:59 an 11 year old video technology company
1:21:01 a video storytelling company
1:21:04 um I was an innovator in the early days
1:21:06 of the world wide web and one of the
1:21:08 reasons I'm so passionate about this
1:21:10 [ __ ] is I was there at the beginning of
1:21:12 the World Wide Web coming up I was in
1:21:14 New York City I started one of the early
1:21:16 agencies in the mid 90s in New York City
1:21:19 and kind of rode the internet wave up up
1:21:22 and down and through
1:21:26 um
1:21:27 and
1:21:28 the energy of what's going on with
1:21:31 generative AI is very very very similar
1:21:34 to the early days of the web
1:21:37 but it's way more profound
1:21:39 and it's coming way faster and it's
1:21:43 I you know
1:21:46 I get I get
1:21:48 very philosophical with this stuff
1:21:50 because the World Wide Web
1:21:52 change the world right it fundamentally
1:21:54 changed the world
1:21:58 that's going to look quaint compared to
1:22:00 what this generative AI stuff's gonna do
1:22:04 it's gonna it's gonna transform
1:22:05 everything like even how I think about
1:22:08 how computers work and how programming
1:22:11 works and how product development works
1:22:14 and how writing works and how the
1:22:16 creative process works
1:22:18 and how creative products are
1:22:21 conceived
1:22:23 and curated and delivered all of it's
1:22:26 going to change all of it
1:22:28 all of it filmmaking at the highest
1:22:31 level down to the lowest level
1:22:33 um
1:22:35 at all of it
1:22:36 and so
1:22:38 I don't know
1:22:40 so I'm just here trying to get as many
1:22:42 people
1:22:43 as excited as I can not to listen I
1:22:46 don't give a [ __ ] if you drink the
1:22:48 Kool-Aid or not if you go play with this
1:22:50 and actually understand what it is and
1:22:52 how it works and then decide [ __ ] that
1:22:55 I actually like the process of sitting
1:22:57 at my typewriter and writing for eight
1:22:59 hours like for me that's
1:23:01 for me creativity is sitting at the
1:23:04 keyboard for hours at a time torturing
1:23:07 myself with words if if that's the thing
1:23:10 you enjoy
1:23:11 [ __ ] God bless you
1:23:15 but not everyone's like that and these
1:23:17 tools are like like this tool is like
1:23:20 having
1:23:21 by far the most powerful writing
1:23:24 assistant you've ever had like the most
1:23:26 powerful writing assistant in the in the
1:23:29 history of humanity
1:23:31 is now a click away like if if you still
1:23:34 haven't gone there if you're here at
1:23:36 this channel listening to this fat old
1:23:38 58 year old Gen X
1:23:40 [ __ ]
1:23:42 if you haven't gone here yet go there
1:23:44 right now and start playing with it
1:23:47 and don't think of it like oh it's like
1:23:49 fancy Google or don't think of it like
1:23:51 oh but I heard it hallucinates and it
1:23:53 lies and it does this and the robots are
1:23:56 gonna kill us don't be in that cynical
1:23:58 place right now just go there and start
1:23:59 playing with it learn it well enough to
1:24:03 actually understand
1:24:07 what it makes possible
1:24:10 and then kind of aim it at what you do
1:24:14 and see how good it is at the [ __ ] you
1:24:18 do
1:24:19 and even if it's only 80 is good
1:24:22 or 60 it'll do that 80 in 30 seconds
1:24:27 and then you can still spend the next 20
1:24:29 you know two hours refining it
1:24:31 but you owe it to yourself to go there
1:24:33 and do that and I know I'm kind of soap
1:24:35 boxy but we just oh we even have more
1:24:37 people now hey everybody Welcome uh I'll
1:24:40 get off the soap box now anyway if you
1:24:42 have
1:24:43 um
1:24:44 questions pop them in the comments below
1:24:46 I do try to answer them all I'm more
1:24:48 than 99 messages behind so if it seems
1:24:50 like I'm not answering you that's just
1:24:52 because I'm a jerk I'm a big fat poo poo
1:24:55 head prompt engineering meaning
1:24:57 will expand over time it will come to
1:24:59 mean interfacing with AI models yeah and
1:25:02 I think so so again because we're so
1:25:04 early in this generative AI phase
1:25:07 right now all of these tools are all
1:25:10 these disparate
1:25:11 sort of pieces of [ __ ] software that are
1:25:14 sitting on top of these profoundly
1:25:15 powerful models so the output of what
1:25:18 all of these tools are doing is
1:25:19 profoundly powerful but the interface
1:25:21 itself sucks it just sucks right like
1:25:25 like in chat gbt right now it gives me a
1:25:28 history of all my prompts but I can't
1:25:30 search them like if I want to go find
1:25:32 what I'm one of my prompts I can't right
1:25:34 I've got to just scroll and scroll and
1:25:35 scroll
1:25:36 and and they're also
1:25:39 um they're not fully integrated in the
1:25:41 tools we use but but they're going to be
1:25:42 if so if you look at generative fill
1:25:44 within Photoshop
1:25:46 they're starting to sneak these
1:25:47 generative generative tools within the
1:25:50 the tools we use they're going to be in
1:25:51 office 365. they're going to be in the
1:25:53 Google office suite
1:25:55 um they're going to be in everything
1:25:57 um they're going to be in operating
1:25:58 systems right so so so this stuff is
1:26:00 coming so prompt engineering isn't going
1:26:03 to look a year from now what it looks
1:26:05 like today but the idea of understanding
1:26:08 oh
1:26:09 what I need to do as a human
1:26:12 is know what I want enough to ask the
1:26:15 machine to go do that for me right and
1:26:18 right now that requires a little bit of
1:26:20 you know
1:26:21 critical thinking and Engineering
1:26:23 thinking but that'll get easier over
1:26:26 time but but that's that's the skills to
1:26:28 start to get curious about this start to
1:26:31 explore it and and start to understand
1:26:33 that you can bend this tool to your will
1:26:36 Microsoft calls the the gpt4 that
1:26:40 they're incorporating into their tools
1:26:42 co-pilot right so GitHub co-pilot is
1:26:45 owned by Microsoft I think it's a
1:26:47 brilliant name because that's that's
1:26:49 really what the relationship is you're
1:26:51 the pilot and this AI tool is your
1:26:53 co-pilot so it's it's like having a it's
1:26:56 like having an incredibly smart intern
1:27:01 um who needs a lot of guidance so you
1:27:03 guide it and then you say by the way do
1:27:06 you know anything about the Mesopotamian
1:27:07 War and it will it knows it all
1:27:10 by the way can you speak Babylonian yep
1:27:14 oh
1:27:15 can you translate this Babylonian
1:27:17 scripture into English sure here you go
1:27:19 anything else you'd like me to do there
1:27:21 boss
1:27:22 uh yeah put it in a table and compare it
1:27:26 to the King James Bible version version
1:27:28 of the same passage sure
1:27:32 um yeah so that's what's coming use four
1:27:34 not 3.5 so yes I agree with that uh
1:27:38 tilde l
1:27:40 that's not a very creative name but I I
1:27:42 like the brevity uh can it write a
1:27:44 WordPress website yeah it like I don't
1:27:46 so
1:27:48 certainly if you can copy the the code
1:27:51 from a WordPress website and paste it in
1:27:53 there and say just rewrite this it can
1:27:54 do that for sure but it probably knows
1:27:57 enough to just write a WordPress website
1:28:02 um yes I would think so yes um Can AI
1:28:05 ask you questions for brainstorming yes
1:28:07 so so a really cool thing that you can
1:28:10 do with with chat GPT
1:28:13 you can say
1:28:16 I want you to act as
1:28:19 my brainstorming partner
1:28:21 and you're an expert at and whatever it
1:28:25 is if you're in the the I don't know
1:28:28 Remnant carpet business we're gonna do
1:28:31 some brainstorming and and you're an
1:28:34 expert brainstormer in the remnant
1:28:35 carpet business and what I want you to
1:28:39 do is
1:28:41 um I want you to help me brainstorm and
1:28:43 you're gonna ask me a series of
1:28:44 questions about my business
1:28:47 so that you can understand my business
1:28:49 and then once you understand my business
1:28:51 you're going to lead me in a
1:28:53 brainstorming exercise I don't want you
1:28:55 to write any questions yet but do you
1:28:57 understand and it'll say yes I
1:28:59 understand I'm going to be your
1:29:00 brainstorming partner and I love the
1:29:02 remnant cart carpet business and then
1:29:04 once I understand your business then I'm
1:29:07 going to lead you through a
1:29:08 brainstorming thing you're like okay
1:29:09 great now I want you to ask me one
1:29:11 question at a time and don't go on to
1:29:13 the next question until you're satisfied
1:29:17 with my answer to the previous and I'll
1:29:19 say okay great here's your first
1:29:21 question what's the name of your
1:29:23 business
1:29:24 and then it will just start asking you
1:29:26 questions and you answer it and the more
1:29:29 context that you give it
1:29:31 the better and better it will get so
1:29:33 that by the time it turns into a
1:29:35 brainstorming partner it won't be giving
1:29:37 you generic brainstorming exercises
1:29:40 it'll be giving you very specific ideas
1:29:43 for your business so absolutely
1:29:45 brilliant at that
1:29:48 is clutter Cloud plus worth using if
1:29:50 you're using gpt4 right now yeah let me
1:29:52 show you why okay
1:29:54 let's go over
1:29:57 to the internet
1:30:02 okay so if you go to chat gbt you go
1:30:05 here
1:30:07 you got your little chat window you got
1:30:09 GPT 3.5 and 4. it's connected to the
1:30:11 internet it's got plugins
1:30:14 you don't need Poe
1:30:18 if you've got access to this you really
1:30:20 don't need anything else
1:30:22 for right now like that that'll change
1:30:24 over time but right now like this is the
1:30:25 this is kind of the Cadillac experience
1:30:27 just in terms of the quality of the
1:30:30 model and how they've got it fine-tuned
1:30:32 and how it does what it does certainly
1:30:35 with writing
1:30:37 um what Poe has so Poe is a similar kind
1:30:40 of
1:30:41 um engine
1:30:42 where you've got
1:30:45 you know your your text entry at the
1:30:47 bottom and it answers questions but then
1:30:50 over here on the side oh they actually
1:30:51 changed the interface from yesterday
1:30:52 that's wild so this this is new here but
1:30:56 the these first
1:30:58 um one two three four five six these
1:31:00 first six things are different
1:31:04 um chat models right so Sage is from Poe
1:31:06 which is uh quora
1:31:09 um you've got gpt4 Claude plus and
1:31:12 Claude instant 100k this is the one I'm
1:31:14 gonna come back to these three are
1:31:17 subscription only so you so you have to
1:31:18 pay for this and then you got Claude
1:31:20 instant which is super fast and you've
1:31:22 got chat EBT so the Claude models are
1:31:25 from a company called anthropic which is
1:31:27 re really interesting company X openai
1:31:30 people
1:31:31 Claude instant and chat GPT are
1:31:33 essentially the same thing Claude plus
1:31:35 and gpt4 are essentially the same thing
1:31:38 again gpt4 is better in my opinion Sage
1:31:41 is okay not great but whatever
1:31:43 Claude instant 100K that 100K means it's
1:31:47 a hundred thousand token
1:31:49 context window meaning
1:31:51 you can put in up to 75 000 words
1:31:55 into your prompt so that's like an
1:31:58 entire novel
1:31:59 right so if I click on this
1:32:02 let me go I'm gonna go find something
1:32:03 let's go find
1:32:05 um
1:32:07 uh screen play
1:32:11 PDF free
1:32:15 um let's say Scarface
1:32:21 Scarface script is this actually going
1:32:24 to give it to me
1:32:26 where's the PDF
1:32:32 click to view and download View
1:32:48 yeah but I I don't think that's actually
1:32:51 just a PDF that's their stupid website
1:32:53 let me see if I can find a PDF of it
1:32:59 screenwriters Network I think all of
1:33:01 these so that this is this is one of my
1:33:04 this is one of my absolute frustrations
1:33:06 about
1:33:08 um yeah about Google right now is that
1:33:10 everything in Google is like an SEO
1:33:12 [ __ ] try to scam some money off you
1:33:15 kind of stuff I just want to find this
1:33:17 the the PDF
1:33:20 and today
1:33:24 scripts to screen
1:33:28 Scarface
1:33:30 I had script Scarface transcripts at
1:33:33 scripps.com let's see if this is it
1:33:36 PDF
1:33:40 okay I think
1:33:47 okay that's the actual PDF let's see if
1:33:49 this will actually work okay
1:33:51 so
1:34:00 is this it so it's not formatted
1:34:03 but let's go grab all of this
1:34:10 oh I can't huh
1:34:13 ah this is frustrating all right let me
1:34:16 go
1:34:17 let me go back to Claude so in here I've
1:34:20 got so all of the text that you see here
1:34:23 this this was 2001 A Space Odyssey so
1:34:28 I've got all of 2001 A Space Odyssey
1:34:31 paste it in here
1:34:33 so let me go
1:34:40 grab it
1:34:42 this is some other [ __ ]
1:34:45 hold
1:34:52 this was a bunch of LinkedIn profiles so
1:34:54 that was the other thing I did I pasted
1:34:55 in like six different LinkedIn profiles
1:34:58 and then started querying against it
1:35:01 and this is a big long transcript of a
1:35:04 meeting okay here's
1:35:07 okay title card
1:35:09 Africa three million years ago so that's
1:35:11 the beginning of 2001 A Space Odyssey
1:35:15 so I'm gonna go down here the bottom
1:35:19 and I reset the context here this is why
1:35:21 I can't search on it right now so I'm
1:35:22 going to repaste this in
1:35:25 all right so there's the end of 2001 A
1:35:29 Space Odyssey so I'm going to say copy
1:35:31 okay so now
1:35:33 in my little text box here I'm going to
1:35:35 paste the entire screenplay of 2001 A
1:35:39 Space Odyssey
1:35:41 oh and I didn't ask it a question so
1:35:43 it's actually going to answer something
1:35:44 here let's see let's see what it answers
1:35:45 I didn't I didn't actually give it a
1:35:47 prompt
1:35:51 still waiting let me give it a prompt um
1:35:55 who are the main
1:35:58 characters
1:35:59 and what happens to them question mark
1:36:06 um
1:36:13 summary of the key events Frank Poole
1:36:15 has accidentally killed Dave Bowman
1:36:17 realizes the onboard computer Hal is
1:36:19 malfunctioning and decides to disconnect
1:36:21 Hal's higher functions leaving Hal in a
1:36:24 childlike state
1:36:25 so it just read that entire screenplay
1:36:28 here's a summary of the main characters
1:36:31 let's see
1:36:32 um
1:36:36 um
1:36:37 what
1:36:40 um does how
1:36:43 do
1:36:45 to the crew
1:36:47 question mark
1:36:59 still waiting still waiting I'm sorry
1:37:02 Dave I'm afraid I can't do that so
1:37:04 anyway you get the basic idea so this is
1:37:06 so I pay for Poe and I pay for gpt4 and
1:37:11 this this is why
1:37:14 I do that so let's see what does Hal do
1:37:17 how malfunctions by falsely reporting
1:37:19 issues
1:37:20 while Frank Poole is doing an Eva to
1:37:23 replace the unit how causes the pod bay
1:37:25 doors to open depressurizing the ship
1:37:29 a pod crashes into Frank pool killing
1:37:31 him instantly this leaves Dave Bowman as
1:37:33 the only surviving crew member Dave
1:37:35 Bowman realizes how is malfunctioning
1:37:37 and disconnects him
1:37:39 so
1:37:40 pretty pretty amazing stuff
1:37:43 crazy right
1:37:45 so I know that was a long answer to a
1:37:47 short question but you know
1:37:49 I'm losing my voice too what's going on
1:37:51 here people
1:37:56 wow wow wow wow welcome everybody
1:38:00 all right
1:38:02 um does it have a voice to text it looks
1:38:04 painful sometimes watching your typing
1:38:06 it does but I'm too lazy
1:38:09 um
1:38:10 and
1:38:11 I don't know if you don't watch me type
1:38:13 don't don't don't be here it's like I'm
1:38:15 I'm just doing what I'm doing like all
1:38:18 of that requires me to do setup and work
1:38:20 and learn new key commands and I've got
1:38:23 80d
1:38:24 foreign
1:38:27 if I had that level of executive
1:38:29 function I probably couldn't do this
1:38:30 tick tock
1:38:33 I seriously think that I'll be out of a
1:38:35 job if I continue copywriting I'm very
1:38:37 worried um
1:38:38 so here's the deal
1:38:43 um
1:38:46 I think you'll be out of
1:38:47 a job if you continue copywriting the
1:38:50 way you've done it before
1:38:54 um but I think that
1:38:56 does the world is the world still going
1:38:58 to need copywriters yes
1:39:00 but they're just going to use different
1:39:01 tools so just like if you're a
1:39:03 copywriter and you used
1:39:05 um a typewriter and then this along
1:39:07 Along Came this thing called a word
1:39:08 processor and a big part of your skill
1:39:10 was you didn't make typos so you didn't
1:39:13 have to use white out and you were
1:39:14 better than other copywriters if you
1:39:16 were still married to that Paradigm of
1:39:17 well I've got to use a typewriter and
1:39:20 then these young whippersnappers are
1:39:22 coming in with their fancy computers and
1:39:23 they can you know copy and paste
1:39:26 documents and you know do spell check
1:39:28 and things like that you you'll be left
1:39:31 behind so these tools even though they
1:39:33 do a lot of writing
1:39:35 my experience with them is
1:39:38 they can get you to about 80 percent
1:39:40 really really quickly and then it's
1:39:43 still up to the human to curate and
1:39:45 refine and you know finish and
1:39:48 understand
1:39:49 what they want to produce so I think
1:39:51 it's just more like shifting your
1:39:52 mindset that the tools that you'll use
1:39:55 to do copywriting will be different
1:39:58 but if you just stick to how you do what
1:40:01 you do now
1:40:03 the competition that you will face is
1:40:05 someone will come along that can write
1:40:07 10 times faster
1:40:08 and you might argue that well it's ten
1:40:11 times faster but is it as good
1:40:13 at some point the boss won't care right
1:40:16 at some point the bus is going to go a
1:40:18 10x increase in productivity is worth
1:40:21 the quality dropping slightly so that's
1:40:23 the risk that you face if if you don't
1:40:26 embrace it
1:40:27 um but yeah it's it's scary one of the
1:40:29 so I started this thing called the AI
1:40:31 salon and I'll give you the
1:40:33 um oops not that hang on
1:40:39 so the AI Salon this is the URL for the
1:40:42 salon
1:40:43 right here we meet bi-weekly and
1:40:47 um and we have an online Discord we can
1:40:49 geek out and ask questions and hang out
1:40:50 with other people one of the very early
1:40:52 members of the salon she's one of the
1:40:54 organizers now is a copywriter
1:40:56 and when she started out she was
1:40:58 absolutely terrified
1:41:00 and
1:41:01 the more she got into it the more she
1:41:03 understood what chat CPT was was and how
1:41:06 it worked
1:41:07 she said holy [ __ ] this changes
1:41:08 everything about how I do my job but
1:41:10 she's like busier than she's ever been
1:41:12 largely because she's AI literate and
1:41:15 she's really efficient and she said it's
1:41:17 completely transformed how she does
1:41:19 copywriting but she's working more than
1:41:21 ever but it's completely different she
1:41:23 actually spoke at the salon a couple of
1:41:26 weeks back
1:41:27 um and and talked about her experience
1:41:29 doing that so
1:41:30 um you can if you go to the AI salon on
1:41:33 YouTube you can actually see that
1:41:35 there's a video of her talking about
1:41:36 being a copywriter and and uh
1:41:39 how different things are now so it's
1:41:42 just it's wild but I you know again
1:41:45 I I would not put your head in the sand
1:41:47 just because it's scary like both feel
1:41:50 the fear but also learn this enough to
1:41:52 understand how exciting it actually is
1:41:55 um
1:41:57 but I totally feel your pain I mean this
1:42:00 is this is this is existential kind of
1:42:02 stuff
1:42:03 this this is what why people are both
1:42:06 excited about it and terrified about it
1:42:07 because it's going to change a lot
1:42:10 um I can't get it to add up numbers
1:42:12 correctly consistently it keeps messing
1:42:14 up so I would say use the Wolfram Alpha
1:42:16 plug-in and have have Wolfram Alpha do
1:42:20 actual math the reason these large
1:42:22 language models are not good at things
1:42:24 like
1:42:26 um like uh like logical functions like
1:42:28 that it's just the nature of how they
1:42:30 work so
1:42:32 when they put all that stuff on screen
1:42:34 they're doing it one word at a time it's
1:42:35 a it's a prediction engine so
1:42:38 how it how it actually functions is is
1:42:41 almost identical you know when you're
1:42:42 typing in Google and it like finishes
1:42:44 your sentence for you like it like adds
1:42:46 those three words at the end of your
1:42:48 sentence
1:42:49 that's all it's doing but like on
1:42:51 steroids
1:42:52 so it's it's literally predicting one
1:42:55 word after another to write all of that
1:42:57 copy that it writes
1:42:59 so if it gets it wrong along the way
1:43:01 it's not going back and looking at what
1:43:03 it wrote and saying did I get that right
1:43:05 it doesn't have that function in there
1:43:06 the only function it has is predict all
1:43:08 the next words
1:43:10 It's So Sophisticated that it looks like
1:43:12 it's doing the right thing I like to
1:43:14 describe it as mansplaining as a service
1:43:16 that that it will confidently answer you
1:43:19 even if it's wrong
1:43:21 um like every good man in your life
1:43:25 it will answer you even if it's wrong
1:43:28 um
1:43:30 and because of that it will kind of do
1:43:33 math in a way that looks like it should
1:43:35 be right now
1:43:37 this might be something to try so I
1:43:40 heard Andre carpathi who's who's one of
1:43:42 the founders of or the lead engineer at
1:43:45 openai one of the inventors of all this
1:43:47 stuff
1:43:49 um he was talking to the Microsoft
1:43:50 developers conference and he was talking
1:43:52 about a mathematical equation and he
1:43:53 said
1:43:54 the way these large language models work
1:43:56 if you if you say hey here's this really
1:43:58 complicated mathematical equation go
1:44:00 solve that for me
1:44:02 it will give you an answer
1:44:05 why it gets it wrong sometimes is
1:44:08 it's been trained on all of the internet
1:44:10 right
1:44:12 so what do you have on all of the
1:44:13 internet well one of the things you have
1:44:16 on all of the internet is you have a
1:44:17 bunch of math professors
1:44:19 that put that equation in there and
1:44:22 solved it correctly
1:44:24 but what you also have on the Internet
1:44:26 is all of the students
1:44:28 that solves that equation and got it
1:44:30 wrong
1:44:31 and so if you just say hey chat GPT
1:44:33 solve this math equation it will kind of
1:44:36 look at all of its knowledge all of the
1:44:38 possible answers and it will kind of
1:44:40 amalgamate them together and give you an
1:44:42 answer and so it may mix in some student
1:44:44 bad answers with some Professor good
1:44:46 answers
1:44:47 so one of the things you can do is you
1:44:49 can say hey I want you to act as as a um
1:44:53 Stanford trained meth Professor that
1:44:58 specializes in this branch of
1:44:59 mathematics and then I want you to solve
1:45:02 this problem
1:45:03 what that very possibly will do is
1:45:06 narrow What's called the latent space or
1:45:09 narrow the kind of uh the region of
1:45:12 answers that is more likely to have
1:45:14 right answers in it so there's there's
1:45:15 probably ways that you can improve its
1:45:18 performance by narrowing the window of
1:45:20 what it's looking at so that might be
1:45:22 something worth trying but that's why it
1:45:23 [ __ ] up like that over the next two
1:45:25 years or so there will be increasing
1:45:28 layers built into these tools that do
1:45:30 things like fact and error checking so
1:45:33 that it won't be so Random right now
1:45:35 it's it's a pretty random mess again
1:45:36 we're early
1:45:38 Kentucky is regulating AI maybe buy a
1:45:41 property in Kentucky in five years the
1:45:43 only place to find work well or you know
1:45:47 Kentucky will get increasingly left
1:45:50 behind because
1:45:51 the the world the world is here's the
1:45:55 deal with technology you can't uninvent
1:45:58 it
1:46:00 you can't uninvent technology right you
1:46:03 can try to resist it
1:46:05 I you could make that same argument at
1:46:07 the turn of the last century when the
1:46:09 steam engine was out and you could say
1:46:11 well Kentucky's we're not going to be
1:46:14 going with the steam engine we like our
1:46:15 horses and our and our plows we're good
1:46:17 we're an agricultural Community we're
1:46:20 not going to go with that fancy steam
1:46:22 engine [ __ ]
1:46:23 it that's the same argument right you're
1:46:26 not going to uninvent this stuff so the
1:46:28 best thing that you can do is understand
1:46:31 it
1:46:33 and then you can choose not to use it
1:46:36 but if you just say hey we're going to
1:46:37 block it I think you're screwed
1:46:40 go look at if you if you want to see
1:46:42 something interesting Gregory
1:46:44 um go Google
1:46:45 um city of Boston
1:46:48 AI guidelines so the city of Boston
1:46:51 about a month ago put out a guideline
1:46:54 saying
1:46:55 if you're going to use generative AI
1:46:57 here's how to do it here's what it is
1:46:59 here's what it's good at here's what
1:47:00 it's not good at here's some use cases
1:47:03 for things you can do with it here's
1:47:04 some things you shouldn't do with it
1:47:06 here's a code of ethics about how to
1:47:08 think about it
1:47:09 I think that's a really important
1:47:10 document so rather than just saying Hey
1:47:13 [ __ ] science and and you know [ __ ] all
1:47:16 this highfalutin stuff and we're just
1:47:17 gonna put our heads in the sand
1:47:20 um this stuff's coming whether you want
1:47:21 it or not and it ain't going to be
1:47:23 uninvented I would strongly encourage
1:47:25 you to go go read those guidelines from
1:47:27 the city of Boston I think they're
1:47:29 they're quite good and I think they
1:47:31 could serve as a road map for for a lot
1:47:34 of us
1:47:37 um so you're basically saying sell my
1:47:38 Google stock I wouldn't listen Google is
1:47:41 absolutely behind on this stuff but they
1:47:43 invented it and they've got a lot of
1:47:45 money and a lot of really smart people a
1:47:48 lot of really smart people I would not
1:47:50 count them out but they are definitely
1:47:52 behind the eight ball with this with
1:47:53 this generated AI stuff Bard sucks right
1:47:56 now
1:47:57 um I can't get it to add numbers
1:47:58 correctly we answered that for someone
1:47:59 who's terrible at Discord navigation
1:48:01 instead of going to Discord can you have
1:48:05 I don't see the rest of that question oh
1:48:07 wait can you have
1:48:09 something else I don't know what it said
1:48:11 but
1:48:12 um can you can't do mid-journey outside
1:48:14 of Discord right now that will change
1:48:16 but right now it's in Discord only
1:48:20 um
1:48:21 and uh yeah that that sucks discord's
1:48:25 Discord sucks
1:48:27 I'm a chat GPT newbie will you tell me
1:48:30 what it's good for so if you're a newbie
1:48:33 go to this site prompts.chat
1:48:36 this will teach you about prompting but
1:48:39 what this really is is it's a giant long
1:48:41 list of personas that you can tell chat
1:48:45 GPT to be so go be a mathematician go be
1:48:49 an author go be a business coach go be a
1:48:53 data analyst and in
1:48:57 in setting chat GPT up as that persona
1:49:01 a will give you ideas about the things
1:49:03 it can do but it can do lots it could it
1:49:06 can create a marketing plan for you it
1:49:08 could create a meal plan for you if you
1:49:09 create a fitness plan for you
1:49:12 um it could write all your social media
1:49:14 content it could write your next book it
1:49:16 could help you edit your current book it
1:49:19 can code for you it can document code
1:49:22 that you've already written it can debug
1:49:24 code you've already written it can
1:49:25 refactor code you've already written
1:49:27 into a more modern language like the
1:49:30 depth of it so so here's the deal
1:49:32 imagine
1:49:34 so you know those really annoying
1:49:36 students in your class that just [ __ ]
1:49:39 aced everything they like studied
1:49:41 everything and learned [ __ ]
1:49:43 everything and every class they took
1:49:45 they were good at
1:49:46 imagine that student that learned all of
1:49:50 it
1:49:52 that could go read the entire internet
1:49:54 and just knew it all
1:49:57 they had zero social skills and and they
1:50:00 didn't know what to do with all this
1:50:01 knowledge but they knew all that [ __ ]
1:50:03 that's kind of like chat GPT
1:50:05 so would you say what is it good for
1:50:09 if you had the world's smartest intern
1:50:11 working for you that would do anything
1:50:14 you wanted like that
1:50:16 even though sometimes it would get
1:50:18 things wrong and sometimes it would be a
1:50:20 little stupid and sometimes you'd need
1:50:23 to guide it but it [ __ ] knew
1:50:25 everything
1:50:26 what would you do with that intern I
1:50:28 don't know you could do lots
1:50:30 you could start the world's most
1:50:31 successful lemonade stand you could
1:50:33 probably go solve some huge societal
1:50:37 problem you could probably cure some
1:50:40 horrible disease right so it can do lots
1:50:45 like almost anything
1:50:47 so what I would encourage you to do is
1:50:48 go there and just start playing and go
1:50:50 to prompts.chat this will give you lots
1:50:52 of ideas like it's literally hundreds of
1:50:55 different ideas of the kinds of things
1:50:57 this can do
1:50:59 and and you could take any one of those
1:51:02 so if you say to this thing you know act
1:51:04 like an expert I don't know philosopher
1:51:08 spend the evening talking to the world's
1:51:11 smartest philosopher
1:51:15 um that's what it's that's what it's
1:51:16 like
1:51:18 I've been using cheche upt all day to
1:51:20 help rewrite resumes and cover letters
1:51:22 outline a novel Trilogy so okay that's a
1:51:25 perfect experience with chat GPT I've
1:51:27 Rewritten my resumes and cover letters
1:51:29 okay that's done you know what I'm gonna
1:51:31 go outline my novel Trilogy exactly like
1:51:34 that's that's the experience like you
1:51:37 can go in with one thing and then you
1:51:38 can go you know
1:51:39 I wonder if it could help me with that
1:51:41 old
1:51:42 Trilogy that I never got around to
1:51:44 finishing the outline oh [ __ ] there's my
1:51:47 outline
1:51:49 I wonder if we could do the first
1:51:50 chapter oh that's not bad
1:51:53 that's it
1:51:56 um will it auto my taco truck we talked
1:51:58 about that
1:52:00 uh you don't need to ask questions many
1:52:02 many times I try to get to them although
1:52:03 I'm hundreds of questions behind
1:52:05 my workplace where I write I use AI to
1:52:08 help with insane project oh with this
1:52:11 insane project management tool that's
1:52:13 cool I can beat up AI congratulations I
1:52:16 think we all can it's a bit of a dweebie
1:52:18 geek right now
1:52:22 where can we find a career for this
1:52:26 um
1:52:28 so that's that's a that's a really
1:52:30 interesting question
1:52:33 um
1:52:38 these AI tools are going to be
1:52:40 ubiquitous they're going to be in
1:52:41 everything so the answer to that
1:52:44 question is is almost the same as if you
1:52:46 had asked in 1994 where could you find a
1:52:49 career in the world wide web
1:52:52 back in 1994 it was like learn HTML and
1:52:55 you can go build websites and maybe you
1:52:57 can start an agency or you could maybe
1:52:58 start a.com company
1:53:01 so it's kind of like that right now
1:53:02 where it's not super obvious where it's
1:53:05 going to be but if you think about where
1:53:06 could I get a job on the internet right
1:53:08 now it's almost a weird question you're
1:53:09 like well the internet's everywhere so
1:53:11 like any job you have is going to be
1:53:12 with the internet
1:53:14 it's going to be the same with AI for
1:53:16 the next two to three years being AI
1:53:19 literate is going to be novelty right so
1:53:21 so if you're a copywriter but you're an
1:53:24 AI literate copywriter you're gonna have
1:53:26 a hell of a lot better chance than
1:53:27 someone who's not AI literate so what I
1:53:29 would say back to you is
1:53:32 what do you want to do or what do you
1:53:33 currently do and then I would point you
1:53:35 to to this or if it's in the art field
1:53:38 something like mid Journey or stable
1:53:39 diffusion on the image side
1:53:42 and I would I would start to learn what
1:53:44 can this do as good as I do or better
1:53:47 than I do
1:53:48 figure out with this how you would
1:53:51 replace yourself
1:53:53 for whatever it is you want to do and in
1:53:55 figuring that out
1:53:58 you'll find the career forward
1:54:01 for lack of a better term and the other
1:54:03 thing that I really encourage people to
1:54:04 do is as you're learning about this
1:54:06 stuff it's why this channel exists start
1:54:08 teaching others as you learn something
1:54:10 just say hey I'm learning this chat gbt
1:54:13 thing I I know you were struggling with
1:54:15 your small business would you mind if I
1:54:17 like came up with some marketing ideas
1:54:19 for you using AI I don't care if it'll
1:54:24 help great
1:54:25 and then go up and try to help someone
1:54:28 solve their marketing problems using
1:54:29 that
1:54:30 that could be the beginning of a career
1:54:32 a bunch of people that I know like
1:54:35 literally started learning this stuff
1:54:37 put it in their LinkedIn profile and
1:54:40 within a month started working why
1:54:42 because we're so new everyone's trying
1:54:44 to figure this out right now so just go
1:54:46 start playing take what you know
1:54:48 try to replicate it with that and in
1:54:51 doing that you'll kind of discover oh it
1:54:53 does this really well maybe I could go
1:54:55 tell people how to do that yeah go do
1:54:57 tell 10 people for free and then the
1:54:59 11th one charge them 50 bucks and then
1:55:02 the 20th one charge them 500 bucks
1:55:05 that might be a career I don't know I
1:55:07 mean it's it's it's a it's a really
1:55:10 broad question if you're more technical
1:55:13 you can go on the technical side of this
1:55:15 but but we're past the inflection point
1:55:17 now but Chachi piece T has done is you
1:55:20 don't need to be a computer scientist to
1:55:22 use AI anymore
1:55:24 you just need to go to that website and
1:55:26 be able to type in words and in the
1:55:29 future you won't even need have to do
1:55:30 that you'll just be able to talk [ __ ]
1:55:32 and it'll just go do it
1:55:34 but right now you got to learn that [ __ ]
1:55:38 um
1:55:39 all right
1:55:44 can you get chat EBT to give you a
1:55:46 prompt template you can absolutely have
1:55:48 chat GPT teach you how to prompt it
1:55:52 you can say I want you to do this what's
1:55:54 what's a way that I can prompt you so
1:55:57 that you do this better and it will tell
1:55:58 you that
1:56:00 and then you can say could you just go
1:56:01 do that
1:56:02 oh
1:56:04 all right it's a gift and a curse
1:56:06 exactly it's a dessert topping and a
1:56:09 floor wax
1:56:10 uh
1:56:12 um yes me too wrote several new business
1:56:15 plans in three months all right I've
1:56:16 already seen all these questions
1:56:18 oh for some reason it jumped me way up
1:56:21 ahead sorry I mean you know what I'm
1:56:23 going to do I'm going to scroll all the
1:56:24 way to the bottom
1:56:26 so and then I'm just going to scroll up
1:56:27 a little bit I'm jumping a bunch of
1:56:29 questions just because I got lost
1:56:30 there's a Emilio's wife hello Emilio's
1:56:33 wife
1:56:36 okay I'm just gonna grab a random one
1:56:39 here hold please lots of conversation
1:56:42 amongst yourselves which I love to see
1:56:43 that's cool I'm using chat GPT to run
1:56:46 DMD campaigns I give it a prompt and it
1:56:48 creates a story I can play in mold
1:56:50 that's awesome yeah act as a as act as a
1:56:52 dungeon master and a lot of times I
1:56:54 think it'll even ask you like what you
1:56:56 know what version do you want to play
1:56:57 because it knows all the rules so you
1:57:00 can go back and play different versions
1:57:01 of DnD that you like
1:57:03 which is super cool AI will kill
1:57:05 everyone after Elites merge with silicon
1:57:08 because it will Doom the non-cyborgs
1:57:11 elite there's someone who's been
1:57:13 watching some movies yeah Hollywood's
1:57:15 been training us for 50 years that the
1:57:17 robots are going to kill us so that's a
1:57:19 very common
1:57:20 um
1:57:21 uh
1:57:23 idea but
1:57:24 I think there is there is an equal
1:57:27 possibility that the robots will free us
1:57:31 from the mundane and the soul-crushing
1:57:34 work that we [ __ ] hate doing imagine
1:57:36 if everyone could just do work that
1:57:38 inspired them we very well may be going
1:57:41 in that direction it's one of the funny
1:57:42 things I'll hear about people saying is
1:57:44 like well if the robots do all the work
1:57:46 what are people going to do I don't know
1:57:48 they're [ __ ] human beings they'll
1:57:50 start contributing in other ways they'll
1:57:52 realize oh maybe that job was crushing
1:57:55 my soul and I hated it you know what
1:57:57 I've always wanted to do huh and then
1:57:59 they'll go use that to figure out how to
1:58:01 do
1:58:02 and then
1:58:04 we enter a great Renaissance where
1:58:06 everyone is creatively expressed and all
1:58:08 of humanity loves one another and we
1:58:10 sing Shangri-La and dance in circles
1:58:14 in rainbows and silk ribbons fly through
1:58:16 the air
1:58:18 or not or the robots kill us
1:58:22 which way is it gonna go well I think
1:58:25 again why this channel exists I think
1:58:28 the more of us that step into this
1:58:31 and sort of bend it to the light
1:58:34 it will actually bend it to the light if
1:58:37 we all just sit back and stay cynical it
1:58:39 will bend to the dark
1:58:43 I added the plug-in functionality cool
1:58:45 if you discover anything cool and
1:58:47 interesting please share
1:58:49 it's so good with a tier I know like if
1:58:53 hopefully someone in here tonight has
1:58:55 had the Kevin McAllister moment where
1:58:58 it can do that
1:59:01 I've told this story before but
1:59:03 I was teaching a group of like I don't
1:59:05 know
1:59:06 10 10 to 12 year olds how to do like
1:59:08 image prompting and stuff like that and
1:59:10 it was a school for like gifted but you
1:59:13 know like ADHD you know not neuro spicy
1:59:16 kids right so they're like bouncing off
1:59:18 the walls teaching them how to do this
1:59:20 stuff
1:59:20 and at the end of like two hours this
1:59:22 one kid he'd been quiet for the whole
1:59:25 time I was teaching and he said he
1:59:28 looked at me and he goes you know
1:59:31 I'm writing an application in C sharp I
1:59:34 said really because yeah
1:59:36 yeah it's a clone of Flappy Birds but uh
1:59:39 I got a bug in the code so I'm just
1:59:41 gonna throw that out it's like 10. he
1:59:43 was like too cool for school I'm like oh
1:59:45 that's cool I said
1:59:46 have you have you
1:59:48 tried you know playing with that in chat
1:59:50 GPT yeah yeah yeah yeah I said here come
1:59:53 here let me show you something so so I
1:59:56 open up chat GPT and I I type in
1:59:59 um write a clone of Flappy Birds and C
2:00:01 sharp that can be run in unity because
2:00:03 he was running it immunity
2:00:05 and it wrote the code and he his mouth
2:00:07 literally did this he went
2:00:11 and he goes I'll be right back I gotta
2:00:13 get my laptop so he runs down four
2:00:15 flights of stairs runs back up four
2:00:18 flights of stairs he's my laptop
2:00:21 so we get out his code
2:00:23 his c-sharp code for his Flappy Birds
2:00:25 clone
2:00:26 and we copy it out of unity we paste it
2:00:30 into chat GPT and I say
2:00:33 um what's wrong with this code and it
2:00:35 found a typo there was like you know two
2:00:38 characters were swapped or there was
2:00:39 something missing
2:00:41 and it fixed it
2:00:43 and I said we'll copy it you know paste
2:00:45 it back into Unity so it copy copies it
2:00:47 out of chat gbt paste it into Unity hits
2:00:50 run and Flappy Bird starts playing and
2:00:53 he starts playing it and then he crashes
2:00:55 and he and he looks at me and he goes
2:00:57 this changes everything
2:01:01 that's the moment you want to have
2:01:03 that's the moment you want to have
2:01:05 that's why you go there
2:01:06 have that moment where you're like it
2:01:09 can do that
2:01:11 because the minute you have that and you
2:01:13 realize this isn't just a [ __ ] Google
2:01:15 search this is something more
2:01:17 then you start to think about well wait
2:01:19 a minute
2:01:20 if it can do that thank you if it can do
2:01:23 that then maybe it can do this and if it
2:01:24 can do that then maybe it can do this
2:01:26 and holy [ __ ] if it can do that maybe I
2:01:29 can do that
2:01:31 that's what this is about this is about
2:01:34 like Humanity
2:01:36 Amplified
2:01:38 amplifying your Humanity
2:01:41 but you got to get over the hurdle of
2:01:43 this is a limited piece of [ __ ] [ __ ]
2:01:46 the robots are gonna kill us I'm cynical
2:01:49 these AI nerds are dorks
2:01:52 Kentucky is going to shut it out
2:01:55 Kentucky will be the last place you can
2:01:57 get away from this stuff
2:02:01 um one of the bottom two links behind
2:02:04 you so prompts.chat is um
2:02:07 a document that will teach you about
2:02:09 prompting it'll also teach you that you
2:02:11 can sort of create personas that chat
2:02:13 gbt can act like and then future PD is a
2:02:16 directory of like almost 4 000 AI tools
2:02:19 they just keep adding to it it's so so
2:02:22 the thing about all these AI tools right
2:02:24 now so there's almost 4 000 tools at
2:02:26 Future pedia
2:02:28 um three thousand you know 700 of them
2:02:31 are complete abject pieces of [ __ ] and a
2:02:34 waste of your time and then there's like
2:02:36 300 in there that are [ __ ]
2:02:37 mind-blowing so you kind of have to sift
2:02:40 through them so I would say look at
2:02:41 ratings look at things like that but if
2:02:43 you want to geek out you can do that but
2:02:45 but
2:02:46 quite honestly
2:02:48 just playing that chat GPT and
2:02:51 especially now that it's got plugins
2:02:53 just learning how to do that and what
2:02:55 how deep it actually is
2:02:58 is I could spend the next decade there
2:03:00 and and not even probably scratch the
2:03:02 surface
2:03:05 I think the difference is the complexity
2:03:08 of human multitasking we don't think AI
2:03:12 can do that well
2:03:16 hmm
2:03:17 I think AI can probably do that way
2:03:20 better than humans can I I think humans
2:03:24 multitasking is a fallacy I think we do
2:03:26 fast
2:03:28 task switching but we're doing mono
2:03:31 tasking and we're just switching between
2:03:33 them
2:03:34 um the thing that I think humans
2:03:37 that AI can't touch right now is
2:03:39 critical thinking and like the wisdom
2:03:43 part like I feel like
2:03:45 Google gave us information on demand
2:03:48 generative AI chat GPT gives us
2:03:50 knowledge on demand but we've still got
2:03:52 the wisdom so we're still in charge
2:03:54 we're still saying here's what I want it
2:03:56 to do here's why I want it to do that
2:03:58 here's what good looks like here's what
2:04:00 great looks like so all of the critical
2:04:03 thinking
2:04:04 that's still the human piece the wisdom
2:04:06 piece
2:04:08 but the generative AI can do all the
2:04:10 Tactical execution not all of it a lot
2:04:12 of the Tactical execution and make it
2:04:14 way more efficient so then humans move
2:04:17 into more of a conductor role than a
2:04:19 solo violinistral right if you're a
2:04:22 copywriter you're a solo violinist right
2:04:24 if you're using multiple multimodal AI
2:04:27 tools that do images and music and words
2:04:30 and programming
2:04:33 and you understand hey I wanted to do
2:04:35 these things if you can think
2:04:36 horizontally across disciplines and use
2:04:39 these tools you're going to have
2:04:40 superpowers
2:04:41 super powers
2:04:45 no more negative self-talk with the
2:04:49 FN Gen X stuff
2:04:51 why
2:04:53 shame is my love language
2:04:56 so I can't put myself down
2:04:59 uh uh how will change accountants jobs
2:05:04 wait how how will this change
2:05:06 accountants jobs besides analysis
2:05:10 um
2:05:11 it's gonna be really good at accounting
2:05:13 if so if you haven't seen it go to
2:05:16 YouTube and look look for a uh do a
2:05:19 search for code interpreter chat GPT
2:05:22 that's a new mode that's coming for chat
2:05:25 GPT called code interpreter that will do
2:05:28 full-on data analysis so all of the
2:05:31 accounting stuff like chat GPT will be
2:05:33 able to know all of the accounting rules
2:05:35 and you'll be able to interact with
2:05:38 basically a virtual accountant that will
2:05:40 know all the rules so what I would say
2:05:42 is if you're an accountant get AI
2:05:44 literate start learning how it does that
2:05:48 stuff what is it good at what is it not
2:05:50 good at go to futurepedia.io and look up
2:05:53 any AI powered accounting tools there's
2:05:57 probably some in there
2:05:59 they may suck right now but what you'll
2:06:01 probably start to see is glimmers of
2:06:03 what's possible so I would say just get
2:06:06 up to speed as much as possible as much
2:06:08 as possible on what's coming go look at
2:06:10 code code interpreter and look at what
2:06:13 that's going to do
2:06:15 um and just start getting AI literate I
2:06:16 mean it's my message essentially for
2:06:19 everyone on this channel is get AI
2:06:21 literate go there start playing long
2:06:24 enough until you have this moment like
2:06:25 holy [ __ ] I didn't know it could do that
2:06:27 and then from there you can start to
2:06:29 make this some decisions
2:06:31 like like some people
2:06:35 are going to de-prioritize this stuff
2:06:37 after they learn what it is
2:06:39 they're going to say I actually Paul
2:06:41 rates are the guy from marketing AI
2:06:43 Institute he says
2:06:44 I actually like the process of writing
2:06:47 without that because it's how I process
2:06:50 things it's how I think so for him the
2:06:54 act of writing
2:06:56 is how he thinks right so he doesn't
2:06:59 need to use that for the act of writing
2:07:02 but he uses that for the active editing
2:07:04 and for the act of project management
2:07:06 for the act of sorting things and
2:07:08 sifting through things and he's using
2:07:10 this a ton but he's found a certain part
2:07:13 of who he is where he's like okay I'm
2:07:15 gonna not do that for this part of my
2:07:17 life but I'm going to use that in all
2:07:20 these other areas in my life right and
2:07:22 this is a guy that's deep into this
2:07:23 stuff and he's already figuring out
2:07:25 there's places where this is valuable to
2:07:27 me in places where it's not I think
2:07:29 we're all going to have to find our own
2:07:31 boundaries for that
2:07:34 make a calculator for my webs
2:07:36 for my website
2:07:38 and pro rata calculator you can do that
2:07:42 just go to chatgpt tell it you want to
2:07:45 make a calculator for your website and
2:07:48 have it teach you how to do that it'll
2:07:49 write the code and it will teach you how
2:07:51 to incorporate it
2:07:54 um interacting with AI is helping me to
2:07:56 ask better questions to humans too yeah
2:07:58 I listen I I it is training training me
2:08:02 to be a better person
2:08:04 like I'm I'm
2:08:06 if you haven't done this yet it's it's
2:08:08 fun
2:08:10 um
2:08:10 I find myself saying please and thank
2:08:12 you to it when it writes something that
2:08:14 I like I'm like great job
2:08:17 like I tell it it did a great job
2:08:20 um
2:08:20 and yeah there's there's something about
2:08:23 the nature of interacting with it that
2:08:26 like a new intern
2:08:31 it's completely clueless right it knows
2:08:34 everything but it's an idiot
2:08:38 and so what do you do with people that
2:08:39 are completely clueless you're like okay
2:08:41 well let me start at the beginning
2:08:43 so I work for a company that does this
2:08:46 I've got this particular kind of problem
2:08:49 I'm hoping that you can help me solve it
2:08:51 in this particular way is that something
2:08:53 that you can do oh sure I've done that
2:08:55 before okay great right and now the
2:08:58 conversation begins so just like you
2:09:00 would tell an intern okay you know I
2:09:02 like my coffee this way and I want you
2:09:04 to go down to the corner store and I
2:09:05 right
2:09:06 and if they're a really brilliant intern
2:09:08 you only have to tell them that once
2:09:11 and if they're dumb dumb you got to tell
2:09:13 them that once a week
2:09:15 you know it's kind of like you know
2:09:17 Super Genius dum dum
2:09:20 laughs
2:09:22 but but again start learning the
2:09:24 boundaries of what it's good at and what
2:09:26 it's not just get literate about it and
2:09:28 start talking to other people about it
2:09:29 and it will serve you well not only did
2:09:31 I use AI to get a new job and double my
2:09:33 salary go on
2:09:35 agility with Amanda this is awesome I
2:09:38 used it to get my product-based business
2:09:40 into seven new stores this month that's
2:09:43 huge agility with Amanda that's super
2:09:46 cool I mean yeah that
2:09:49 that
2:09:51 not only did I use AI to get a new job
2:09:54 and double my salary I used it to get my
2:09:57 product based business into seven new
2:09:58 stories this month
2:10:01 and if she's smart
2:10:03 she's telling people about that
2:10:05 and if they're smart they're going hey
2:10:08 Amanda could you teach me how to do that
2:10:11 and if she's smart she is
2:10:15 because I gotta tell you the more I
2:10:17 teach this stuff the more I learn like
2:10:19 like hearing stuff like that is like yes
2:10:22 [ __ ] yes
2:10:25 and all those stupid [ __ ] videos that
2:10:27 are like you can make money
2:10:30 [ __ ] you
2:10:32 how you make money with chat GPT is you
2:10:34 figure out what you're passionate about
2:10:35 what you're good at then you learn this
2:10:38 stuff and you learn okay oh that's scary
2:10:41 oh this is this is like standing on the
2:10:44 Big Cliff and getting ready to jump into
2:10:45 the water
2:10:49 yeah it's gonna feel like that like this
2:10:51 is absolutely like like
2:10:56 when you use this stuff we are
2:10:58 absolutely all going into new territory
2:11:01 where we've never been so if you feel
2:11:03 completely clueless and completely
2:11:05 screwed
2:11:07 that's okay
2:11:09 like it will help you out of that
2:11:12 but you kind of have to let go of
2:11:15 in a lot of ways your identity like I've
2:11:17 had to do this with my business where a
2:11:20 big part of my current business it's
2:11:21 called storyvine and it's an automated
2:11:23 video storytelling platform people
2:11:26 answer questions in our app it goes up
2:11:28 to the cloud five minutes later there's
2:11:29 this fully edited video
2:11:32 the automation part of what we do we
2:11:34 call it auto magic editing which you
2:11:36 know I love that line people love that
2:11:38 line they remember that line a year from
2:11:41 now nobody's gonna give a [ __ ] about
2:11:42 automation because everything's gonna be
2:11:44 automated so I'm sitting here with a
2:11:47 business I've spent 11 years building
2:11:49 and I'm looking at one of our core value
2:11:51 propositions and I'm imagining a year
2:11:53 from now
2:11:54 that value proposition goes from novel
2:11:57 and interesting to yeah whatever
2:12:00 holy [ __ ] what am I supposed to do with
2:12:03 that
2:12:05 so I think everyone that starts to get
2:12:07 their head around this stuff faces that
2:12:09 existential moment where they're like oh
2:12:11 this might affect me
2:12:14 and then if you can get through that if
2:12:16 you can flip to the other side of that
2:12:18 and just start to look at what it's good
2:12:21 at what it's not good at what it makes
2:12:22 possible
2:12:24 then you can do what Amanda did that's
2:12:26 really [ __ ] cool really cool
2:12:28 stop being a j and invite me please okay
2:12:31 crypto size are you
2:12:34 here you go you're in well the little
2:12:37 dots are spinning
2:12:42 you there
2:12:45 wait hang on
2:12:47 are you there crypto
2:12:50 hello
2:12:52 I can't hear you
2:12:54 I stopped being a j oh it looks like my
2:12:56 thing's slowing down
2:12:59 I invited you up here but I can't hear
2:13:00 you
2:13:02 hello
2:13:04 what's that
2:13:07 I can't hear you
2:13:12 can't hear you buddy
2:13:14 all right
2:13:16 I tried
2:13:17 all right
2:13:19 um I get totally lost and can't find
2:13:21 where I write prompts half the time ADHD
2:13:23 sucks I know but ADHD is also like the
2:13:26 superpower it's both it sucks and it's a
2:13:28 superpower but yes our frontal lobe is
2:13:31 like a block of cheese
2:13:32 mine is anyway uh can you post the
2:13:36 Boston dock shirt let me go show you the
2:13:38 Boston dock
2:13:40 this was from ages ago but
2:13:45 all right oh apparently I need to stop
2:13:47 touching my screen that's pretty ugly
2:13:49 all right we're gonna go to the Google
2:13:53 or the Bing and we're gonna go Boston
2:13:55 wait uh
2:13:57 of Boston
2:14:02 AI
2:14:04 policy
2:14:09 so it's at boston.gov so just search for
2:14:12 city of Boston guidelines for using
2:14:15 generative AI so it's so it's an eight
2:14:17 page document published in May of 2023
2:14:22 there's your purpose
2:14:24 some sample use cases so you can use it
2:14:27 to write a memo you can write a job
2:14:29 description
2:14:30 um here's some principles empowerment
2:14:34 inclusion and respect transparency and
2:14:36 accountability Innovation and risk
2:14:39 management privacy and security public
2:14:41 purpose guidelines fact check and review
2:14:45 all content generated with AI
2:14:47 especially if it will be used in Public
2:14:49 Communication or decision making so like
2:14:52 they're they're describing right here
2:14:54 okay these tools hallucinate and will
2:14:57 lie to you so [ __ ] check your work
2:14:59 right
2:15:01 pretty straightforward
2:15:02 here's open ai's GPT 3.5 here's Google's
2:15:06 Bard right what's good what's bad don't
2:15:08 share sensitive information do's and
2:15:10 don'ts right really good document
2:15:13 really concise I think it's eight pages
2:15:18 um every city in the country should have
2:15:19 one of these the country should have one
2:15:21 of these every state should have one of
2:15:23 these every school district should have
2:15:25 one of these
2:15:26 every company should have one of these
2:15:28 if you're learning about this AI stuff
2:15:32 you could okay here's okay ah here you
2:15:35 wanna have some fun
2:15:37 this is exciting okay
2:15:38 let's have some fun
2:15:41 go find that document
2:15:45 city of Boston Boston interim guidelines
2:15:49 for using generative AI
2:15:51 go find that document
2:15:55 in in chat EBT type in based on this
2:15:59 document
2:16:01 and paste the text of that
2:16:03 rewrite this for the city of
2:16:07 New York or for my school district and
2:16:11 give it a little bit of information
2:16:12 about your school district and tell it a
2:16:14 little bit about how you want to modify
2:16:15 it
2:16:17 you could generate one of these for your
2:16:18 company for your school district for
2:16:20 your town for your city
2:16:22 for your company for your division
2:16:26 do that tonight
2:16:28 it's really good at doing that [ __ ] take
2:16:30 that rewrite it and then just modify it
2:16:33 any way you want
2:16:34 that would be cool wouldn't that be cool
2:16:36 so we have we have we have 79 people in
2:16:38 here right now so imagine if half of us
2:16:40 took that and went to our bosses or the
2:16:44 powers that be or we are the powers that
2:16:46 be and we put one of those forward and
2:16:49 say Hey what if we did something like
2:16:50 this
2:16:51 that would be cool
2:16:53 uh can you move aside or pause to allow
2:16:56 for a screenshot sure
2:17:00 there you go
2:17:06 I've been at this for two hours my voice
2:17:08 is crispier than thought
2:17:11 I'm sleepy Kyle but I have fomo health
2:17:14 sorry Amelia's life I I have no good
2:17:18 answer for you every day is fomo day
2:17:21 oh my God it really is
2:17:24 um yeah
2:17:25 I haven't been particularly uh uh you
2:17:29 know funny tonight I've been a bit more
2:17:31 soap boxy but whatever
2:17:33 for the technical side of the Mila
2:17:36 m-i-l-a in Quebec is really good oh for
2:17:39 the technical side the m-i-l-a in Quebec
2:17:42 let me see what that is
2:17:44 um okay back
2:17:48 m-i-l-a
2:17:53 Quebec artificial intelligence Institute
2:17:56 inspiring the development oh that's cool
2:18:01 hey
2:18:02 all right I'll check that out that's
2:18:04 pretty cool
2:18:05 inspiring the development of AI for the
2:18:07 benefit of all oh cool located in the
2:18:09 heart of Quebec's AI ecosystem Melia is
2:18:12 or Mila is a community of more than a
2:18:15 thousand researchers specializing in
2:18:16 machine learning and dedicated to
2:18:18 Scientific excellence and Innovation
2:18:19 super cool that's great
2:18:22 um
2:18:24 what have you always wanted to do Mr
2:18:26 Shannon well
2:18:27 um
2:18:28 I'm doing it to a great degree I you
2:18:30 know
2:18:33 um when I first went to college I went
2:18:36 for computer science and then
2:18:38 um what I knew about computers was like
2:18:40 you know the Apple II and video games
2:18:43 and then I went to Penn State for
2:18:45 computer science and I walked in a room
2:18:46 into a room full of IBM mainframes and
2:18:49 they had to be a stack of Punch Cards
2:18:52 and they said you know
2:18:54 your life is going to be generating you
2:18:56 know
2:18:57 reports for big corporations and my
2:19:00 stomach turned and I'm like the [ __ ] it
2:19:02 is I'm not gonna that's not gonna be my
2:19:04 life so I switched my major to theater
2:19:06 and I got a degree in acting but I've
2:19:07 always had this sort of
2:19:10 um
2:19:11 curiosity about technology and
2:19:13 interesting ways to use it combined with
2:19:17 um
2:19:18 like this storytelling core
2:19:22 so I started and ran a theater company
2:19:25 um for four and a half years in New York
2:19:27 City then I wrote a bunch of screenplays
2:19:28 then I started an online magazine then I
2:19:31 started an online agency
2:19:33 then I went and did a bunch of
2:19:35 um film Marketing in Hollywood and then
2:19:38 I started storyvine which is a
2:19:41 storytelling technology company so I'm
2:19:44 kind of doing it I mean it's it's you
2:19:46 know some like for me with my ADD
2:19:50 I if if I'm if I don't give a [ __ ] about
2:19:53 it I can't do it I just can't like I
2:19:56 just can't it's like I'm
2:19:57 constitutionally unable I yeah I just
2:20:02 you know
2:20:03 so I have just I guess Gotten Good at
2:20:07 finding ways to make money at combining
2:20:11 technology and storytelling so
2:20:14 in a lot of ways I'm doing what I want
2:20:16 to do but I'm you know I would like for
2:20:19 storyline to be farther than it is and
2:20:21 I'd like to be Farther Along on
2:20:22 understanding this AI stuff like I feel
2:20:25 even though I do this Channel and I talk
2:20:27 like I know what the [ __ ] I'm talking
2:20:29 about I'm clueless
2:20:30 I'm clueless there's so much to learn
2:20:33 there's so much possibility here
2:20:35 like all the questions about what's my
2:20:37 job going to look like what's the future
2:20:39 going to look like what's business going
2:20:40 to look like I don't have a [ __ ] clue
2:20:41 I know it's going to be different
2:20:44 I know it's going to be way different
2:20:50 uh move a bit sideways so I can see your
2:20:52 background there you go neuro spicy I
2:20:54 love it thank you
2:20:55 I stole that from who did I steal that
2:20:58 from well there's a tick tocker here
2:20:59 that talks about it
2:21:01 and then uh who else talks about it
2:21:03 what's his name um David Shapiro on
2:21:06 YouTube he's an AI he's a brilliant AI
2:21:08 guy working on AGI artificial general
2:21:11 intelligence he's brilliant this is the
2:21:14 ultimate cheat code rod
2:21:16 this [ __ ] is the ultimate cheat code for
2:21:19 everything it's a that's actually a
2:21:20 great way to think about chat GPT chat
2:21:23 GPT is the ultimate cheat code for
2:21:25 [ __ ] everything for everything it's
2:21:28 the ultimate cheat code it's that's the
2:21:30 perfect description
2:21:32 um that neuro spicy that neuro atypical
2:21:34 kid lost his motivation to learn to code
2:21:36 after an encounter with AI oh that's
2:21:38 interesting yeah well so
2:21:41 so here's the thing that kid might have
2:21:44 lost
2:21:45 his motivation to learn to code
2:21:48 but
2:21:50 if he can find a way to rekindle his
2:21:53 motivation to continue development and
2:21:56 let it do the coding but he becomes the
2:21:59 sort of Puppet Master
2:22:01 having this do all the Tactical [ __ ]
2:22:05 in their Hills
2:22:07 metaphorically and you know otherwise
2:22:10 I used it from for some Arduino coding
2:22:12 recently and it worked better than I
2:22:14 expected exactly refined inquiry
2:22:17 terrible navigating Discord can Chachi
2:22:20 PT do the mid Journey function for me
2:22:24 um
2:22:25 no but let me show you something here
2:22:27 Clint hopefully you're still here
2:22:31 you can have chat EPT write the prompts
2:22:34 if what you're talking about is writing
2:22:35 the prompts then potentially so I'm
2:22:37 going to show you a couple of things
2:22:38 here okay
2:22:40 so we're going to flip over to
2:22:45 let's see
2:22:48 I'm gonna go to Claude okay so Claude
2:22:51 or to uh to Poe so poe.com
2:22:55 gives you access to to all these
2:22:57 different models but there's a big chunk
2:22:59 of it's free like Claude instant chat
2:23:01 GPT and Sage are all free but what also
2:23:03 is free at poe.com are these are these
2:23:07 Bots
2:23:08 and so
2:23:10 um what the Bots are is just very
2:23:13 specific
2:23:15 prompts that that you can interact with
2:23:18 so I've got two here there's one called
2:23:21 MJ captain mid-journey captain and
2:23:23 there's another one called mid Journey
2:23:24 so if you go to poe.com MJ captain
2:23:28 I'm just going to click on it here
2:23:31 and I'm going to clear my context
2:23:34 I can now type in a prompt that like if
2:23:39 I don't know anything about mid journey
2:23:41 I can just type in
2:23:43 um let's see
2:23:45 um
2:23:47 tropical jungle bird
2:23:53 um in the rain forest which is not a
2:23:58 very good prompt but like that's what I
2:23:59 want and I don't feel like learning all
2:24:01 that [ __ ] what this bot does is it
2:24:04 rewrites that so
2:24:07 but what the painting of a toucan bird
2:24:10 perched on a branch surrounded by lush
2:24:12 green foliage in a misty rainforest in
2:24:14 the style of James Audubon cool using a
2:24:17 wet on wet technique post-impressionist
2:24:19 art movement with a title Tropics alive
2:24:22 focal point on the toucan warm right so
2:24:26 so it made it not just a tropical bird
2:24:28 it made it a toucan it said James
2:24:30 Audubon and then it's got you know dash
2:24:32 dash AR which is aspect ratio for 4.3 no
2:24:37 dry no dry life less monochromatic style
2:24:41 70 right version 5.2 so this rewrote my
2:24:46 shitty prompt into something that's
2:24:48 mid-journey specific now I'm going to
2:24:50 pop over to Discord which I agree with
2:24:53 you that Discord sucks oops
2:24:57 but here we are in Discord and I'm going
2:25:00 to go to my AI Salon Discord server and
2:25:04 in here oops sorry
2:25:06 um I'm in the AI Salon this is my
2:25:08 Discord server and in here we have a mid
2:25:09 Journey channel so I'm going to go to
2:25:11 the mid Journey Channel
2:25:14 and then it looks like Ali's in here
2:25:16 playing right now
2:25:19 thank you
2:25:20 oh that's cool and Kyle is doing
2:25:23 something some other Kyle
2:25:25 and Ally's in there okay so
2:25:27 how you make a an image in mid journey
2:25:30 is you type slash imagine
2:25:33 and then it gives you a prompt and yeah
2:25:36 you have to know all these things so I'm
2:25:38 just going to literally hit paste so
2:25:39 this was the thing that
2:25:42 um that gave me
2:25:44 um I don't want it to be a toucan bird
2:25:46 but I want it to be a of an exotic bird
2:25:54 and then let's see
2:25:56 focal point on the bird
2:26:00 and I'm gonna make it I want it
2:26:03 um
2:26:04 I want it taller than it is I'm gonna go
2:26:07 three or two
2:26:10 because I want it taller than it is wide
2:26:13 and then I'm going to leave the rest the
2:26:15 same and so now I'm gonna hit return so
2:26:18 what that what that bot did was did a
2:26:22 lot of the kind of coding sort of [ __ ]
2:26:24 that you have to do in in
2:26:27 um
2:26:28 in Discord for mid-journey
2:26:31 um so that might help you
2:26:34 the other thing that you can do is if
2:26:36 you don't know it you can so slash
2:26:38 imagine
2:26:39 gives you
2:26:41 an image if you do slash settings
2:26:48 it will pop up
2:26:50 all of your settings so like here's mid
2:26:52 Journey 5.2 5.154321
2:26:56 niji which is the the anime looking
2:26:59 stuff different styles remix mode High
2:27:02 variation low variation things like that
2:27:05 so
2:27:07 um hopefully that helps a bit let's go
2:27:08 look at our little birdies here so like
2:27:11 78 percent done
2:27:14 they're getting better they're looking
2:27:15 more and more Autobahn like which is
2:27:17 interesting that's 93 percent
2:27:24 [Music]
2:27:27 and there we go there's our Autobahn
2:27:29 birdies
2:27:31 and so I can grab any one of those and I
2:27:34 like this one with the with the fruits
2:27:36 and then so so what happens the other
2:27:39 thing in mid Journey if you don't know
2:27:41 this you get these four variations and
2:27:43 then you've got U one through four and V
2:27:46 one through four u means upscale and V
2:27:49 means variation so if I hit V2 it would
2:27:53 give me variations of this image but I
2:27:55 like it good enough so I'm just going to
2:27:56 hit U2
2:27:59 and now it's going to give me a high-res
2:28:01 version of that and I'm going to show
2:28:03 you one other cool thing
2:28:05 so so there's our bird you know our our
2:28:08 sort of you know
2:28:12 was informed basically and there's a new
2:28:15 feature called zoom out so I can zoom
2:28:18 out two times 1.5 times I can do a
2:28:21 custom Zoom let's do a custom Zoom
2:28:23 I'm going to zoom out
2:28:25 um two times but I'm going to make my
2:28:27 aspect ratio
2:28:30 um 16 by nine
2:28:34 and submit it and so now it's going to
2:28:36 actually stretch the width of so right
2:28:38 now we've got a square image so the 16
2:28:41 by 9 it's going to give us more on the
2:28:43 width and it's going to zoom out from
2:28:44 the bird so we'll see a lot more of the
2:28:46 rain forest around this bird
2:28:50 and this is a brand new feature with
2:28:52 version 5.2 of mid Journey which is
2:28:56 super cool
2:28:57 and look there's our bird off in the
2:28:59 distance and notice in this cool so the
2:29:02 the the detailed part of the image is
2:29:04 still detailed but all this is blurry
2:29:06 around him
2:29:07 and and it's it's going through these
2:29:10 cycles of increasingly making this more
2:29:13 and more specific so that's the way the
2:29:15 diffusion models work they start as kind
2:29:17 of random noise and then over multiple
2:29:19 Cycles
2:29:21 and they basically denoise the image
2:29:24 into more and more
2:29:27 um clarity
2:29:28 it's like how it works is that somebody
2:29:31 figured this out like I I don't know I
2:29:34 just feel like we're in this amazing
2:29:36 amazing time in history and we're all
2:29:39 lucky to be able to play with this [ __ ]
2:29:42 like really lucky
2:29:44 it's amazing
2:29:46 all right we got one more one more cycle
2:29:48 here it's at 93 percent
2:29:55 boom there it is so let's see which of
2:29:58 those do we like I don't know
2:30:01 the one with the fruits yeah I guess
2:30:03 that one number one said I'm going to
2:30:05 say upscale number one
2:30:10 and there you go a beautiful beautiful
2:30:14 like you know
2:30:17 picture of a tropical bird where
2:30:20 my input was whatever it was a bird in
2:30:24 the Tropic in the rainforest like you
2:30:26 know an exotic bird in the rainforest it
2:30:28 wrote The Prompt it did all that and
2:30:30 then what I did was I like my
2:30:32 contribution here was curating which of
2:30:34 those options I liked using the zoom
2:30:37 feature to get us this bigger version of
2:30:39 the thing you know
2:30:41 so that's the that's the magic of
2:30:43 mid-journey mid journey is mind-blowing
2:30:45 and and yes your your frustration with
2:30:51 um
2:30:52 with mid-journey as an interface is very
2:30:55 Justified it sucks it's horrible
2:30:59 but
2:31:00 if you want that [ __ ] right now you just
2:31:03 gotta dig through it
2:31:06 um
2:31:07 [Music]
2:31:11 looks like we're winding down here
2:31:14 they're using chat GPT at work the
2:31:16 number of half-baked apps has grown
2:31:19 exponentially it's glorious I know
2:31:22 that's that's one and like that's what
2:31:24 it's going to be like for the like so so
2:31:26 remember when Photoshop came out and
2:31:28 everybody thought they were a [ __ ]
2:31:29 designer like anyone that could launch
2:31:31 an app thought they were a Photoshop
2:31:33 designer and it was just all this bad
2:31:35 art for a long time it's gonna be like
2:31:37 that with everything now right because
2:31:38 it's so easy to produce something at a
2:31:41 like an 80 level
2:31:43 so you're just gonna have
2:31:46 like long articles that are written that
2:31:48 you're going to be halfway through
2:31:49 reading it you're gonna be like this
2:31:51 sucks oh [ __ ] this is one of those and
2:31:54 then it's gonna be like there's all
2:31:55 these Half Baked apps and there's all
2:31:56 these Half Baked articles then there's
2:31:58 gonna be all these half big books and
2:32:00 there's gonna be right
2:32:02 and then what's going to happen is this
2:32:04 I think
2:32:06 within all the
2:32:08 noise of infinite content generation
2:32:13 artists are going to emerge
2:32:16 people that
2:32:18 hopefully learn how to use these tools
2:32:20 in a way that no one had ever thought of
2:32:22 and and just shit's gonna rise out of
2:32:25 the out of the ashes of just all this
2:32:28 you know horrifically mundane [ __ ] that
2:32:31 all looks Fancy on the surface but it's
2:32:33 all just [ __ ] and something
2:32:34 remarkable is going to emerge out of
2:32:36 that and and I and it's it's not going
2:32:38 to be any different than it is now like
2:32:39 you know why did Billy eilish take off
2:32:42 right like why did she rise above the
2:32:44 noise I don't know there was some
2:32:45 magical combination of her and her
2:32:47 brother and how they wrote and how they
2:32:48 produced and how she sang and how she
2:32:50 performed and who they were
2:32:53 that
2:32:55 that happens right and that's going to
2:32:57 continue to happen but
2:33:00 it's gonna be when you when you
2:33:02 democratize creative tools to the point
2:33:06 that anyone can do like I don't have
2:33:09 illustration or painting skills at all
2:33:12 but I could generate something that
2:33:15 looked good enough to put in some sort
2:33:17 of book about birds right there okay so
2:33:20 you know
2:33:23 it's gonna be wild I think we're
2:33:25 entering a great New Renaissance a great
2:33:27 Renaissance
2:33:28 I really do
2:33:32 um it's gonna be exciting all right
2:33:34 seven more minutes I'm leaving at 10.
2:33:36 it's here at Denver 10 is my witching
2:33:38 hour I'm done I'm you've people have
2:33:40 demanded enough you're demanding you ask
2:33:42 and you ask and you ask because I told
2:33:44 you to ask so thank you for that Tap
2:33:47 Away on the screen I appreciate all of
2:33:49 you being here I'm just being a smart
2:33:50 ass all right did he say Paul rateser or
2:33:54 Mark rates or Paul
2:33:55 rates are Roe
2:33:58 t-z-e-r marketing AI Institute
2:34:03 and there's there's a the marketing AI
2:34:06 Institute podcast is a weekly podcast
2:34:09 with Paul rateser and Mark Caputo I
2:34:12 think is his co-host
2:34:15 um for me
2:34:18 that podcast is the one to listen to
2:34:20 it's for a couple of reasons
2:34:23 he's been doing it for five years he
2:34:25 started the marketing AI Institute five
2:34:28 years ago so he was four years early
2:34:30 right so machine learning and ai's been
2:34:32 out for a couple of decades gpt3 and
2:34:36 tools like it and and things like
2:34:37 mid-journey and stable diffusion like
2:34:39 the fact that he's been doing this for
2:34:41 five years and talking about marketing
2:34:43 with AI for five years he's way ahead of
2:34:46 the Curve
2:34:47 so his perspective right now he's just
2:34:50 he's just he's very grounded and when
2:34:53 new stuff comes out he's very thoughtful
2:34:54 about okay here's what it might look
2:34:56 like in education here's why educational
2:34:59 institutions shouldn't block chat gbt
2:35:02 but why they should embrace it here's
2:35:04 how they might embrace it he's just
2:35:05 really good at analyzing the reality of
2:35:08 how we're incorporating this into our
2:35:10 everyday lives so Paul rates are
2:35:13 marketing AI Institute is chat gbt a
2:35:16 programming language that's actually a
2:35:17 great question it's not chatgpt is a
2:35:22 small
2:35:23 application sitting on top of what's
2:35:26 called a large language model
2:35:28 the large language model that it's gpt4
2:35:31 is is what they're what they call the
2:35:34 new fancy one and 3.5 is the one before
2:35:36 it
2:35:37 the large language model is a machine
2:35:39 learning
2:35:41 algorithm
2:35:43 that's been trained on like all of the
2:35:46 public internet and so it kind of knows
2:35:49 all the [ __ ] that was out there good bad
2:35:50 and indifferent right
2:35:52 and then what chat GPT is is it's a very
2:35:55 simple interface sitting on top of that
2:35:57 it's just got the chat box and your chat
2:36:00 history and then it responds with [ __ ]
2:36:04 um
2:36:05 what's interesting about your question
2:36:08 is that
2:36:11 effectively what you're doing when
2:36:13 you're using chat GPT
2:36:14 is your programming a computer using
2:36:18 your native language
2:36:21 you're in here you wrote in English
2:36:24 you're basically using the English
2:36:25 English language as a programming
2:36:28 language so you're saying hey chat GPT I
2:36:32 want you to act like a physicist
2:36:35 and I want you to explain to me the
2:36:38 difference between what Einstein thought
2:36:41 of the cosmos and what Stephen Hawking
2:36:44 thought of the cosmos
2:36:46 and it'll just tell you now how you
2:36:48 would program that traditionally is you
2:36:51 would have to learn a computer
2:36:52 programming language you would have to
2:36:54 identify some data sources you would
2:36:56 have to come up with the logic to say
2:36:58 okay I'm going to identify
2:37:01 um
2:37:03 um
2:37:03 papers from both Einstein and Stephen
2:37:07 Hawking and then I'm going to logically
2:37:09 come up with ways that I might compare
2:37:11 their different theories and
2:37:16 I'm gonna either exclude or include
2:37:18 quantum mechanics because Einstein
2:37:20 couldn't quite figure that out and
2:37:22 Hawking did and right whatever it might
2:37:25 be
2:37:26 historically what you would have to do
2:37:28 is you would have you would have to come
2:37:30 up with all that logic you would have to
2:37:31 identify the data sources you would have
2:37:33 to write the programming language you
2:37:35 would have to write an interface for it
2:37:37 so that it actually displayed something
2:37:40 um and now all you do is you type in
2:37:43 make me a table
2:37:45 that actually let's go do it
2:37:48 because
2:37:49 this is this is I I this continues to
2:37:53 blow my mind every time I use it I still
2:37:55 can't believe
2:37:56 we get to do this we as human beings get
2:37:59 to do this
2:38:00 so I'm in chatgpt 3.5 so this is just
2:38:03 normal chat GPT this is the free one so
2:38:06 I'm going to say make a table
2:38:10 comparing
2:38:12 um
2:38:14 the
2:38:17 Theo theories
2:38:20 of
2:38:22 Albert
2:38:24 Einstein
2:38:27 and Stephen
2:38:30 Hawking
2:38:32 on the cosmos
2:38:36 right so I wrote a single sentence there
2:38:41 and it not only knew
2:38:46 what their theories of the cosmos were
2:38:49 it also knew how to categorize
2:38:53 different elements of those theories
2:38:57 it knew how to program a table and it
2:39:00 knew how to display the table in in
2:39:03 what's called markdown language so that
2:39:05 I could literally copy and paste this
2:39:07 into a spreadsheet right now and let's
2:39:11 see major contributions theory of
2:39:13 relativity Hawking radiation black hole
2:39:16 thermodynamics Right theory of gravity
2:39:19 um Albert Einstein general theory of
2:39:21 relativity Stephen Hawking Quantum field
2:39:23 theory in curved space-time right so so
2:39:26 it actually is getting the the core
2:39:28 elements of their theories of the cosmos
2:39:31 correct put them in a table now could
2:39:34 these be more detailed could they be
2:39:36 cited could but yeah but like here was
2:39:39 my input
2:39:41 right if you knew nothing about this
2:39:43 other than there are two dudes that know
2:39:45 physics
2:39:46 and they talk about the cosmos you now
2:39:48 know way more than you did before right
2:39:52 so
2:39:53 in a sense
2:39:55 it is a programming language but it's
2:39:58 chat GPT is not the programming language
2:40:01 is the programming language Chachi BT
2:40:04 just
2:40:05 knows how to take your English
2:40:07 and
2:40:09 do [ __ ] with it it's like it's just
2:40:13 I I don't know I find this stuff
2:40:15 mind-boggling I
2:40:16 every time I do it
2:40:20 it
2:40:21 I don't know anyone else like that
2:40:24 like it
2:40:28 am I wrong here
2:40:31 it's insane it's insane we're we're
2:40:34 lucky that we get to experience this
2:40:39 as a tech leader I can prototype python
2:40:42 features hand off to Dev team for
2:40:44 refinement and integration exactly like
2:40:46 that's that's how I use chat GPT pretty
2:40:49 pretty consistently now is I quickly get
2:40:52 whatever I'm thinking about to eighty
2:40:54 percent and then either I either take it
2:40:56 over the finish line or I just hand it
2:40:58 off to someone and go give this a proof
2:41:00 tell me what's wrong with it you know
2:41:02 tighten that up
2:41:03 went from weeks to hours exactly that
2:41:05 can't hear
2:41:08 yeah I know someone I brought up on
2:41:09 stage I couldn't hear them either
2:41:13 I was at PSU during the punch card era
2:41:16 too shocking things have come so far so
2:41:18 fast I know but when I when he walked in
2:41:21 with that stack of Punch Cards and
2:41:23 basically said this is your life
2:41:25 I just had this flash of I'll jump off a
2:41:28 building like I'm like that's not gonna
2:41:30 be my life no [ __ ] way no [ __ ] way
2:41:35 for me computers were entertainment and
2:41:37 what he said was that computers are
2:41:39 processing data
2:41:41 and in this really boring horrific way
2:41:45 so yeah so I'm glad I got my acting
2:41:48 degree from Penn State
2:41:52 um your abilities are exactly what I
2:41:54 need for my project can we work together
2:41:56 sure I've got my contact infos in my um
2:41:59 what's it called
2:42:01 profile up there or Kyla storyvine
2:42:04 storyvine.com kylestoryvine.com
2:42:09 wrote the automation code very cool why
2:42:12 don't I like Discord
2:42:13 um discord's fine it's it's
2:42:16 um
2:42:20 it's really hard to find stuff that
2:42:23 you've said before like the search
2:42:24 function's bad
2:42:26 the fact that the conversations are so
2:42:30 disjointed or bad
2:42:33 particularly for mid-journey
2:42:36 like like if if you go look at if you
2:42:38 haven't played with leonardo.ai so
2:42:40 leonardo.ai is a web interface for
2:42:42 stable diffusion
2:42:44 that's what state that's what
2:42:46 mid-journey needs to be like mid-journey
2:42:48 needs to have a web interface or an
2:42:50 application interface that you point and
2:42:53 click on icons and pictures or go use
2:42:56 um Adobe Firefly right you you put in
2:43:00 your prompt but then you can click on I
2:43:01 want this style or that Styler that
2:43:03 style right
2:43:04 you're generating visual stuff with a
2:43:07 command line interface if you're a geek
2:43:09 and an engineer a command line interface
2:43:11 is Perfectly Natural to you but if
2:43:14 you're a human being that's like I just
2:43:16 want a pretty picture from a kid's book
2:43:18 you don't want to have to [ __ ] learn
2:43:21 how to do emojis in Discord and talk in
2:43:24 Discord and find people in Discord and
2:43:26 you go to a mid-journey website that
2:43:28 says join the beta and when you click on
2:43:30 that it takes you to this whole other
2:43:32 application called Discord which you've
2:43:34 never heard of before and you create an
2:43:36 account on it and then you have to jump
2:43:38 through all these hoops to even be able
2:43:39 to let in to use it and then there's all
2:43:42 these conversation threads and you don't
2:43:44 even know where to begin and it's just a
2:43:46 [ __ ] [ __ ] show it's just a nightmare
2:43:49 and and for mid-journey in particular
2:43:51 it's about creating the most beautiful
2:43:53 images you can imagine no one should
2:43:57 have to be doing command line interface
2:43:59 to generate images but that's where we
2:44:01 are right now so Discord is fine as a
2:44:05 precursor to slack but like I use slack
2:44:08 in my business all the time right it's
2:44:10 like business Discord and it's like it's
2:44:13 great
2:44:14 until it's not like try describing a bug
2:44:18 that you discovered in Discord and not
2:44:21 get in a fight with someone you can't
2:44:22 right it's horrible with that so
2:44:26 I don't know
2:44:28 it means it literally just needs to be
2:44:31 not in Discord but it is and so until
2:44:33 it's not
2:44:35 can you copyright AI pictures uh as if I
2:44:38 think the copyright office just
2:44:40 reaffirmed that they're they're not
2:44:42 allowing AI images to be copywritten I I
2:44:46 can't imagine that's going to stay for
2:44:48 that long because especially so so if
2:44:51 I'm using photoshop's generative fill if
2:44:54 I start with a photograph that I took of
2:44:56 a fire hydrant right
2:44:59 and then
2:45:00 it's the the picture is vertical and
2:45:04 then I go into generative fill and I say
2:45:05 make it horizontal and it fills in some
2:45:07 city background but it's still my fire
2:45:09 hydrant but 70 of that picture is not my
2:45:14 photograph but 30 is but my photograph
2:45:18 is the
2:45:20 um focal
2:45:21 Center of that new image
2:45:24 is that AI generated or not
2:45:30 70 of it is AI generated so the
2:45:33 copyright office might say that's not
2:45:35 and then and then okay but if they say
2:45:37 well the primary subject is the
2:45:39 photograph you took so that is
2:45:40 copywritable okay great
2:45:42 how small does that
2:45:44 fire hydrant get before it's no longer
2:45:47 my photograph like if that fire hydrant
2:45:50 is 10 pixels high is that still my
2:45:53 original artwork so I just I think the
2:45:55 whole
2:45:56 the right now
2:45:59 generative AI is like this ghetto off to
2:46:02 the side like I just did that little
2:46:04 rant on on Discord right right now to do
2:46:07 mid Journey you have to go into this
2:46:09 geeky place called Discord it's way off
2:46:11 to the side
2:46:12 but now that it's rolled into Photoshop
2:46:14 and it's just another tool within
2:46:16 Photoshop
2:46:17 how do you know which part is AI
2:46:19 generated and which parts not if I do 75
2:46:22 different layers in there and and
2:46:26 you know ninety percent of it is AI
2:46:28 generated but but I close it and all
2:46:31 that [ __ ] together isn't that mine isn't
2:46:33 that
2:46:34 created with intent with creative intent
2:46:38 wasn't I the one going it was option A
2:46:40 not option b so I don't know where the
2:46:43 boundaries are but currently the
2:46:45 copyright office is saying [ __ ] you to
2:46:47 the AI artists
2:46:48 so we'll see I don't know how long it's
2:46:50 going to last there's gonna be I just
2:46:51 that that to me feels like a 20-year
2:46:54 battle but both directions the zoom out
2:46:57 feature is so cool it really is thanks
2:46:59 for the demo you're welcome where did
2:47:00 you go and put
2:47:02 to help generate the prompt I went to
2:47:04 poe.com
2:47:06 uh Slash MJ captain or slash mid-journey
2:47:10 which are just Bots that people made
2:47:13 so you can probably just go to Poe
2:47:15 there's an area where you can explore
2:47:17 the Bots
2:47:18 and just look up
2:47:20 um let me go see what it it's called so
2:47:23 if you go to Poe
2:47:24 they now move these up to the at the top
2:47:27 left corner there's a button that says
2:47:29 explore
2:47:31 and then there's AI
2:47:35 yeah it looks like the mid-journey one
2:47:37 is is the top trending bot right now
2:47:40 so poe.com mid-journey
2:47:44 it doesn't look like you can actually
2:47:45 search for them here but they have
2:47:46 different categories
2:47:48 so I'd go to the AI category there's the
2:47:51 dolly bot there's the prompt genius
2:47:54 so poe.com slash
2:47:58 explore underscore Bots
2:48:05 yeah
2:48:06 or just go to poe.com and then in the
2:48:09 top left hand corner hit the explore
2:48:11 button
2:48:14 so half of the populations will be made
2:48:16 redundant by this how will that be a
2:48:18 good thing for them well so similar to
2:48:21 how the steam engine displaced most of
2:48:24 the Agricultural jobs
2:48:27 those people found new things to do I'm
2:48:30 not saying that
2:48:32 displacement's not going to be painful I
2:48:34 think it's going to be very painful
2:48:36 but AI is probably going to generate
2:48:39 more jobs than it displaces
2:48:42 um they're going to be different though
2:48:44 so people are going to have to change
2:48:46 and change is hard and some people won't
2:48:48 change and some people will just be
2:48:50 displaced and that's going to be painful
2:48:52 and that's why I I think there's
2:48:54 probably a need for some form of
2:48:57 Universal basic income or some sort of
2:48:59 some sort of transition while while all
2:49:02 of this disruption happens
2:49:04 so we'll see I don't I don't know that
2:49:06 it's going to be half the population but
2:49:07 it's probably going to be 30 to 40
2:49:08 percent are gonna are gonna face
2:49:10 significant displacement but here's the
2:49:13 other deal
2:49:14 so I'm not an economist but this is kind
2:49:17 of how I'm understanding it right now
2:49:19 right now with our current Technologies
2:49:21 you kind of have
2:49:24 the the able-bodied workers of the world
2:49:27 putting in this many hours to produce
2:49:30 this much productivity right and and
2:49:33 whatever that is there's some
2:49:34 unemployment and there's some right it
2:49:35 does it does a little bit of this but
2:49:37 but for the most part we're we're in a
2:49:39 state of stasis
2:49:43 with generative AI like like what I've
2:49:46 heard is tech companies like engineering
2:49:48 companies that are fully embracing AI
2:49:50 are seeing a 30 to 40 productivity
2:49:52 increase right
2:49:54 so if for the same amount of input you
2:49:58 see a 30 to 40 percent increase in
2:50:00 productivity
2:50:02 you can do one of two things you can
2:50:04 either produce more money more stuff do
2:50:07 more right or or you can lower your
2:50:09 prices for the same amount of work or
2:50:12 you can just lay off like that amount of
2:50:14 people right
2:50:16 um and then increase your profit so
2:50:18 there's going to be some companies the
2:50:19 ones largely owned by private Equity
2:50:21 firms that are just going to say [ __ ] it
2:50:24 fire the people and let's let's take all
2:50:26 that profit for ourselves there's going
2:50:28 to be a lot of other companies go that
2:50:30 upskill these workers to either that
2:50:33 they say
2:50:34 you can work less
2:50:37 so I'll let you work 30 hours a week
2:50:39 instead of 40 because we're producing so
2:50:41 much more so as long as we can maintain
2:50:43 what we were doing before or maybe even
2:50:45 increase that slightly if I can drop
2:50:47 your hours to 30 we'll do that or I'm
2:50:50 going to keep your hours the same and
2:50:51 we're going to go get more clients or
2:50:52 we're going to deliver a higher quality
2:50:54 product or whatever it might be so so AI
2:50:56 is going to create this Gap and then the
2:50:58 question is what happens with that Gap
2:51:00 do those people get laid off do they get
2:51:02 upskilled you know
2:51:03 does the does the country make more
2:51:06 money and we realize that and pour that
2:51:08 back in I don't know that I don't know
2:51:12 all right I'm being told I gotta go it
2:51:15 happens it happens it's been two hours
2:51:17 all right everybody
2:51:19 um it's mind-blowing it is
2:51:22 let's see so what was that question
2:51:26 do you have a YouTube I do not
2:51:29 so we can see all of your show this is
2:51:31 all of my show this is it thank you for
2:51:33 the crown oh that's nice
2:51:35 um no this is it I I try to go live once
2:51:38 a night uh let's see it's great it's
2:51:40 really bad all right everybody I'm out
2:51:42 of here peace out thank you for time
2:51:44 thank you for your time thanks for
2:51:46 hanging out talk to you later