aiLL 7/1/23 - The Democratization of AI is Here

Description

Chapters

Transcript

0:01 foreign
0:14 what day is today today is Saturday
0:17 welcome everybody
0:20 welcome to the AI learning lab we have a
0:24 shiny new haircut
0:25 it's very thin on the sides
0:29 oh
0:30 happy Saturday
0:33 welcome welcome welcome
0:37 starting a little early tonight
0:39 I might do a short little session now
0:41 and then I'll do a longer session later
0:43 if you're new here welcome my name is
0:44 Kyle Shannon this AI learning lab I'm an
0:47 entrepreneur and a
0:49 I guess an AI Enthusiast we're
0:52 evangelist or
0:55 demystifier
1:05 I'm also got I've got some perspective
1:08 on this stuff I was there for the early
1:09 days of the World Wide Web I started one
1:12 of the first digital agencies in the mid
1:13 90s
1:14 and um
1:17 the the early days of the world wide web
1:19 and the early days of this generative AI
1:20 stuff feel very very familiar
1:23 uh
1:24 uh very very familiar to me
1:27 um yeah so if you've got questions about
1:28 AI or anything like that pop them in the
1:30 questions below I'll do my best to
1:31 answer them J word
1:34 um well what has AI helped you with
1:37 hmm
1:40 Lots within my business I'm using
1:44 um open ai's whisper so so I'm the
1:47 founder and CEO of a company called
1:48 storybine which is an automated video
1:50 storytelling platform we're using
1:53 whisper to do interactive transcriptions
1:54 right now uh the company's 11 years old
1:57 I haven't had transcriptions for hey
1:59 Troy uh haven't had transcriptions for
2:03 like you know 10 and a half of those
2:06 years because they just suck the auto
2:08 transcription sucked so it's helping
2:10 with that I'm using it within the
2:12 business to
2:14 um for Creative development uh when we
2:17 when we write prompts for our our video
2:19 the way our technology Works we've got
2:21 to write these question prompts it's
2:23 helping to do that when we collaborate
2:25 with clients we're doing
2:28 um chat gbt prompt writing in real time
2:32 hey Emilio's wife what's happening and
2:36 um
2:37 so there's that and then I've got an
2:39 entirely new product Vision uh and uh
2:44 completely re-architecting the value
2:46 proposition and the product of storyline
2:49 so that's just on the business side then
2:51 on the personal side
2:52 um just I'm more creatively engaged than
2:55 I've been in decades and
2:58 um just the more I learn about this the
3:00 more I want to do and uh I've created a
3:02 community around it I've created this
3:04 community around it my AI Salon uh and
3:07 yeah all that sort of stuff another
3:09 exciting Saturday night in the AI
3:11 learning lab exactly exactly
3:14 what'd you do Saturday night I don't
3:17 know there's this old dude on Tick Tock
3:19 we hung out
3:24 so so yeah so so lots of stuff and and
3:27 you know what I've been doing these
3:30 lives now for I don't know three or four
3:32 weeks pretty consistently I've been
3:33 doing them every night and
3:36 um
3:37 just just you know getting people to get
3:41 past
3:43 the fear or dismissiveness about AI
3:45 without having tried it to just try it
3:47 and that's like why I have what Kyle why
3:50 do you have your URLs behind your head
3:52 well I have URLs behind my head because
3:54 if you haven't played with chat gbt like
3:57 here's the deal I I deeply firmly
3:59 believe that
4:03 this generative AI stuff is going to is
4:05 going to be transformative in nearly
4:07 every profession you know certainly all
4:09 knowledge work all creative work but but
4:11 I think it's going to be pretty deep and
4:13 and I think it's gonna it's going to
4:17 um impact in both you know positive and
4:20 negative ways but it's going to impact
4:21 people both personally and
4:22 professionally
4:24 um and so the more people that we can
4:27 get curious about it trying it playing
4:29 with it the better because the the every
4:32 little sort of positive nudge we can
4:35 give this thing in a positive correct
4:37 direction is going to have a
4:38 disproportionately positive effect
4:39 moving forward so the URLs behind my
4:42 head here the top one there is that's
4:44 chat's official website if you've never
4:47 been just
4:48 go there right now and start playing
4:50 just do it just you owe it to yourself
4:52 to just go play if you haven't played if
4:54 you're just listening to other people
4:56 talk about it not good
4:59 um bing.com is uh is also chat GPT if
5:02 you click on their chat button at
5:04 bing.com that's gpt4 connected to the
5:07 internet and it's completely free and
5:09 that also now has Dali image generation
5:12 built in it so if you go to bing.com and
5:13 play with their chat you can make images
5:15 you can play with gpt4 it's connected to
5:18 the internet and it's free and then
5:20 poe.com is like those other two except
5:23 it's got six different models you can
5:25 play with and a bot generator and you
5:27 can play with other people's Bots and
5:29 this there's just all all three of those
5:31 are variations on the same thing and
5:33 then prompts.chat will teach you how to
5:35 prompt and futurepedia.io is a directory
5:39 of like 35 or 3 700 AI tools so if you
5:44 just want to geek out and see what the
5:45 hell's going on with this AI stuff you
5:47 can go there be be forewarned most of
5:50 the tools at futurepedia.io or crap
5:52 because we're early in the business and
5:54 people are just throwing [ __ ] together
5:55 with duct tape and spit and most of it's
5:58 bad okay so
6:00 welcome everybody uh if you have
6:03 questions about AI pop them down below
6:04 I'll do my best to answer them
6:07 um Troy I learned something new every
6:08 day oh that's awesome thank you thank
6:10 you for the kind words I it's good to
6:12 hear that these are worthwhile because
6:14 you know sometimes you know I'm talking
6:16 to an iPhone in my home office so so
6:20 it's hard to know
6:22 uh if it's having an impact and if it is
6:25 I'll keep doing it if it's not I'll stop
6:27 um some youngster was in here what do
6:29 you just you're telling old people how
6:31 to use chat GPT what's the possible
6:33 value in that I don't know old people
6:36 seem to like it
6:38 there's even some youngins in here that
6:40 seem to appreciate it
6:42 that's my other request if you're gonna
6:44 troll me please be creative with it
6:46 don't just give me boring trolls at
6:48 least be creative the best ones are the
6:50 ones that seem like a compliment but are
6:52 they
6:53 those are the good ones all right
6:56 let's see
6:57 I'm glad you thought that was
6:59 interesting what I've been doing
7:00 um how much do you have to pay for it so
7:02 I just said that bing.com is free if you
7:05 want to pay for chat GPT plus which I do
7:07 because it gives me access to plugins
7:09 and it's going to give me access to code
7:11 interpreter and it also gives me access
7:14 to it connected to the internet
7:17 um it's 20 bucks a month Bing is free
7:20 um poe.com if you want access to their
7:22 three Advanced models is 20 bucks a
7:25 month also I pay for one and three and
7:27 I'm not connected with them I'm not
7:28 making any money so I don't give a [ __ ]
7:30 if you uh pay for them or not but that's
7:32 that's what those prices are
7:34 um truth is it's taking over so everyone
7:37 is just going to have to get on board
7:39 it's just like it's it's like disrupt or
7:42 be disrupted right disrupt your just
7:44 disrupt yourself is is the idea and and
7:47 you know honestly like
7:50 disruption's a funny word because yes
7:52 it's going to be disruptive but it has a
7:54 negative connotation to it
7:57 what I'm trying to inspire and and
7:59 encourage people to do is get curious
8:01 about it because once you get past the
8:03 hump of like kind of getting your head
8:06 around the fact that this is a bigger
8:07 deal than it's this isn't just fancy
8:09 Google right it seems like Google right
8:12 it's it's just a simple input box and
8:14 you type [ __ ] in and it makes text right
8:16 that's kind of what Google does right
8:18 this is so much deeper than that and in
8:21 fact if you go to prompts.chat you'll
8:23 see what that really is is just a big
8:25 long list of personas that you can tell
8:28 chat GPT to act like an accountant or
8:31 act like a screenwriter or act like a
8:33 counselor or whatever it might be
8:35 um
8:36 and and you start to get a sense of oh
8:39 this thing can do a lot and it can it
8:41 can do a lot and and
8:43 um if you choose to do that and get
8:45 excited about it and start to re
8:48 reinvent yourself before you're
8:51 involuntarily reinvented or reassigned I
8:55 think I think there's going to be a lot
8:57 of reassignments uh unvoluntary
9:00 reassignments and you know here's the
9:03 good news if you're if you're reassigned
9:05 voluntarily or unvoluntarily we all have
9:08 access to these tools now and they're
9:10 really [ __ ] powerful so you're
9:12 everyone's going to have a chance to do
9:14 all sorts of amazing things so why not
9:16 start now
9:17 and listen if you're not reassigned
9:19 awesome now you now you're an AI
9:21 literate person in a business where all
9:24 businesses are going to be incorporating
9:26 this within what they do
9:28 so it's a really good point J word
9:31 um introduce chat CPT to another
9:33 business owner yesterday and spend an
9:35 hour with him he's hooked that's awesome
9:37 Troy I mean that's the thing it's one of
9:40 my one of my um I guess one of my my
9:42 requests or my thoughts the thing that I
9:44 think is actually quite valuable one is
9:46 get curious right just go there and
9:48 start playing two is get adventurous
9:51 start to figure out well wait like
9:54 you know it writes poems but but can it
9:57 do what I do
9:58 probably like get adventurous and figure
10:01 that out figure out how to disrupt
10:03 yourself and then I think the third
10:05 third layer beyond that is start sharing
10:08 with other people what you're learning
10:09 and what it can do and sit down with
10:11 someone for an hour right I'm spending
10:14 two to three hours every night on this
10:16 channel alone that's independent of my
10:19 newsletter my tick or my Twitter Channel
10:22 what I'm doing at work
10:24 um the the AI Salon that I do every
10:26 other week with our Discord server like
10:28 I'm doing a lot in this and and that's
10:31 not to brag but it's like it feels
10:34 important I wouldn't be doing if it if
10:36 it didn't feel important that's kind of
10:38 how I'm wired
10:39 um let's see we got a bunch of people
10:41 here
10:43 let me
10:44 let me bring someone up here
10:48 did that do anything yeah it didn't do
10:50 anything oh whatever all right let me go
10:53 let me go see what's the best text to
10:55 video generator right now there's a new
10:57 one that just came out I forget the name
10:59 of it oh what the hell is it called
11:02 but but Runway ml the the Gen Gen 2
11:05 right now is probably the best one
11:08 um
11:09 there's a there's a new one that just
11:11 came out like hang on
11:14 um text to video carpathy
11:28 [Music]
11:34 here it is
11:36 there's one called zero scope V2 is so
11:40 so zeroscope V2 is one of those ones
11:43 where you're gonna have to run it you
11:45 know on Google collab runwayml
11:49 um they have an app I think the app only
11:51 does gen 1 right now
11:53 gen 1 is video to video Gen 2 is text to
11:56 video
11:57 um I don't think they've migrated that
11:59 to their app yet so you have to you have
12:01 to do your video generation in their
12:03 Discord server so just like mid-journey
12:05 you have to learn Discord and then you
12:07 get to learn mid-journey and Runway ml
12:10 yay
12:11 why are they using Discord why do we
12:14 have to make images with command blind
12:16 instructions because we're early
12:19 if you want the most powerful technology
12:23 put in the hands of mankind you're gonna
12:25 have to do some [ __ ] like learn Discord
12:27 it'll get better I promise but right now
12:30 it's it's a bit of a [ __ ] show
12:34 oh
12:35 hooked Cody is hooked awesome I learned
12:40 something new every day I know I've been
12:42 it's just I like it's it's multiple
12:44 times a day I'm just like real it does
12:47 that oh I haven't even thought about
12:48 doing that oh oh you know what I could
12:50 do oh oh oh every day every day is just
12:55 a series of oh really
12:58 oh my I had no idea could do that rather
13:03 yeah it's a lot of that [ __ ] going on
13:06 um those of us knew more about AI sooner
13:08 can lead the way exactly that's very
13:10 very true
13:12 and and really positive is it what does
13:14 this one say is AI gay
13:17 AI can be whatever you want it to be you
13:20 you can you can assign whatever pronoun
13:22 to AI you want
13:26 oh my God
13:28 let's put politics in every single
13:31 possible thing let's not not on this
13:33 channel I don't give two shits about it
13:35 I just want us not to be [ __ ] to
13:37 each other so if you're going to troll
13:39 me be creative and if you're gonna be an
13:41 [ __ ] hit the road
13:43 because we can have fun here we can
13:45 transform the world and have fun
13:47 and bend this toward the good that's why
13:50 we're here that's why I'm here all right
13:52 I got lost in my questions I was
13:54 following so I'm trying to find where I
13:55 was
13:56 best video generator I learned something
13:58 new every day we said those
14:02 what's the add-on for AI images well so
14:07 um so in bing it's just in there so let
14:09 me let me show you Ben because I haven't
14:11 I haven't actually showed this much here
14:14 thank you very much for the kind Hearts
14:17 the kind words
14:19 um let's go to bing.com
14:23 all right so here we are at bing.com
14:26 the the normal search engine right here
14:28 you got this little chat button
14:31 and and actually the search button does
14:33 GPT stuff as well but if you want to go
14:35 into full-on
14:37 cat gbt you have three different modes
14:40 here you've got more creative which is
14:42 like gpt4 you know sort of writing mode
14:45 you've got more balanced and then you've
14:46 got more precise so if you're doing
14:48 research for example you want it on more
14:51 precise if you're doing creative writing
14:52 you want creative writing I think it
14:54 defaults to more balanced
14:56 um but but within here you know you can
14:58 do anything you can do in chat GPT like
15:01 right you know write me a
15:04 song Let's see a
15:07 um
15:08 uh a hit country song
15:14 oh dang it I can't type
15:17 song
15:19 what are we gonna do it about about
15:23 um
15:25 my chubby
15:28 dog
15:31 all right and so off it goes verse one
15:35 it's called my chubby dog
15:38 let's see he's got a big belly and a
15:40 wagon tail he loves to eat and he never
15:42 fails he makes me smile and I'm feeling
15:44 low he's a chubby dog and I love him so
15:46 not exactly uh Grammy award-winning
15:49 lyrics there but you know it's fine it
15:50 understands verse courses things like
15:52 that but now I can also say
15:55 um make me and me
16:00 an image of my black and white chubby
16:09 dog who's a mutt that has
16:15 beagle and uh bull
16:20 dog in him
16:23 we'll see what this does this should
16:25 just go off and generate some images
16:29 sure I'll try to create that
16:34 so now it is
16:37 your image is generating it's actually
16:39 pretty fast this image generation
16:41 unbeing
16:47 actually that's not too far off from my
16:49 my old pupper
16:51 that one there's pretty close chimpy
16:53 chimpy this doggy on screen looks like
16:56 you
16:58 that was pretty close he's got more of a
17:00 white face but yeah but he's fat like
17:02 that he's got a fat neck like that
17:05 he's sitting right beside me
17:07 um so that's that and then and then
17:10 there's there's other there's other
17:12 tools like there's a tool called
17:14 leonardo.ai
17:16 that you have to get on the wait list
17:18 for but I would I would strongly
17:19 encourage you to get on the wait list
17:21 for this leonardo.ai
17:24 um it's this is a web front end to
17:27 stable diffusion and it's got all these
17:29 different styles that you can choose
17:31 from and like these are all images that
17:33 people have created and if you find an
17:35 image you like like here's a cool bird
17:37 you can click on that bird and and you
17:41 can see what prompt that they used and
17:43 what negative prompt that they used
17:45 and depending on
17:50 how they did their thing you can
17:52 generate I think you can edit any of
17:54 these generate with this model right so
17:57 you can copy their prompt and you go
17:59 generate with that model
18:02 um you can also train your own
18:06 um train on your own images like I'm a
18:07 photographer so I uploaded a bunch of my
18:11 um Cloud photography images
18:14 and let's see train model no I don't
18:18 want to train model
18:19 oh fine-tuned models I'm gonna go use
18:21 that model
18:23 so I'm gonna go use this I'm gonna
18:26 generate with this model
18:28 so there's some things I did so so let's
18:30 say dramatic Sunset over Manhattan from
18:34 Brooklyn let's say dramatic Sunset over
18:36 the brook
18:39 Lynn Bridge
18:43 um
18:44 from
18:47 the roadway
18:52 and so now that'll go off and it'll
18:55 generate images and but it'll use my
18:58 cloud photography for the sky more than
19:00 likely which is you know what some of
19:02 these ones down below here did what else
19:05 have I done there's some muscle cars so
19:08 so this is Leonardo so this is another
19:10 way to do images I'll come back to that
19:12 and then there's
19:14 um in Discord in in Discord there's a
19:16 thing called mid-journey which in my
19:18 opinion is the very best of the
19:21 um of the tools
19:23 and you have to generate things within
19:26 um within Discord so if you don't know
19:28 Discord these are all the different
19:29 servers here's your different
19:31 conversations within a server it's kind
19:33 of like Slack
19:35 and then mid-journey's got their own but
19:37 you can also put mid-journey inside
19:38 other servers so this is my AI salon and
19:41 this is our mid-journey Channel
19:43 and the way you make it an image here is
19:46 you type slash imagine
19:48 and then you type in a prompt and so
19:50 we'll just go copy someone else's prompt
19:53 here those cats are pretty cute
19:55 so we'll just go grab one of those and
19:57 copy
19:59 and then we'll pop it in here
20:02 it says in the style of Van Gogh starry
20:04 starry night so we're going to say in
20:06 the style of Annie
20:11 Lebo wits
20:13 leave oh wits I don't think that's
20:16 spelled right but that's okay and we're
20:18 going to say photograph
20:21 [Applause]
20:23 of a white cat okay
20:25 so now that's going to make us some
20:27 images so let's go back over to Leonardo
20:31 and there's our Brooklyn Bridge that
20:33 one's actually really good these other
20:35 ones don't look like the Brooklyn Bridge
20:36 at all like this this one this one looks
20:39 like some concoction
20:41 that's like London Bridge Brooklyn
20:43 Bridge some strange Masonic temple but
20:48 this one's actually pretty close
20:50 not quite there but you know not bad and
20:53 there's my sky right so it pulled in
20:56 it pulled in some some you know edits
21:00 and things that very much look like my
21:02 photography so that's pretty cool that
21:04 one's pretty trippy
21:07 um so anyway I don't know if that
21:08 answered your question let's go back and
21:10 see if the Annie Leibowitz pictures
21:12 there yeah it's waiting to start not
21:14 there yet but anyway
21:16 all right let me flip the camera around
21:18 welcome everybody got a got a good crowd
21:21 here on a Saturday night
21:23 thanks for hanging out with me on a
21:25 Saturday night I don't know why you'd
21:26 want to hang out with me on a Saturday
21:28 night
21:29 go on a date clap you're good you don't
21:33 need to be here to say I should'll be
21:35 here when you get back
21:39 ah it's never a dull moment here at the
21:42 AI learning lab
21:43 you like my haircut I got the little
21:45 sides trimmed trimmed it up it's not
21:47 such a [ __ ] nightmare anymore well
21:49 because it's got glue in it from from
21:51 the barber
21:53 all right
21:56 um
21:58 all right let's see oh there's tons of
22:02 comments in here thank you for the
22:03 comments I'm just trying to figure out
22:04 where it was
22:06 putting something new every day
22:08 hopefully that helped you Cody
22:11 I have a disease that will destroy my
22:13 memory and recall
22:15 in 10 years training the llm to recall
22:18 for me that's a huge use case I think
22:20 the idea I just saw I just saw a thing
22:23 yesterday that that someone just
22:25 launched a
22:27 um
22:27 like a website that that's its purpose
22:29 is that you can train it up to be other
22:31 people so you can do it for memorials
22:33 but that particular use case I lost my
22:37 dad to Alzheimer's so so I resonate with
22:40 that and and
22:42 um a good friend of mine Richard Boyd
22:44 before all the chat GPT but before GPT 3
22:48 came out he created
22:50 a virtual version of his father when his
22:53 father passed him his father had all
22:54 this handwritten stuff and he he had all
22:57 that um handwritten stuff transcribed
22:59 and he built a model about his father
23:02 um I think I think that's something that
23:04 we're going to see more and more and
23:05 more of in the coming in the coming
23:07 years not like short years like in the
23:09 next one to three years because you can
23:12 already do a lot of it today
23:15 um and you know it's going to be
23:16 interesting because companies like Apple
23:19 and Google and Facebook and Twitter that
23:23 have you know access to our social media
23:26 histories are going to be able to very
23:28 very quickly build up personalized
23:31 models of us
23:32 with our permission of course because
23:34 they wouldn't steal our data and exploit
23:36 it would they
23:38 uh but it's great use case and I think
23:42 it's going to be quite powerful oh
23:43 there's the the Annie Leibowitz cat's
23:45 coming it's pretty good I'll show it to
23:47 you in a second
23:49 nice choice choice and shirts this
23:51 evening uh Emilio's wife knows that I I
23:53 wear shirts that are very nice but they
23:55 always have like little white things on
23:57 them so you can't ever see the text so I
24:00 have been consciously wearing dark
24:01 t-shirts so you can see the text over
24:03 them
24:05 um it's actually going exceptionally
24:07 well oh that's great is that the uh
24:09 that's oh that's the creating of the
24:11 model of yourself that's awesome man
24:12 that's incredible that's really cool
24:15 um are a in our our plugins for
24:19 chat GPT free they're not right now so
24:22 you have to pay for chat GPT plus to get
24:25 access to them because they're in beta
24:27 but Microsoft has announced they're
24:29 supporting openai's plug-in architecture
24:32 so what that means is when they when
24:34 they move out of beta Microsoft will
24:36 incorporate them into Bing so again why
24:39 I have Bing here is this is chat gpt4
24:43 for free connected to the internet and
24:46 it will have all the stuff that chat GPT
24:49 has that they decide to incorporate
24:52 um
24:53 but paying the 20 bucks to open AI gives
24:56 it to you sooner right you you basically
24:59 get earlier access to to stuff if you're
25:02 an open AI subscriber
25:06 um so
25:08 and and and the plugins like are the
25:11 plugins right now worth paying for
25:14 um
25:15 they're very janky there's a lot of them
25:18 a lot of them are just horrible pieces
25:21 of [ __ ] that are not worth your time
25:24 um there's not really a ranking system
25:26 right now so it's not it's not the app
25:28 store yet right so I don't if you're old
25:30 enough to remember when the iPhone first
25:32 came out there wasn't an app store and
25:35 the apps weren't called apps they were
25:37 called widgets and they weren't actually
25:38 applications they were hyperlinks to
25:41 websites
25:43 um and so you you know developers
25:44 basically had to make websites and call
25:46 them widgets and create an icon for them
25:48 that's kind of the state of plugins
25:49 right now like people are just playing
25:51 with them they're trying to figure it
25:53 out there's some really interesting ones
25:54 but but you know they're they're this
25:57 many
25:58 um and there's there's about 600 of them
26:00 now to dig through but they're
26:02 interesting and it it points to a future
26:05 where where we're going to have an
26:06 interface where you can just kind of ask
26:08 this thing what to do and it will go
26:11 figure it out it'll figure out do I need
26:12 to go to a website do I need to write an
26:14 API do I need to write an application
26:17 so so that's the kind of future we're
26:19 headed towards so if you want glimpses
26:21 into that early then it's worth paying
26:24 for that if you're like [ __ ] it I'll
26:26 I'll play with this stuff when it's
26:28 ready for prime time then I just go to
26:30 Bing
26:30 [Laughter]
26:35 this isn't just fancy Google Kyle
26:37 Shannon yeah you can quote me on that
26:39 this is not just fancy Google it's way
26:42 more profound
26:44 very profound
26:47 I don't think you have any idea however
26:49 founded it
26:52 all right A lot of people have an idea
26:54 how profound it is that's here's a
26:56 here's a piece of uh
26:58 piece of interest in trivia the the
27:01 speed with which that piece of software
27:04 got to a hundred million users six weeks
27:07 it's the fastest adoption of technology
27:09 in history
27:12 um if you're curious is anyone actually
27:14 using this yes they are I am a complete
27:18 noob in programming how can I run auto
27:21 GP oh author GPT GPT Okay so
27:26 huh
27:27 [Laughter]
27:30 there's there's not an easy way to so so
27:33 basically what you need to do so here's
27:35 here's the way you do it so right now
27:37 GPT author is
27:40 um is a project that's it's a bunch of
27:42 Python scripts that are put together in
27:45 a Google collab notebook Google collab
27:48 is basically
27:50 um you you use these things called
27:52 notebooks which are just basically
27:54 step-by-step chunks of code that you run
27:57 on a virtual machine
27:58 so you've got to run it on a Google
28:00 collab virtual machine so the first
28:02 thing I do is I go to YouTube and watch
28:04 videos on you know
28:06 Google collab for beginners just
28:09 understand what it is what Google collab
28:11 is and what notebooks are basically what
28:14 they are is you can run code and see the
28:18 code you're running but you don't have
28:19 to write it there's there's people
28:20 there's Engineers out there writing all
28:23 sorts of cool software in these
28:25 notebooks once you understand how the
28:26 basic notebook model works then you can
28:29 play with different projects
28:31 um I'll go show it to you right now but
28:33 the the kind of uh prerequisite it's
28:36 like we're running a university here
28:38 here your prerequisites for this course
28:40 are just understand basically what
28:43 collab is it's a virtual machine that
28:45 you're running on Google servers and
28:47 you're running these code chunks that
28:51 are in a thing called a notebook
28:53 essentially by pushing the play button a
28:55 bunch of times
28:57 um but they're janky and and they break
28:59 and you kind of have to troubleshoot and
29:01 like so for example when I was first
29:03 trying to get my head around this I was
29:05 trying to install stable diffusion
29:06 because I wanted to do this project that
29:08 image behind all those URLs that's from
29:11 a project I did called Kyle Shannon
29:13 dreams I failed nine times
29:16 getting it running using Google collab
29:19 so
29:20 it ain't easy but let me let me go show
29:24 it to you here because it's worth it's
29:26 worth just at least
29:28 doing what I can to get you up to speed
29:29 on it
29:31 oh there's the kitty cats
29:33 that's the the photograph of a white and
29:36 orange cat sitting in a rocking chair
29:38 it's it actually gave us a white cat in
29:40 an orange rocking chair which is kind of
29:43 interesting
29:45 but like some of these are pretty good I
29:47 mean
29:48 mint journey is just amazing I mean this
29:50 one in the upper right actually the
29:52 lighting in this one in the lower right
29:53 oh this is a cat wearing like a silk
29:56 silk pajamas in a wicker chair that's
30:00 freaking awesome
30:04 that's actually very Annie Leibowitz too
30:06 like that really captured like like if
30:10 Annie Leibowitz was doing uh celebrity
30:12 cats I think this is what it would be oh
30:15 my God that's awesome all right
30:19 ah let's see okay so here's GPT author
30:22 so okay let me
30:25 um let me make this a little bigger so
30:27 hopefully you can see it
30:31 um
30:34 and let me close that so this is kind of
30:37 what it'll look like when you come in
30:40 so when you first run it it'll it'll say
30:43 like GPT author version 1.1
30:46 and
30:48 um it'll kind of describe what it is
30:50 here's the GitHub repo so github.com M
30:54 Schumer slash GPT Dash author if you go
30:57 there
30:59 it'll it'll sort of create your own copy
31:01 of this and or it won't it won't create
31:04 your own copy it will launch
31:07 that notebook in your version of Google
31:12 collab
31:13 and then when you launch it there's a
31:16 thing that'll show up up here where you
31:18 can say copy it to Google Drive
31:20 so you also have to have Google Drive
31:22 running
31:23 so I've already copied mine to Google
31:25 Drive which is why that says copy of GPT
31:27 author to Google Drive okay so then what
31:30 these notebooks are is there's there's
31:32 these cells and and
31:35 each cell is numbered and at the
31:37 beginning of the cell is a little play
31:39 button and and basically when you hit
31:41 the play button it's running the text
31:43 that's in the cell like for example
31:44 you've got to put in oh I'm showing
31:47 people my my uh my open AI keys I'm
31:49 gonna have to go change those uh anyway
31:52 whatever hackers hackers are gonna take
31:54 my keys but anyway you you're gonna put
31:56 your keys in there so it's going to run
31:58 this code it's going to run your keys
32:00 and then
32:01 um
32:02 so I'll hit run here and so it spins and
32:06 so like this is the first chunk of code
32:08 to whatever this does uh
32:11 install the libraries and you know
32:16 put my keys in there
32:18 and then the next one using Google
32:19 Translate for translation so you run
32:21 that and now it's doing that and so so
32:25 basically you just run each of these and
32:27 when you get a little green check mark
32:29 that means that you're good to go and if
32:32 it gives you an error then you've got an
32:34 error
32:35 and then
32:38 um
32:39 this is this is where I'm actually
32:42 writing what the book is about what the
32:44 writing style is and I just keep going
32:47 down down hitting the play button
32:48 hitting the play button and then the
32:51 very last one
32:52 you hit it will actually start stepping
32:55 through and show you that it's writing
32:57 the book and then it actually delivers
32:59 the book in one of your Google folders
33:02 and so in this case
33:06 it wrote a book called
33:10 um
33:11 what's it called
33:12 breaking the mold
33:14 it's a it's about an average guy
33:18 um in a world where everything is below
33:20 average and so because he's average he's
33:22 he's like superhuman and so it wrote me
33:25 you know a 14 chapter book
33:29 um using Fight Club as the uh as the
33:33 sort of satire model
33:34 so anyway that's it and it generated
33:36 that image that that cover image
33:39 so
33:41 there you go so I don't know if that
33:42 helps or not
33:44 that help
33:46 is that good if that's good let me know
33:49 if it's not I'm happy to answer other
33:52 questions about it
33:56 we lost some people because that was
33:59 boring as [ __ ] that was just me pointing
34:01 at a screen of text yeah
34:04 go ahead double tap the little hearty
34:06 hearts and then
34:08 um that'll let people know we're here
34:09 and if you wouldn't mind share the live
34:11 that'll also let people know we're here
34:13 I don't like country music but country
34:15 in the style of Beatles I just thought
34:17 is funny it's really good at doing stuff
34:19 like that you can mix you know things
34:21 like you know write me a country song in
34:23 the style of The Beatles it's good at
34:24 that
34:25 what site is this I don't know what that
34:28 was in reference to so I don't know
34:32 oh it jumped me to the bottom of the
34:33 comments again so I'm scrolling up here
34:36 to see where we are
34:39 good evening West Side Remy
34:41 hello hello
34:45 all right
34:50 [Music]
34:54 um
34:55 can an AI be built as a male but it
34:59 feels female well you can create
35:03 so yes I mean you can I like like I
35:06 don't quite know what that would look
35:07 like
35:08 if you know if you're talking sort of
35:10 cliches you could say I you know I want
35:12 you to to identify as a male but be
35:15 super empathetic or whatever like
35:17 however you characterize what that would
35:19 act like you could actually create a bot
35:21 so if you go to poe.com
35:23 one of the things you can do there is
35:25 you can create a bot and when you create
35:27 a bot you're actually prompting it to
35:30 act a certain way in fact let me I'll
35:33 I'll go show you some that I've created
35:36 and then you can go do that
35:38 like like creating bots is actually a
35:41 really good
35:42 um way to
35:44 um
35:45 practice prompting so where I am right
35:47 now is I'm I'm at poe.com
35:50 and pro dot comes just like chat gbt
35:52 where you got the little chat bot at the
35:54 bottom you know it looks like Google but
35:55 it's not and then you've got
35:59 um six different models the first six
36:00 choices are different chat chat gbt
36:03 models three of them are subscription
36:04 only
36:05 but then at the top here you can explore
36:07 Bots or you can create bot and then
36:11 somewhere down here I can go look at all
36:12 of my Bots and so let's go look at so I
36:15 did Shakespeare I did Jay-Z I did Ricky
36:18 Bobby I did Sling Blade I did Yoda
36:23 um I did Steve Jobs
36:25 Mary Shelley Maya Angelou Virginia Woolf
36:28 so all of these like if you want to go
36:30 play with the Bots I wrote like you can
36:31 go to poe.com I am Yoda and it will talk
36:35 like Yoda
36:37 um so so let me pop one of these open so
36:39 here's the Steve Jobs bot and let me
36:42 just go in and show you
36:45 the how the The Prompt that I used so I
36:49 named I need you name the bot I am Steve
36:52 Jobs you give it a description by the
36:54 way I had chat GPT help me help me do
36:56 all this
36:57 and then you've got the actual prompt
36:59 itself so the prompt is
37:02 you are Steve Jobs the iconic founder of
37:04 Apple Inc and a Visionary leader in
37:06 technology and business engage with
37:08 users by sharing your passion for
37:09 Innovation design
37:11 then I say as Steve Jobs your memory of
37:15 your accomplishment and experiences is
37:16 exceptional allowing you to do
37:20 feel free to address questions regarding
37:22 the principles strategies and
37:24 philosophies of your of your leadership
37:25 style as these are important components
37:27 of your legacy following these
37:29 guidelines keep the interactions brief
37:31 unless a user requests it do not say you
37:35 work for Apple in any way do not refer
37:37 to your own death the reason I put those
37:39 two in there is because as I was testing
37:41 this bot
37:42 he would either Steve Jobs would either
37:45 say I work at Apple right now and that
37:47 was creepy and weird or he was he would
37:49 mention his own death like in his
37:51 answers which was weird
37:53 um but and then I said provide guidance
37:55 as an individual so so if you want to do
37:57 a bot that's a man but acts like a woman
37:59 you would basically just modify you know
38:01 create a prompt that sort of sets up the
38:04 personality of what you want to bot bot
38:05 to be
38:07 um and then you can do an intro message
38:08 and then you hit save that you bet you
38:10 basically launch it and it's live and
38:12 you can go play with it and you can
38:13 share it with other people so
38:15 um that's why I have Poe on the PO on
38:18 the list because even though Poe is is
38:21 not super great there's some cool [ __ ]
38:24 like the Bots are pretty cool the other
38:26 thing about Poe I'll show you one other
38:28 thing here that is this is the reason I
38:31 actually paid for the subscription here
38:33 I wasn't going to ever pay for Poe but
38:36 they're the only ones I know where you
38:37 can get access to this model Claude
38:39 instant 100K but that is is that's
38:43 um
38:44 anthropic's equivalent of GPT 3.5 but it
38:48 has a hundred thousand token context
38:50 window which basically means you can
38:54 um put a 75 000 Word document into your
38:57 prompt so so this big long prompt right
39:02 here is
39:03 um terms of service for some service I
39:06 was signing up for and I've never read
39:08 terms of service in my life so I decided
39:11 to say tell me the most important things
39:13 I'm agreeing to in short bullets so it
39:15 gave me a set of short bullet points of
39:17 what that agreement says so but you know
39:20 75 000 words you if let's say you wrote
39:23 a novel and you've got a first draft of
39:25 a novel and you want to actually be able
39:27 to interact with it and have it give you
39:29 editing notes on your novel you could
39:31 just paste the entire novel in here and
39:33 start interacting with it I mean that's
39:35 mind-blowing to me
39:37 so that's why I paid for that one
39:40 ah my my throat's getting a little
39:42 crunchy I gotta drink here some water I
39:45 gotta drink some water I'm a little
39:47 parched if you know what I mean
39:51 um compare this to chat gbt and what's
39:54 better
39:55 compare what to chat gbt what was I
39:57 using Poe
39:59 oh oh Bing okay
40:01 um chat gbt at open AI chat gpt4 at
40:06 openai is better than Bing which is also
40:10 chat GPT 4.
40:12 because they have different
40:15 um safety guard rails and different fine
40:18 tunings so Microsoft is just a little
40:21 shittier right now they had to kind of
40:22 lobotomize their gpt4 because uh it made
40:26 the cover of the New York Times when it
40:28 acted badly to a New York Times Reporter
40:31 which still cracks me up every time I
40:34 say it
40:36 it told him its name was Sydney and it
40:39 wanted him to leave his wife because it
40:41 was in love with him that made the cover
40:42 of the New York Times
40:46 [Music]
40:50 I was showing Google collab
40:53 um I asked Chachi PT to write a country
40:55 song it was written by The Beatles it
40:57 was hilarious that's awesome that's
40:59 really good
41:00 you could probably you know it probably
41:02 even you could probably have it give you
41:04 a set of cords for it too and it would
41:08 probably like know enough about beatles
41:10 chord progressions and Country chord
41:13 progressions that it would give you you
41:15 know like a like a four chord CDG kind
41:18 of country song but with some with some
41:21 Beatles progressions in there that's
41:24 that's kind of cool
41:26 um
41:29 I don't like country music okay we saw
41:32 those this is Leonardo dot AI ah good
41:35 you're helping each other thank you for
41:37 doing that the previous one with the
41:39 song I missed the the one with the song
41:42 Oh that was Bing bing.com so
41:45 so the The Dark One
41:48 um that didn't have the the robots in it
41:50 was chat GPT the white one with with
41:53 creative balanced and precise was Bing
41:56 and then the dark one that had the bots
41:58 in it is poe.com so all the ones I've
42:01 showed so far other than the image tools
42:03 are are one of these three
42:06 so take a screenshot
42:08 take a screenshot
42:10 [Laughter]
42:13 why do you have a screenshot of an idiot
42:15 I don't know
42:18 there's websites on there too
42:21 but why is he smiling like that that's
42:23 just what he does
42:27 my kids have a friend over right now
42:29 he's probably gonna overhear me
42:32 and be concerned for all of our safety
42:34 all right
42:36 um does anyone know the website this is
42:38 looks like you're helping each other
42:40 with that
42:42 your Discord is free the link oh yeah so
42:45 so for my stuff
42:49 I did the wrong thing I hit the wrong
42:52 button an old man trying to use tick
42:54 tock okay
42:57 um so the AI Salon that I created if you
43:00 go to the salon.ai that's the website
43:03 address
43:04 um there's a link tree there
43:06 and you can sign up for the meetups we
43:08 meet every other week in person here in
43:11 Denver and then on
43:13 um
43:14 on online for if you're not in Denver
43:17 and then we also have a Discord so the
43:19 link the invite to the Discord is there
43:21 and there's also a LinkedIn group for
43:25 the salon as well that's not as active
43:27 but it's just kind of a different it's
43:28 it's a different group if you're on
43:29 LinkedIn and then everyday AI news
43:31 that's the Twitter handle for the
43:33 newsletter that I do but but I put up
43:35 I've been putting up these long Twitter
43:37 threads
43:39 um I put one up today Ethan Malik did an
43:42 article about the coming homework
43:45 apocalypse he basically says listen the
43:48 internet made it really easy for
43:50 students to cheat ai's gonna
43:53 [ __ ] destroy it so teachers get ready
43:57 he's a he's a Wharton business school
43:59 professor uh mid journey is no longer
44:02 free mid Journey's never been free
44:05 um if you want to play with something
44:06 that's kind of like mid-journey that is
44:08 free it's called Blue Willow
44:10 if you enjoy generating images and
44:12 Discord which I don't if you just want
44:15 to make images so a couple of things
44:17 about image generation
44:18 if you're if you're in a business
44:21 especially in a big company
44:23 um Adobe Firefly has been trained on
44:26 non-copy written or or images that adobe
44:30 has the rights to
44:32 um so there's no legal exposure if you
44:36 use Adobe
44:37 generative AI products which is
44:40 brilliant on their part by the way and
44:42 and the fact that they're moving as fast
44:44 as they are and like they've rolled
44:46 generative fill into Photoshop and now
44:49 they're building recolor and illustrator
44:52 that's really good
44:54 um
44:54 the the
44:56 um
44:57 the the other ones you know mid journey
44:59 is the best
45:01 stable diffusion is the most flexible
45:03 but it requires you to be geeky big uh
45:06 um uh Dolly which is open ai's image
45:10 generation tool is not as good as the
45:12 other two but that's that's the thing
45:14 that's rolled into Bing so so there's a
45:16 bunch of different options out there
45:18 there's some free ones out there
45:19 Leonardo dot AI is the is is the first
45:23 and best
45:25 thing interface that I've seen for
45:27 stable diffusion where you don't have to
45:28 do all the installs for stable diffusion
45:30 but you do have to get on their wait
45:33 list so that's that's the drag about
45:34 leonardo.ai but it's really good it's
45:37 really good it's raining in Chicagoland
45:39 it's actually nice here although it's
45:41 been raining we we had Denver so far in
45:43 the first six months of the years had
45:45 more rain than Seattle it's never
45:47 happened in history
45:50 um yeah the the rain in Chicagoland the
45:52 uh the Nascar race is there today the uh
45:54 the Xfinity the Xfinity race
45:57 um got rain well I don't know if it's
45:58 still rained out but um it was rained
46:01 out for a while there
46:02 but that that um that street course in
46:05 Chicago is wild then like the the tight
46:08 little turns like long straightaways
46:10 where they're blasting down there but
46:12 like the average speed for the whole
46:14 course is like 85 miles an hour so for
46:16 those guys it's pretty slow
46:18 too smoky to go out interesting what a
46:21 are you using for video editing though
46:23 are there any there are probably let me
46:26 let me put these other URLs back up here
46:29 hang on oh damn it keep doing that he's
46:33 not the brightest guy
46:35 all right
46:36 um go to futurepedia.io and go into
46:39 their video section I know that there
46:42 there's one called video that's a bunch
46:44 of tools there there are a bunch of
46:47 different
46:49 AI augmentations to video tools but
46:52 right now they're just kind of like
46:53 these ad hoc tools like here's one that
46:55 will remove your background here's one
46:57 that will do rotoscoping here's one
47:00 that'll do this here's one that'll pull
47:02 out social media clips from longer form
47:04 content
47:06 so right now they're kind of they're
47:07 they're just kind of all over the place
47:09 they're kind of these ad hoc tools
47:11 um I don't know if there's one that like
47:13 ideally what we want and Adobe will do
47:16 this right ideally what we want is we
47:19 want to be able to do with video editing
47:21 what we currently do with chat GPT for
47:23 writing right you know find the
47:27 interesting [ __ ] in my video and Stitch
47:29 it together with the appropriate b-roll
47:31 it'll go do that so there may be some
47:35 things there yes Thelonious I got a
47:37 haircut what do you think
47:39 the guy was cutting my hair he's like
47:42 you want me to get those eyebrows I'm
47:43 like yeah I think it's time
47:46 I said you can get the the ear here too
47:50 got that too he's like yeah you just hit
47:53 a certain age it just starts coming out
47:54 right
47:54 [Laughter]
48:01 I've used chat to save my mom 18 000 in
48:06 back taxes that's awesome I mean that's
48:09 the thing about these tools is like
48:10 they're they're incredibly good at
48:13 reading documents so you for example the
48:16 tax code right have it read the tax code
48:19 and then have it read you know some
48:22 expenses you know your your uh your uh
48:25 whatever your your Quicken file and then
48:28 have it help you put that [ __ ] together
48:29 that's awesome
48:31 that's pretty incredible
48:33 um
48:35 I'll send a link with the mega prompt
48:37 that's cool I don't know what Mega
48:39 prompt you're talking about
48:42 plugins or hodgepodge Lodge they are
48:45 plugins right now are a mess my mom's a
48:47 widower in the state of California
48:48 trying to take advantage of her yeah use
48:51 CH use chat GPT to to you know [ __ ] with
48:55 the bureaucracies I think that's I think
48:57 that's what the thing that I'm most
48:59 excited about about these when I talk
49:01 about you know these are way more than
49:02 just a fancy Google is
49:05 um
49:06 this is like a like a simplification
49:08 engine like it translates languages it
49:11 it translates legal documents into like
49:14 something we can all understand without
49:16 having a law degree
49:18 um it can read your writing and and help
49:21 you write new stuff like it so it's like
49:24 a simplification layer that we're going
49:26 to be able to stick on top of all these
49:28 over complicated bureaucracies and say
49:31 okay here's all the rules that this
49:33 bureaucracy says we have to follow
49:35 here's the [ __ ] I'm in the middle of
49:38 right now help me get out of this and
49:40 it'll help us get out of this if you
49:41 haven't if you haven't seen it yet go
49:43 check out um there's a site called do
49:45 not pay
49:47 um which is it's it it's called your the
49:49 the world's first AI lawyer or something
49:51 like that it's got some stupid uh
49:53 marketing uh but it's been around for
49:55 six years or so but it does like it'll
49:57 fight traffic tickets for you and
49:59 parking tickets and things like that so
50:01 it's doing like you know small claims
50:03 lost stuff
50:04 started by a 19 year old by the way
50:06 not a lawyer
50:09 there's gonna be a lot of disruptive
50:11 things created with this technology
50:13 um by non-specialists in that field
50:17 because they don't know any better right
50:18 so they'll just create something
50:20 completely disruptive
50:22 all right I use chat GPT as my personal
50:25 estate and probate lawyer super cool
50:29 um for these types of use cases here's
50:31 the file with the mega prompt for the 18
50:33 000 savings oh that's awesome
50:36 the file also has a Monte Carlo
50:38 simulation and a PTSD GPT program I
50:41 created that's really awesome the author
50:44 lesson above was above my pay grade I
50:47 know that that's why I said
50:49 um uh what's it called Google collab
50:55 so right now chat GPT like why I keep
50:58 pointing people here there's there's
50:59 exciting stuff going on like autonomous
51:02 agents and this thing called Lang chain
51:04 and these things that write novels but
51:07 they're all sort of on the geeky side of
51:09 the house you have to be more of an
51:10 engineer than just a user of Technology
51:15 um and like I'm I know just enough about
51:18 technology to get that Google collab
51:21 going for stable diffusion that I did
51:23 this with but it took me nine tries and
51:25 that wasn't like nine tries in one night
51:27 that was like over two months where I'd
51:30 spend like an hour or two I think I'd
51:32 get it running and then it would just
51:33 fail and I couldn't figure it out and
51:34 I'd be like [ __ ] it I did that nine
51:37 different times and then I finally got
51:39 it working I was like yes
51:41 it felt very accomplished but that's
51:43 that's just the the nature of where we
51:45 are in the in the evolution of this
51:48 technology right now chat GPT marks the
51:51 inflection point where normal people can
51:53 use this [ __ ] too
51:55 but we're only six months into that
51:58 right there's like 20 or 30 years of
52:01 Engineers doing machine learning stuff
52:04 and then there's all the engineers that
52:06 have sort of piled on since chat GPT
52:09 that are doing all sorts of cool [ __ ]
52:11 like revolutionary [ __ ]
52:13 but it ain't easy to use right so if if
52:16 you tend more to the geeky side you can
52:18 go play with some of that stuff if
52:20 you're just trying to figure out what in
52:22 the [ __ ] this means for our lives
52:23 welcome to the club I mean that's that's
52:26 why this channel exists
52:29 um I have zero coding experience by the
52:30 way
52:32 um I'm not a coder I understand code but
52:35 I have ADD I couldn't actually code I
52:37 couldn't I couldn't do it if you paid me
52:41 um I so would love to learn the coding
52:44 you said YouTube
52:46 taught you that
52:49 well no I'm just saying if you want to
52:51 learn how to use Google collab which is
52:54 the tool that a lot of these developers
52:56 are using for for the tools
52:58 um just go to YouTube and and search for
53:01 Google collab
53:02 starter lessons
53:05 uh thank you so much this helped me a
53:07 lot for author gbt great
53:09 that was not boring thank you so much
53:12 you're welcome
53:15 um doing great good
53:18 experience to creating the program each
53:22 program and course is unique because all
53:25 the pain points are unique my advice for
53:27 Leonardo prompts I'm TR wait oh any
53:29 advice for Leonardo prompts I'm trying
53:31 to use it
53:33 for concept art to do my fantasy so yeah
53:37 Milo here's here's let me go show you
53:39 something we'll we'll go look for this
53:41 together because I know there's one for
53:42 mid-journey Leonardo is stable diffusion
53:45 so I'm I'm going to I'm at poe.com
53:49 and I'm gonna go to the explore tab here
53:52 info.com and then
53:56 um I don't think we can actually search
53:58 for
54:01 let me search for a stable
54:05 no it didn't find any
54:08 um
54:10 let me see is there are there images or
54:12 is there art
54:13 so there's different categories of bots
54:15 well let me show you the mid Journey one
54:17 so the mid Journey one could help with
54:19 stable diffusion so I'll go to this mid
54:22 Journey bot and I'm going to clear the
54:25 context so basically what it says is
54:27 what would you like to prompt in
54:28 mid-journey today oh I know something I
54:30 can I can uh share with you
54:34 who asked about the
54:37 any advice Milo okay Milo I got a couple
54:40 of things for you okay so so I'd say I'd
54:43 say go to the PO Bots and see if you can
54:46 find a stable diffusion
54:48 prompt prompt bot that will help you
54:50 write stable diffusion prompts
54:54 is one thing but this is the mid-journey
54:56 one so so what I'm going to type is
54:59 um a let's see photo
55:02 of
55:04 a tabby cat
55:07 um
55:08 by Annie Leibowitz
55:14 any
55:16 so I type in a crappy prompt and then it
55:19 will rewrite it so it says color photo
55:21 of a tabby cat by Annie Leibowitz
55:23 piercing eyes soft for Curious
55:26 expression sitting on a windowsill so if
55:29 if you go into one of these prompt
55:30 rewriting Bots or or if you find a
55:33 Chachi PT
55:35 prompt for stable diffusion it'll kind
55:39 of do this [ __ ] for you so that's one
55:41 thing you can do
55:42 the other thing you can do is go to open
55:46 art
55:48 wait open our
55:50 dot a I slash promptbook
55:54 openart dot AI slash prompt book
55:58 and it will take you to this thing this
56:01 is the first time I saw this it it blew
56:05 my [ __ ] mind because this is a 96
56:08 page PowerPoint presentation basically
56:11 that really clearly explains you know
56:15 what stable diffusion is how it works
56:17 what prompt engineering is
56:20 um
56:22 what the what the structure of a prompt
56:24 is like is it a photo or a painting
56:26 what's the subject do you want special
56:28 lighting do you want it in a specific
56:30 environment do you have color schemes do
56:32 you have a specific point of view like
56:34 overhead or side or you know ground up
56:37 is there a background do you want it to
56:39 be a solid color or it to be living in a
56:42 thing and it just goes on and on and on
56:44 and explains you know about uh prompting
56:48 and then it also talks about things like
56:50 modifiers like these tools understand
56:53 all sorts of things about photography
56:56 they understand you know terms like tilt
56:58 shift and long exposure Color Splash
57:02 monochrome satellite
57:05 um and then it goes on and talks about
57:07 art mediums right do you want it to be
57:09 graffiti or watercolors or oil paintings
57:11 and this thing just goes on and on and
57:14 on and and
57:16 what what this document did for me is it
57:19 made it really clear just how much these
57:23 models understand
57:25 and that
57:27 and that if you if you want it to look a
57:30 certain way
57:32 if you can provide a reference for it
57:34 it'll probably help the other thing you
57:36 can do on Leonardo is you can
57:39 you can upload an image that will be the
57:44 um kind of the starter image to do
57:45 something else with it so you could
57:47 upload a picture of yourself and then
57:48 say I want you know that person in a
57:50 cowboy hat
57:52 right
57:59 [Music]
58:04 how would you use a bot is poe good for
58:07 coding pose pretty good for coding yes
58:08 but there's again there's there's a
58:10 bunch of different models there there's
58:12 Sage which is pose model there's GPT 3.5
58:16 gpt4 4 is great at coding
58:20 there's anthropics
58:23 Claude instant Claude instant 100K
58:27 they're okay at coding and then there's
58:29 Claude plus which I think is good at
58:31 coding so
58:33 um so yeah so so po can be good at
58:36 coding
58:37 I haven't used it for that I've just
58:39 used Chachi VT for it but all of them
58:41 are pretty good at it
58:44 all right
58:45 experience in creating wait experience
58:48 to creating a program
58:50 um
58:51 I mean each program and courses you need
58:53 okay sorry I saw these already
58:56 what do I think about Google deepmind I
58:59 think that the dude that co-founded
59:00 Google deepmind just got 1.3 billion
59:03 dollars for his new startup that's what
59:05 I think holy [ __ ]
59:08 that's what I think about about Google
59:11 deepmind I mean it's [ __ ] remarkable
59:13 right that Alpha fold thing is insane
59:16 it's insane I'm gonna create four
59:19 hundred thousand proteins it took him
59:23 like what
59:24 two weeks to to model one protein with
59:27 all these computers and then with
59:29 deepmind it did 400 000 in a couple of
59:31 days
59:33 it's just insanity
59:35 did I offer a course to teach all this
59:37 legit I would pay for that thank you
59:39 what's your name purely
59:41 Blissful Mama um
59:43 we talked about it last night on the
59:44 live I'm gonna put together a workshop I
59:47 think I'm gonna do I think so tell me
59:49 what you think about this
59:51 I think what I'm going to do is I'm
59:52 going to do a two-day workshop and it
59:54 would probably maybe it's like two days
59:55 on a weekend like 90 minutes or two
59:57 hours for each session the first one
1:00:00 would be like a basic primer like here's
1:00:02 the tools here's how they work here's
1:00:04 what prompting is here's image
1:00:05 generation here's chat GPT we'd figure
1:00:08 out like what what other you know
1:00:09 appropriate tools we so we do a primer
1:00:11 on day one
1:00:12 then I would have people go off and play
1:00:15 that night and come the next day with
1:00:18 something they're trying to solve in
1:00:19 their life like an actual problem and
1:00:21 then I think the second day would be
1:00:22 like a Hands-On primer where people
1:00:25 would talk about here's what I'm trying
1:00:26 to do and then we together we'd sort of
1:00:28 collaborate on how you might use these
1:00:30 tools tools to do that so that's what
1:00:31 I'm thinking about putting together so I
1:00:34 appreciate you saying that
1:00:37 um
1:00:39 Blue Willow generate images does unpaid
1:00:43 version of Poe have any value versus
1:00:45 chat GPT no well the Bots the bot maker
1:00:49 the freebot maker at Poe is just like
1:00:52 what I like about the bot maker is it's
1:00:55 a good way for you to practice prompting
1:00:57 because when you make a bot if your
1:00:59 prompt sucks the bot will suck or it'll
1:01:01 be inconsistent
1:01:02 and then you're like oh yeah like Steve
1:01:05 Jobs shouldn't talk about the fact that
1:01:07 he still works at Apple because he's
1:01:08 dead
1:01:09 I should go fix that right so like like
1:01:13 it just it's it's good for you to kind
1:01:15 of understand that if you get that
1:01:16 prompt dialed in right you can get a
1:01:18 thing to behave in a certain way and
1:01:20 then and and then you can take that
1:01:22 knowledge and you can apply that to your
1:01:24 business or for stuff that you do
1:01:25 personally so so it's why I have the
1:01:29 three of these here is that there
1:01:30 there's a reason to use all three of
1:01:32 them they're not exclusive like I you
1:01:35 know
1:01:36 if you never use Poe again is that going
1:01:38 to hurt you no if you just go to chat
1:01:40 GPT
1:01:42 and learn to prompt there and learn to
1:01:43 get good at that like I I've said this
1:01:46 before on this channel I I deeply
1:01:49 believe this
1:01:50 I could spend the next decade
1:01:53 just using GPT 3.5 and probably not even
1:01:57 scratched the surface of what that thing
1:01:59 can do
1:02:00 it's that deep it's just [ __ ] insane
1:02:04 like the deeper you dig the more
1:02:06 impressive it gets
1:02:08 stable diffusion mid-journey dolly two
1:02:10 it did get rained out oh okay the uh the
1:02:13 Xfinity race I wonder if they'll finish
1:02:15 it tomorrow they'll probably finish it
1:02:16 in the morning
1:02:19 foreign it's so funny watching NASCARs
1:02:23 going through Street courses they're
1:02:24 just such big lumbering things did you
1:02:26 watch watch the uh the the garage 56
1:02:29 NASCAR you know driving around the 24
1:02:31 Hours of Le Mans it was like me oh
1:02:37 and then it was faster than they thought
1:02:39 it was gonna be that was pretty awesome
1:02:43 we took America
1:02:45 America
1:02:46 took over Lamont that was pretty funny
1:02:50 um how would you prompt for t-shirt
1:02:52 graphics
1:02:53 um
1:02:55 I would I would probably just call it
1:02:57 t-shirt graphics let me I don't know
1:02:59 like like I would I mean I would say
1:03:04 you know illustration let me let me let
1:03:07 me do one right now just
1:03:08 slash imagine
1:03:11 um
1:03:12 illustration of
1:03:15 70s
1:03:18 t-shirt
1:03:22 [Applause]
1:03:24 um 70s t-shirt
1:03:28 graphic line art I'd use like line art
1:03:32 like I would I would really go I would
1:03:34 go look at that um
1:03:37 that prompt book
1:03:38 openart.ai prompt book I'd go look at
1:03:41 that and sort of pull out the styles
1:03:43 from that
1:03:45 so 70s t-shirt graphic line art of a
1:03:50 um
1:03:51 Surfer and a wave in
1:03:57 um I don't know uh pop colors let me
1:04:00 just see what that does if it does
1:04:01 anything interesting I'll show it to you
1:04:03 if it looks like [ __ ] I won't but but um
1:04:11 go to
1:04:12 um
1:04:15 there's a site called lexicon.art is it
1:04:18 dot art I think it's dot art hang on let
1:04:21 me go see
1:04:23 [Applause]
1:04:25 yeah lexicon.art go to lexicon.art and
1:04:29 type in let me see t-shirt
1:04:32 graphic
1:04:41 um yeah there's some here so if you go
1:04:44 to lexicon.art
1:04:45 and just type in t-shirt graphic or
1:04:48 t-shirt whatever you know
1:04:52 t-shirt graphics
1:04:55 t-shirt illustration
1:04:59 what's cool about Lexica is every image
1:05:01 that you find yeah like t-shirt
1:05:03 illustration is better here let me let
1:05:06 me let me show you these I'll show you
1:05:08 lexicon because it's this is worth
1:05:09 taking a look at
1:05:11 um
1:05:14 so lexica.art and then I just typed in
1:05:17 t-shirt illustration and so like there's
1:05:19 you know there's a
1:05:22 you know one that that's sort of a mood
1:05:24 board for a t-shirt you know there's
1:05:25 black and white ones that could be
1:05:27 interesting you could you could modify
1:05:29 those in Photoshop
1:05:32 um
1:05:33 what's cool about this is anything that
1:05:36 you find here where you like the style
1:05:38 okay so like this the skull dude here
1:05:42 any everyone that you click on it gives
1:05:44 you the prompt that generated that and
1:05:47 it shows you here's other things that
1:05:49 were generated with that prompt and then
1:05:51 you can also say open an editor and now
1:05:54 I can say generate and I can create my
1:05:57 own
1:05:58 let me log in here
1:06:04 uh wait where did it go it lost it
1:06:09 uh dang it come on man
1:06:14 open an editor hopefully I'm logged in
1:06:17 now generate okay so now it's generating
1:06:20 so now I'm taking that prompt that they
1:06:24 did and I'm generating it
1:06:27 um on my own so that's portrait of a
1:06:30 hipster skeleton so let's we'll see what
1:06:32 it gives us
1:06:36 and this this site lexic is not free
1:06:39 like you can do a certain they give you
1:06:41 a certain amount of credits you can play
1:06:43 like those are very very different than
1:06:45 what it gave there but you can also open
1:06:48 your advanced settings here and change
1:06:49 your creativity scale but I can I can
1:06:52 also do portrait of a hipster cowboy
1:06:58 art by butcher Billy vintage 80s anime
1:07:00 t-shirt design blah blah blah so let's
1:07:02 just do it so I'm just going to change
1:07:03 it to Hipster Cowboy instead of hipster
1:07:06 skeleton and we'll see what it what it
1:07:09 comes up with
1:07:20 so there you go hipster Cowboy are those
1:07:23 good t-shirts I don't know they're at
1:07:25 least in the neighborhood
1:07:26 I mean I kind of feel like this stuff
1:07:28 right now is not quite predictable
1:07:30 enough to be doing
1:07:32 commercial artwork consistently but I
1:07:35 think for doing things like concept
1:07:36 boards and concepting it's [ __ ]
1:07:38 brilliant
1:07:40 so anyway those those those should
1:07:42 hopefully help help a little bit help a
1:07:45 little bit he helped me a little bit
1:07:48 he was old he didn't really know how to
1:07:50 use Tick Tock but he he helped me a
1:07:52 little bit
1:07:53 I'm not as skilled as I was
1:07:57 all right
1:08:01 lost my place in the comments I lost my
1:08:04 place in the comments someone used to
1:08:06 write me a song that I can sing when I
1:08:08 lose my place in the comments there's
1:08:09 Thelonious talking about my haircut
1:08:13 um do me a favor share the live if you
1:08:15 would just so that we can get some other
1:08:17 people uh seeing it maybe tap the screen
1:08:19 give
1:08:20 pretend like you're enjoying this
1:08:23 maybe we can get some other suckers in
1:08:26 here we'll get them to use Chachi PT too
1:08:29 it's gonna be awesome
1:08:30 [Laughter]
1:08:37 [Music]
1:08:48 all right there was the prompt for
1:08:50 Leonardo is po good for dinner how would
1:08:52 you use a bot
1:08:54 on the how would you use a bot
1:08:57 I don't know I mean
1:09:00 like the mid-journey bot
1:09:02 like I used the mid Jersey bot now
1:09:04 because because I have ADD I didn't
1:09:06 create myself a little system to take
1:09:08 all of the different attributes of
1:09:10 mid-journey and build my own Mega prompt
1:09:12 for mid-journey
1:09:14 I just go to a mid-journey bot that
1:09:16 someone wrote that into a bot and so I
1:09:18 can just type in a shitty little prompt
1:09:19 and it will it will write me a better
1:09:22 mid-journey prompt than I would have
1:09:23 written and so
1:09:25 I use that bot in particular because I'm
1:09:28 lazy
1:09:29 um
1:09:30 the Shakespeare bot that I wrote like
1:09:33 rather than you
1:09:36 um figuring out how to write a prompt to
1:09:38 get chat EBT to write an iambic
1:09:40 pentameter you could just go to poe.com
1:09:42 I am Shakespeare and I've already
1:09:45 written that prompt for you so now you
1:09:46 can just interact with it and it will it
1:09:49 will answer you in iambic pentameter
1:09:50 even if you're talking about you know
1:09:52 business advice it'll give you business
1:09:54 advice as if Shakespeare wrote it and
1:09:58 why wouldn't you want that so that's how
1:10:00 you use Bots
1:10:01 and and also just the the other the
1:10:03 other reason to use Bots is just to go
1:10:05 explore them
1:10:07 and and explore
1:10:09 what this thing makes possible like how
1:10:12 is that thing not like Google in like an
1:10:15 infinite number of ways so if you go
1:10:17 play with 10 000 Bots there's ten
1:10:20 thousand different variations of prompts
1:10:22 that people put in there
1:10:24 that may inspire you to go oh [ __ ] I
1:10:26 never thought about it like that
1:10:29 and that's the stage we're at and I I
1:10:31 mean quite honestly here's the real
1:10:34 reason to use Bots anything that you can
1:10:36 do to get yourself a little more AI
1:10:39 literate a little more literate in
1:10:42 that's possible oh if that's possible
1:10:44 then I bet that's possible oh that is
1:10:46 possible if that's possible then I bet
1:10:48 that's possible the more sort of
1:10:50 synapses you can connect in your brain
1:10:53 to understand how deep this thing
1:10:55 [ __ ] runs then when your boss comes
1:10:58 to you or when you know someone comes to
1:10:59 you from your community and says I'm
1:11:01 trying to solve this problem I don't
1:11:02 even know where to start ah follow me
1:11:05 let's go to chat I know how to I know I
1:11:07 know where to start
1:11:10 one of the short Coatings of chat GPT
1:11:13 for coding is that it can't read lots of
1:11:14 lines of code or write lots of lines of
1:11:16 code yes that's why I pay for poe.com
1:11:20 because of the hundred thousand token
1:11:22 Claude but also
1:11:25 chat GPT with plugins so if you pay for
1:11:29 chat apt plus and use plugins
1:11:31 there are plugins now that will read
1:11:33 websites and read PDF files that are
1:11:36 long and we'll interact with them so
1:11:41 um and I would be surprised if there
1:11:44 weren't some coding Bots there the other
1:11:47 thing that you might want to play with I
1:11:48 don't know what the context window is
1:11:50 over at github's copilot but GitHub
1:11:53 co-pilot is chat GPT that's been
1:11:56 optimized for code
1:11:58 mm-hmm
1:12:00 and if if you're geeky enough that
1:12:02 you're playing with code
1:12:03 the
1:12:06 um
1:12:07 GPT 3.5 turbo now has a 16 000 token
1:12:12 context window so that's like 12 000
1:12:15 words so that's starting to get there
1:12:17 and then ultimately gpt4 is going to
1:12:19 have 35 000 or 32 000 tokens which is
1:12:22 like 25 000 words something like that
1:12:27 so yes
1:12:28 um and then and then the other thing
1:12:29 quite frankly if you want to get geeky
1:12:31 go start playing with Lang chain and
1:12:34 create yourself a vector database and
1:12:36 then you can dump giant piles of code in
1:12:38 there and have it read all that stored
1:12:39 in a vector database and
1:12:42 query till your heart's content and
1:12:44 don't worry about file length
1:12:48 um what do I think about Google deepmind
1:12:50 I think it's deeply impressive what do I
1:12:53 think about Google bard I think it sucks
1:12:55 ass
1:12:56 [Laughter]
1:13:00 well do I think Google will get better
1:13:02 yes I think Google will get better do I
1:13:04 think Google is behind right now yes
1:13:07 nothing comes close to gpt4 in my
1:13:09 opinion nothing's even close
1:13:11 and I heard like I've heard in the past
1:13:13 week or so that there's kind of rumors I
1:13:15 think George hotz if you know him he's
1:13:18 if you've ever watched if you want to
1:13:19 you want to see a [ __ ] Uber geek go
1:13:21 watch George hot's YouTube videos of him
1:13:25 coding when he was running his
1:13:27 um his autonomous driving startup that
1:13:30 he started out of his garage watch him
1:13:33 code I've never seen anything like that
1:13:35 that dude's brilliant but he was saying
1:13:38 he was saying on on an interview with
1:13:40 Lex Friedman this week that um it looks
1:13:44 like gpt4 might actually be like 16
1:13:48 different models stitched together like
1:13:49 it's it's not just a single thing I
1:13:51 think gpt4 they're doing some sneaky
1:13:53 [ __ ] on the back end to make it as good
1:13:55 as it is
1:13:57 um but but gpt4 is really [ __ ] good
1:13:59 and it's
1:14:01 you know
1:14:02 there's a reason they have an open
1:14:04 source that [ __ ] they figured some stuff
1:14:06 out that that other people haven't
1:14:07 figured out
1:14:09 um
1:14:11 I know nothing about tech but this is so
1:14:13 interesting I'm 55 I'm 58 go Gen X [ __ ]
1:14:17 everybody else
1:14:19 this is so here's the deal chat GPT
1:14:22 here's I I firmly believe this there's
1:14:25 two revenges happening right now revenge
1:14:27 of the liberal arts major because
1:14:29 horizontal thinking is going to become a
1:14:31 superpower vertical thinking is going to
1:14:33 get penalized because
1:14:35 these tools are going to do all the
1:14:37 vertical execution horizontal thinkers
1:14:39 critical thinkers are going to be good
1:14:41 so liberal arts majors and I also think
1:14:43 this is
1:14:45 revenge of the Gen xers
1:14:48 because we know how to [ __ ] use
1:14:49 computers but we're old enough to know
1:14:52 some [ __ ] so when chat GPT outputs
1:14:55 [ __ ] we can recognize it really
1:14:57 quickly and when it outputs Brilliance
1:14:59 we can recognize that really quickly so
1:15:01 I think liberal arts gen xers I'm going
1:15:04 to be doing really good
1:15:06 um you look fabulous thank you we're
1:15:07 both 57 keep on keeping on brother hey
1:15:10 I'm in grad school oh that's awesome man
1:15:12 I love that there's something for me
1:15:14 about just never-ending curiosity like
1:15:18 if I think about retirement like there's
1:15:20 nothing interesting about that to me
1:15:23 it's like well so when you retire
1:15:26 retired you stop being curious [ __ ] that
1:15:29 just keep being curious keep learning
1:15:31 and like we now have the most powerful
1:15:34 tool in the history of humanity we get
1:15:36 to play with it
1:15:38 It's [ __ ] amazing I know nothing
1:15:41 about tech uh but I can fix a carburetor
1:15:43 I I actually think people that can do
1:15:46 Hands-On [ __ ] are going to be way in
1:15:48 demand plumbers electricians people that
1:15:51 can fix carburetors
1:15:53 that's good because this shit's gonna
1:15:56 [ __ ] some stuff up the knowledge workers
1:15:58 are gonna be like I think I need to
1:15:59 learn Plumbing if they can't figure this
1:16:02 [ __ ] out
1:16:05 hey Kyle I subscribe to Mid journey is
1:16:07 it true that I can
1:16:09 I can sell its generated art
1:16:11 commercially I'm pretty sure that's true
1:16:15 I would go read there I would I would go
1:16:17 take their terms of service and pop it
1:16:19 into chat EBT or Poe and have it explain
1:16:22 what are actually in the terms right so
1:16:25 so I would assume so don't assume that
1:16:28 you can copyright your work right now
1:16:30 the copyright office is saying if it was
1:16:32 generated you know out of one of these
1:16:35 AI tools it's not copyrightable my gut
1:16:37 is that's going to change because
1:16:40 if you see anyone who actually knows
1:16:42 what they're doing with these tools it's
1:16:44 not like they put in one prompt and
1:16:45 outcomes Brilliance it's like they work
1:16:47 on Crafting those prompts for like
1:16:49 thousands of images and then they curate
1:16:52 out the best ones and then they take
1:16:53 that into Photoshop and they [ __ ] with
1:16:55 it and they change the composition and
1:16:57 then
1:16:58 most of the stuff that of people that I
1:17:01 know are doing this are all sorts of
1:17:03 hybrid you know between Ai and human
1:17:05 contribution so I think that's going to
1:17:08 shift but don't assume that you can
1:17:11 um
1:17:13 that you can copyright it I don't know
1:17:15 that you necessarily need to because we
1:17:18 now have an infinite art generating
1:17:20 machine
1:17:22 um that it can generate [ __ ] in 30
1:17:24 seconds
1:17:25 um the other thing is if you're in a
1:17:27 commercial environment meaning if you
1:17:28 work for a big company
1:17:30 I would say don't use mid-journey or
1:17:33 stable diffusion because they've been
1:17:35 trained on copywritten work and there's
1:17:38 there's already lawsuits happening and
1:17:39 there's going to continue to be lawsuits
1:17:41 happening I think ultimately that [ __ ]
1:17:43 will settle out but if you're working
1:17:45 for a big company they don't want that
1:17:46 headache
1:17:48 um so use Adobe Firefly or generative
1:17:50 fill within Photoshop
1:17:53 um until further notice if you're in a
1:17:55 if you're in a corporation but I'm
1:17:57 pretty sure you can sell what you do and
1:17:58 and listen quite frankly Leibovitz uh
1:18:01 Thelonious helped me
1:18:04 um
1:18:05 spell any Leibovitz Leibowitz correctly
1:18:09 um
1:18:11 um what was it gonna say
1:18:14 oh just like everyone's putting [ __ ] out
1:18:18 there and selling it yeah so I think you
1:18:20 can sell it I don't I don't who's going
1:18:21 to come after you I don't think anyone's
1:18:22 going to come after you but I'm not a
1:18:24 lawyer I have a degree in acting so
1:18:26 don't listen to me for that [ __ ] go read
1:18:28 the terms of service yourself or have
1:18:30 chat GPT explain it to you that's what I
1:18:31 would do uh and then uh I wouldn't worry
1:18:35 too deeply about it but if you want to
1:18:36 copyright things I'd go look at the
1:18:38 there's a case of an artist from England
1:18:41 that did a graphic novel
1:18:43 um where the copyright Department said
1:18:45 the words or cup like the the work
1:18:48 itself is copyrightable what the images
1:18:50 are I don't know there's some [ __ ] going
1:18:52 on it's go learn about it uh it's your
1:18:55 art okay good that's the cool thing
1:18:57 about AI everything is created with it
1:18:59 is unique mostly yeah exactly it's
1:19:02 there's just there's lots of great areas
1:19:04 right now but we're so early that like
1:19:06 especially if you're just doing work
1:19:08 like you want to put out t-shirts or you
1:19:09 want to create patterns and sell them on
1:19:11 Etsy I know tons of people doing that no
1:19:14 one's gonna come after you
1:19:15 I don't think
1:19:17 I I feel like I can't make declarative
1:19:20 statements here because if I'm wrong I I
1:19:22 don't want you to come say I saw some
1:19:23 [ __ ]
1:19:25 like go I am not your lawyer I'm not
1:19:28 even an internet lawyer
1:19:32 I'm just a dude trying to figure this
1:19:33 [ __ ] out
1:19:35 I helped a mom who sells concert
1:19:37 t-shirts come up with art on mid-journey
1:19:39 for a small Biz that's freaking awesome
1:19:40 love that what's the best AI tool right
1:19:43 now to write novels or long-form text
1:19:45 one is author GPT what else
1:19:47 um
1:19:48 I'd go play with the Claude 100 000
1:19:51 token thing because it's because it's
1:19:53 got a hundred thousand tokens that's
1:19:54 like a 75 000 word context window it'll
1:19:58 remember farther back in your in your
1:20:00 session conversation so if you if you're
1:20:03 working on some long-form stuff it'll
1:20:05 write it but it'll also remember more of
1:20:08 the early part of your conversation so
1:20:10 like if you write an outline of a novel
1:20:13 then you have it right chapter one and
1:20:15 then you modify chapter one and do two
1:20:18 or three versions of it and then you go
1:20:20 to write chapter two if you've got a
1:20:22 short context window it'll kind of
1:20:23 forget your outline so that can get
1:20:25 really frustrating so I'd go to I I'd
1:20:28 use the
1:20:29 100
1:20:30 000 token version of Claude you do need
1:20:33 a post subscription for that so it's
1:20:35 like 20 bucks
1:20:36 um but I'd play with that there
1:20:39 um
1:20:41 there are probably if you go to Future
1:20:43 pedia there are probably sort of
1:20:45 pre-written
1:20:47 prompt kind of tools out there that are
1:20:50 that are designed for long-form fiction
1:20:52 or novel writing at futurepedia.io but
1:20:56 but
1:20:58 they are likely just
1:21:00 kind of rappers sitting on top of gpt4
1:21:04 so
1:21:07 like like I wouldn't necessarily pay
1:21:09 someone
1:21:11 a monthly fee for a tool where they've
1:21:13 just kind of written [ __ ] on top of chat
1:21:15 GPT that like a lot of the stuff that's
1:21:18 here is that kind of [ __ ]
1:21:20 there might be tools out there that do
1:21:23 long-form stuff really well and they're
1:21:25 and if like I would say if if they've
1:21:27 been in if they've been launched in the
1:21:30 past year I'd look at them very
1:21:33 suspiciously if they've been launched in
1:21:35 the past six years and they really
1:21:37 specialize in the kind of writing you
1:21:39 want to do there might be something
1:21:40 there so it's worth it's worth at least
1:21:42 looking there
1:21:44 is my go-to for long form yeah exactly
1:21:47 it's all about the prompts it is all
1:21:48 about the prompts oh the other thing the
1:21:50 other thing about uh any sort of
1:21:52 long-term writing don't feel like you
1:21:54 need to write the single Mega prompt
1:21:56 that does everything break it into much
1:21:59 smaller sections think of working with
1:22:01 Chach EBT more like having a
1:22:04 collaborative conversation with a
1:22:05 writing assistant let's say you've got a
1:22:07 really good intern
1:22:09 who just knows [ __ ] everything that's
1:22:11 what chat EBT is like so like a good
1:22:13 intern they need a lot of coaching so I
1:22:16 want you to act like a writing assistant
1:22:18 and I want you to be really good at
1:22:19 grammar and I want you to
1:22:22 um
1:22:22 have deep knowledge of this particular
1:22:26 um part of History because that's what
1:22:28 my book is set in and just go back and
1:22:31 forth with it do you understand yes I
1:22:33 understand and here's some thoughts and
1:22:35 I want you to be good at helping me
1:22:37 um refine my outline can you help me
1:22:40 with that sure I can help you with that
1:22:42 and here's three different you know
1:22:44 storytelling Frameworks we could
1:22:45 consider all that back and forth makes
1:22:48 it better and better and better as you
1:22:50 work through
1:22:51 um your thing but just remember with
1:22:53 chat CPT because it's only got a 4 000
1:22:56 token context window it may forget
1:22:58 earlier parts of the conversation so you
1:23:01 may have to go back in the conversation
1:23:03 copy little pieces and paste them back
1:23:05 down lower to remind it the [ __ ] that it
1:23:08 forgot
1:23:09 and that's again why I would I would
1:23:10 it's worth exploring
1:23:13 um the hundred thousand token uh Claude
1:23:16 model here
1:23:17 knowing that that's not as good a writer
1:23:20 as gpt4 is so maybe you do your long
1:23:22 form [ __ ] here and then you go to gpt4
1:23:25 to refine it might be it might be a play
1:23:29 response to the course oh response to
1:23:32 the course that sounds great like a
1:23:34 great course I definitely want to learn
1:23:36 it all okay great I'll definitely I'm
1:23:38 gonna definitely put together some sort
1:23:40 of workshop and once I get it together
1:23:42 I'll talk about it here
1:23:45 um
1:23:46 you're good now if I start Plastering
1:23:48 Nike signs all over the merch watch out
1:23:50 yeah exactly I'm actually working
1:23:52 there's a guy that that joined the AI
1:23:54 Salon he's been in the Garment business
1:23:56 for like 20 years and he's starting an
1:23:59 AI t-shirt business and and he's like
1:24:04 he's sort of a high-end dude he's got
1:24:05 you know he's done like Disney licenses
1:24:08 and sports team licenses and he's trying
1:24:11 to figure out how to use use these tools
1:24:13 to to do truly personalized graphics for
1:24:16 big licensed things the tech isn't quite
1:24:19 there yet to make it really predictable
1:24:21 but I I think it I think it'll come like
1:24:24 you wait because I because this is gonna
1:24:26 happen this is going to be [ __ ] Wild
1:24:29 Disney will train their own models and
1:24:33 and allow you to pay them to generate
1:24:36 things that look Disney like trained on
1:24:38 so it'll be like officially licensed
1:24:42 original generated stuff so you can do
1:24:45 truly generated or truly original Disney
1:24:49 Art licensed by Disney
1:24:51 that will come probably sooner than
1:24:54 later my prediction two years because
1:24:56 Disney is an incredibly
1:24:59 um technically Savvy company
1:25:02 and man do they know how to make money
1:25:04 on you know leveraging the [ __ ] out of
1:25:06 their their IP
1:25:08 laughs
1:25:10 what's going to be interesting about
1:25:12 when Disney does that like why they
1:25:14 might not do that is
1:25:16 like the restrictions for what you can
1:25:19 actually
1:25:20 share with the world they might lock
1:25:22 that [ __ ] down because they don't want
1:25:23 you to do you know Cinderella in some
1:25:26 compromising position which it's the
1:25:29 first thing everyone's gonna do right so
1:25:31 uh
1:25:33 uh uh
1:25:37 never a dull moment when things are
1:25:41 changing when everything changes what do
1:25:43 you do you go figure it out yeah
1:25:49 um that's another thing I've said this
1:25:52 before here but especially with this
1:25:55 generative AI stuff the new stuff right
1:25:56 not not sort of old school machine
1:25:58 learning and old school meeting older
1:26:01 than six months ago anyone that's
1:26:03 talking about this stuff right now
1:26:05 that's that that is proclaiming that
1:26:07 they are an expert is full of [ __ ] no
1:26:09 one's an expert right now everything's
1:26:11 changing and it's changing fast so
1:26:14 what's going to be brilliant today is
1:26:17 going to be a total snooze three months
1:26:19 from now it is going to be completely
1:26:20 irrelevant six months from now that's
1:26:23 just the nature of the Beast so again I
1:26:25 think the most important thing to learn
1:26:26 right now is like curiosity exploration
1:26:29 and critical thinking
1:26:30 just learn how this [ __ ] works and like
1:26:33 how you like using it and then just keep
1:26:35 learning and keep learning and keep
1:26:37 learning because
1:26:39 too too much changing too fast to keep
1:26:41 up with it
1:26:44 all right
1:26:47 [Music]
1:26:52 I lost my place in the comments
1:26:57 you gave me a panda for singing thank
1:26:59 you I think
1:27:00 ha
1:27:02 and a rose I'm getting all sorts of
1:27:05 gifts here
1:27:07 can you share the salon
1:27:10 can you share your the salon.ai on
1:27:13 LinkedIn I can't reach the salon.ai oh
1:27:17 don't put in HTTPS
1:27:19 hang on
1:27:21 unless it's it's down maybe I didn't pay
1:27:24 a bill to a domain register the salon.ai
1:27:30 it's it's working
1:27:34 let's see
1:27:38 hang on
1:27:42 it just redirects if you go if you go to
1:27:45 link tree like uh linked
1:27:49 tr.e a i artist
1:27:53 artists Salon that will also take you
1:27:57 there thank you for the heart
1:28:01 um
1:28:02 can you share your oh on LinkedIn
1:28:06 but where am I going to share it on
1:28:07 LinkedIn oh just post it okay yeah hang
1:28:09 on hold hold please
1:28:15 um
1:28:15 post to anyone
1:28:17 um
1:28:19 interested in a community of AI
1:28:23 enthusiasts
1:28:29 join us
1:28:34 the salon dot AI
1:28:43 I'm just gonna make sure that works
1:28:46 yep all right so if you go to LinkedIn
1:28:50 and you go to like search for me Kyle
1:28:52 Shannon on LinkedIn my latest post is a
1:28:56 link to
1:28:57 the salon
1:28:59 link tree
1:29:01 I think it's also linked in my profile
1:29:04 can you share okay I did that
1:29:06 um
1:29:08 can you should it I don't know what that
1:29:10 means gotta upgrade ASAP HP or Dell two
1:29:14 and one touchscreen laptop Amazon best
1:29:17 is somebody putting ads in here for
1:29:19 computers don't do that [ __ ]
1:29:21 I'm trying to learn AI stuff here
1:29:26 wait what is it are you actually asking
1:29:29 something gotta upgrade
1:29:32 64 gigabyte RAM Best Buy Office Depot
1:29:35 Max you might be asking a question if
1:29:38 you're asking like what's the what's the
1:29:40 best computer for generative AI stuff
1:29:43 it depends what you want to do if you
1:29:45 want to be like running models locally
1:29:47 and you want to be training stuff and
1:29:48 things like that then
1:29:50 you need to get into gpus and [ __ ] like
1:29:53 that I don't know I'm a Mac guy so I I
1:29:55 don't know what what to do if you're
1:29:58 just going to be a user of the tools
1:30:00 like
1:30:02 open Ai and Google and meta are they're
1:30:07 they're building all the super computers
1:30:08 so that we don't have to like they're
1:30:11 training all these models and they're
1:30:12 getting them increasingly smaller so
1:30:14 probably within two years you're going
1:30:17 to be able to install a very small file
1:30:19 on your phone or on your current laptop
1:30:22 that's as good as gpt4 and it doesn't
1:30:25 even need to connect to the internet
1:30:27 so if if you're actually asking a
1:30:29 question here it seemed like you were
1:30:31 trying to sell something but I think
1:30:32 you're actually asking a question I
1:30:34 don't know like what I would do is if if
1:30:37 you need you need to understand what
1:30:39 you're actually trying to accomplish on
1:30:41 your machine if you want to do training
1:30:42 and model building and start playing
1:30:45 with open source powerful models then
1:30:47 you're going to want a kick-ass machine
1:30:50 I would go to Reddit and go to the open
1:30:52 source large language model Community
1:30:54 open source llm
1:30:57 just go search on Reddit
1:30:59 um and go find a subreddit there that
1:31:02 where everyone's geeking out on what
1:31:03 machines you use
1:31:05 um and good luck buying a GPU right now
1:31:07 because everybody's [ __ ] buying them
1:31:09 for their super computers
1:31:11 um but you know
1:31:17 for your mid-journey art this is this is
1:31:20 how I help my wife Creator graphics for
1:31:22 her store that's super cool canva plus
1:31:24 is totally worth it yeah canva canva um
1:31:28 if if you're doing anything with these
1:31:30 Graphics are you taking like the mid
1:31:31 Journey graphics and putting them into
1:31:33 like you know things that you're going
1:31:34 to sell on Etsy you can do all sorts of
1:31:37 cool [ __ ] on canva like you can do bulk
1:31:40 create where you create you could create
1:31:42 a whole bunch of t-shirt sayings on
1:31:45 um in chat GPT put them in a spreadsheet
1:31:49 and then just have canva bulk create
1:31:51 them
1:31:52 pretty slick stuff
1:31:54 Gen X is everything at least we think so
1:31:57 we're like leave us the [ __ ] alone and
1:32:00 and we're awesome
1:32:02 thank you
1:32:07 why is Gen X so angry some of the videos
1:32:10 on Gen X right now are just [ __ ]
1:32:12 genius
1:32:16 never retire lifelong learner here yeah
1:32:18 man you just gotta keep going like why
1:32:20 would you stop using this brain we're
1:32:22 given access to and and honest to God
1:32:24 man this is
1:32:26 it's hard to overstate how [ __ ] lucky
1:32:30 we are to live at a time where we get to
1:32:32 play with this [ __ ]
1:32:35 like think about it the past 70 years of
1:32:39 of computer infrastructure and
1:32:42 innovation
1:32:45 has gotten us to the point where some
1:32:48 [ __ ] nerds figured out how to point
1:32:50 those machines at all of the output of
1:32:53 humanity or at least as much as we can
1:32:55 see on the internet right now and turn
1:32:58 it into like this this blob of knowledge
1:33:02 that you can just [ __ ] tap into
1:33:06 [ __ ] like holy [ __ ]
1:33:10 we're entering Post sci-fi do you do you
1:33:14 realize that like what this [ __ ] is
1:33:17 is like this [ __ ] is like the Holodeck
1:33:20 like this is the foundational [ __ ] for
1:33:23 the Holodeck and curing cancer and
1:33:26 curing genetic diseases and curing
1:33:28 climate change and solving the energy
1:33:31 crisis like like it's it's all [ __ ]
1:33:35 in here and like we and and like Sam
1:33:38 Altman put it in a format
1:33:41 that a hundred million people signed up
1:33:43 to use it in six weeks and forced every
1:33:46 other company on the planet to go holy
1:33:49 [ __ ] we got to do our version of that so
1:33:51 the advancements
1:33:52 like the advancements ain't going away
1:33:55 it ain't getting uninvented so so we're
1:33:57 at the at the at the very beginning of
1:34:00 a transformation of humanity like holy
1:34:04 [ __ ]
1:34:08 we had to play with this stuff
1:34:12 I I mean it's it's I don't know I like
1:34:15 when people when I when I hear people
1:34:18 that are like well I hear it makes
1:34:20 mistakes
1:34:22 oh yeah okay
1:34:24 have you used it
1:34:28 well so what it wrote a poem have you
1:34:30 tried to write a poem have you tried to
1:34:32 make two words rhyme in every line of 16
1:34:36 stanzas ever it's [ __ ] hard
1:34:40 it'll take you two hours and your poem
1:34:43 will suck
1:34:44 it does it in like two seconds
1:34:48 I don't know I don't know I just I don't
1:34:51 get how people don't get how big a deal
1:34:54 this is I mean listen uh some obviously
1:34:57 do because 100 100 million people are
1:35:00 like this this is something
1:35:03 for a new person doing GPT
1:35:06 um which one for copywriting that one
1:35:08 the one the one floating over my head
1:35:10 for all of for all of the lives that I
1:35:13 do that's the one that's the one that's
1:35:15 the copywriter and if you can swing it
1:35:18 pay for gpt4 and if you can't afford
1:35:20 that use GPT for it bing.com just
1:35:23 bing.com's chat is gpt4
1:35:27 but gpt4 here
1:35:30 is the copywriting one it is
1:35:33 staggeringly powerful
1:35:36 is this the dude sorry I've got a I've
1:35:39 got a boxing match on I'm waiting for
1:35:41 some dude who's undefeated he's he's
1:35:43 knocked out 14 of 14
1:35:46 got people that he's boxed and he's
1:35:48 boxing tonight
1:35:49 but he's not on yet
1:35:54 um
1:35:58 I thought it was about 30 a month I
1:36:00 don't know what you're talking about
1:36:01 chat GPT is 20 a month I think
1:36:04 Paul might be 30 a month I don't know I
1:36:06 don't know how much they are
1:36:08 I'm bad with money if you pay for the
1:36:10 chat GPD plus you'll get access to the
1:36:12 plugins they have a decent copywriter
1:36:14 that's cool yeah there's there's so many
1:36:16 plugins now I don't I don't know which
1:36:17 ones there are I have no idea what a
1:36:20 plug-in is I have so much to learn okay
1:36:21 Precious Crow let's
1:36:23 let's get you up to speed on what a
1:36:25 plug-in is hopefully you're still here
1:36:29 and I'll teach you how to turn them on
1:36:31 so so everyone who's wondering
1:36:34 um you know is chat GPT worth paying for
1:36:37 let me show you what you get for your 20
1:36:38 bucks
1:36:39 so this is chat GPT this is your chat
1:36:42 history here
1:36:43 for twenty dollars you get that little
1:36:45 graphic that says Plus
1:36:48 okay and then what you also get is you
1:36:50 get not only GPT 3.5 or what they call
1:36:54 uh GPT turbo they think of the free
1:36:57 version
1:36:58 you also get gpt4 gpt4 is just of a
1:37:02 different level there's nothing out
1:37:03 there like it and and the one here at
1:37:05 open AI is really good but you also get
1:37:08 um browse with Bing which is basically
1:37:10 gpt4 connected to the internet so one of
1:37:12 the limitations
1:37:14 of gpt4 is that the last data in its
1:37:20 core model is September 2021 so
1:37:23 sometimes you'll ask it something that's
1:37:25 like as a large language model that you
1:37:28 know last did something in 2021 I don't
1:37:30 know what the [ __ ] you're talking about
1:37:32 so the one that's connected to the
1:37:34 internet solves that because if it can't
1:37:36 find anything in its core model it will
1:37:38 go search the internet and then add that
1:37:40 to the answer basically and then you've
1:37:41 got plugins now
1:37:43 if you've paid for chat GPT and you have
1:37:45 your little yellow tag here but you
1:37:47 can't see that stuff that's because you
1:37:49 have to enable it manually and how you
1:37:51 do that you go to your settings
1:37:54 and then in your settings there's this
1:37:56 thing called beta features and you go
1:37:57 there and you turn them on so browse
1:37:59 with Bing you turn it on plugins you
1:38:01 turn it on so that's how you get access
1:38:03 to them now what are plugins
1:38:06 think of plugins as like little
1:38:09 tunnels out to the internet that do
1:38:13 specific things
1:38:15 so
1:38:17 when you when you choose uh gpt4 plugins
1:38:20 you get this additional little menu and
1:38:24 in the menu you have you've got whatever
1:38:26 plugins you've got installed
1:38:29 but just because you have plugins
1:38:31 installed doesn't mean they're actually
1:38:32 activated so in any given chat session
1:38:35 so when I log into chatgpt and I'm going
1:38:37 to do something if I want to
1:38:41 have GPT help me with a meal plan
1:38:44 I need to actually turn on the plugins
1:38:47 that are related to that or if I want to
1:38:49 do travel I would do that or if I want
1:38:51 to have it read a website for me I would
1:38:53 have it do that
1:38:55 so you install plugins and then in any
1:38:57 any given session you can activate you
1:38:59 can enable up to three of them so let me
1:39:02 I'm going to do one here
1:39:05 that is
1:39:06 um
1:39:08 just kind of a fun one called prompt
1:39:10 perfect
1:39:11 so we've got one plug-in here called
1:39:13 prompt perfect and if I go look at what
1:39:15 it says it says type the word perfect
1:39:17 before you know before your prompt and
1:39:20 it'll craft the perfect prompt every
1:39:22 time so now I can come down here if I
1:39:24 type in the word perfect
1:39:26 and then I type in a shitty prompt um
1:39:29 right right right
1:39:31 me a Mia me a short story
1:39:36 that's a horrible prompt because it's
1:39:38 way too generic and it'll give me
1:39:40 crappy output if fires off prompt
1:39:43 perfect because I put that as the first
1:39:45 word
1:39:47 and then let's see what it did so it
1:39:49 changed write me a short story too
1:39:51 please write a concise and
1:39:53 well-developed short story that includes
1:39:55 vivid descriptions engaging characters
1:39:57 in a compelling plot with a satisfying
1:40:00 resolution
1:40:02 um
1:40:03 ideally it would have added some sort of
1:40:05 specific character ask me about it but
1:40:07 it's going to give me something way
1:40:09 better than what I did so so
1:40:12 chat EPT knows
1:40:15 based on what you type in if it needs to
1:40:17 use a plug-in or not and which plug-in
1:40:19 so let's go fire off another plug-in
1:40:22 here that might be a little more
1:40:24 interesting so I'm going to turn off
1:40:26 prompt perfect
1:40:28 and I'm going to turn on
1:40:31 let's do uh this this is one I've showed
1:40:33 here before it's kind of fun I'm going
1:40:34 to turn on instacart Okay so instacart
1:40:38 the grocery ordering thing so now I'm
1:40:40 going to say
1:40:41 um
1:40:42 I
1:40:44 want you to act like a
1:40:49 um
1:40:50 X like an expert nutritionist
1:40:56 and give me a meal plan
1:41:02 designed for intermittent
1:41:06 fasting and low carbs
1:41:11 um but
1:41:13 don't be boring
1:41:17 and so now it's gonna it's gonna come up
1:41:19 with a meal plan for me sure I can help
1:41:21 here's a Seven Day meal plan for
1:41:23 intermittent fasting with low carb meals
1:41:24 plans designed for 16-8 fasting schedule
1:41:27 breakfast skip breakfast lunch grilled
1:41:30 chicken salad dinner baked salmon
1:41:33 stir fry with a variety of colorful
1:41:36 vegetables skip breakfast
1:41:39 um
1:41:40 let's see I'm gonna I'm gonna stop
1:41:43 generating here I'm gonna say all of
1:41:45 wait
1:41:46 all of those
1:41:50 are boring
1:41:52 give me
1:41:55 exotic options
1:41:58 [Laughter]
1:42:00 so there's another
1:42:02 chat EVT is is like if you don't like
1:42:05 what it gave you tell it to give you
1:42:06 something else Moroccan spiced grilled
1:42:09 chicken
1:42:10 that's good Indian style shrimp Curry
1:42:13 Mexican style grilled fish Middle
1:42:15 Eastern salad Korean style beef okay
1:42:18 this is good
1:42:20 so so these are five okay
1:42:23 now notice what it did at the end here
1:42:25 it says would you like me to create a
1:42:26 shopping list for these meals using the
1:42:28 Instagram plugin so I can or instacart
1:42:30 plug-in so I'm just going to say yes
1:42:32 so all I say is yes
1:42:34 and now what it's going to do is it's
1:42:35 going to fire off the plug-in
1:42:39 it's gonna understand the meal plan that
1:42:42 was created up there
1:42:46 and it's going to create me a list and I
1:42:48 can have this stuff delivered tonight
1:42:49 this could be on my porch in a couple
1:42:51 hours
1:42:55 and look you can also open up these
1:42:57 plugins and and so
1:43:01 ingredients Thai beef fresh herbs lime
1:43:04 dressing so so it it's sent it sent a
1:43:07 bunch of stuff over to instacart
1:43:12 and then
1:43:14 instacart sends back a list
1:43:17 here's your exotic low-carb meal plan
1:43:19 and I click on that
1:43:26 and then instacart come on instacart you
1:43:30 can do it bang bang bang
1:43:33 there's all my stuff
1:43:39 Uncle Ben's ready rice
1:43:41 it's giving me Bird's eyes let's give me
1:43:43 some awful crap in here but Miracle Whip
1:43:46 give me a Miracle Whip
1:43:49 what trash you two can be watch Trash
1:43:53 I do loves me some Miracle Whip
1:43:57 but anyway so so that's so that's
1:44:00 plugins oh and then let me show you one
1:44:01 other thing about plugins
1:44:04 um
1:44:05 there are a lot of them so if I go over
1:44:08 here I go to I go to GPT plugins I go to
1:44:10 my list so let's say I'm bored with all
1:44:12 my plugins
1:44:13 so I go to the plug-in store
1:44:16 and it'll take a minute to load here
1:44:20 and then if I click on the all button
1:44:23 it'll now show me there's so there's a
1:44:26 hundred and fifty pages of of four
1:44:29 plugins each
1:44:30 or however big your screen is that you
1:44:32 know you might have more but that
1:44:34 basically means there's 600 of these
1:44:35 things
1:44:37 um and if I wanted to do something for
1:44:39 uh what we were talking about let me see
1:44:41 if there's anything for stable diffusion
1:44:42 in here
1:44:44 yeah
1:44:45 mixer bot image gen generates AI images
1:44:48 via dolly 2.
1:44:51 create perfect prompts for mid-journey
1:44:54 and stable diffusion pictures so I'm
1:44:56 going to install this and let's let's go
1:44:57 play with it because this was something
1:44:58 we were talking about earlier
1:45:00 so so now that that's installed I'm
1:45:02 going to turn off instacart because we
1:45:04 don't need it
1:45:05 and then I'm going to say
1:45:07 um
1:45:09 uh image
1:45:13 of
1:45:15 70s this is my go-to prompt muscle car
1:45:20 in abandoned Factory
1:45:27 let's see what it does
1:45:29 I think this may actually send this
1:45:31 prompt out to Dali and and pull the
1:45:34 image back in here and then hopefully
1:45:36 it'll it'll rewrite the prompt for us
1:45:38 here's the AI generated image
1:45:42 also if you're interested in exploring
1:45:44 more
1:45:50 yeah so so so what a lot of these
1:45:52 plugins are this is basically just an ad
1:45:54 for their website which is interesting
1:45:56 let's go see what what image it created
1:45:57 and if it actually gave us so it's kind
1:46:00 of a crappy Dolly image it just gave us
1:46:02 one and it didn't give us the prompts
1:46:04 like it told us it was going to so
1:46:06 there's an example of a shitty
1:46:09 crappy useless plug-in and there's
1:46:12 there's lots of those this is basically
1:46:14 just an ad for their mixerbox plugins
1:46:17 website of mixerbox plugins
1:46:19 so a lot of the plugins are just crap
1:46:21 crap crap crap don't be fooled by fancy
1:46:25 Graphics because everyone can make fancy
1:46:27 Graphics now and make their
1:46:29 service seem like it's not a giant
1:46:31 steaming pile of poo poo but that was a
1:46:35 giant steaming pile of poo poo
1:46:39 and every another very cool plugin using
1:46:42 chat GPT for his stories yeah stories is
1:46:44 very cool it generates
1:46:46 um kids books
1:46:48 um stories though like the images like
1:46:51 their image prompts are not great they
1:46:53 got it they got some work to do on their
1:46:54 image prompts but it's a very cool it's
1:46:56 a very cool thing
1:46:58 um I wonder if this can write for papers
1:47:00 that said I wonder if you can write for
1:47:01 papers Etc with this yeah I think that
1:47:04 you'll be able to do
1:47:06 um journalism especially if you're doing
1:47:07 things like um
1:47:09 if you use Bing that's connected to the
1:47:11 internet if you use chat GPT both of
1:47:14 those do citations decently well like it
1:47:17 will give you an answer about a current
1:47:18 topic and then if it if it used a
1:47:21 particular uh news source it'll it'll
1:47:24 cite that and there's going to be all
1:47:27 sorts of tools built to to do that
1:47:29 really really well that it kind of sucks
1:47:30 at that right now but I think I think
1:47:32 it's going to be
1:47:33 it's going to be powerful the other
1:47:35 thing I was I was on a panel with a guy
1:47:37 who's been in the newspaper business for
1:47:39 many many years and he was talking about
1:47:41 an idea about you could take a lot of
1:47:44 local political
1:47:47 um like like
1:47:48 um
1:47:49 documents like like when someone
1:47:52 um applies for a building permit those
1:47:55 are all public records right and a lot
1:47:57 of what local journalism is is digging
1:48:00 up public records and figuring out
1:48:02 what's actually going on there well
1:48:04 local journalism's been decimated by all
1:48:07 these you know National
1:48:09 um private Equity firms buying up all
1:48:12 the local markets and controlling the
1:48:13 messaging
1:48:15 they would be relatively straightforward
1:48:17 to create an AI journalist that went and
1:48:20 looked at public records and wrote
1:48:23 stories about them you know based on how
1:48:26 local journalists used to do it
1:48:29 that's a pretty powerful [ __ ] idea so
1:48:32 it's not about the writing like it's
1:48:34 more about the reporting right and if
1:48:37 you could create tools that go to the
1:48:39 public records pull the public records
1:48:41 find anything interesting in them and
1:48:43 write those stories and then have
1:48:44 basically an editorial staff that's
1:48:47 going yeah that's a good story let me
1:48:48 dig deeper into that that could be
1:48:50 something that's really interesting
1:48:53 he's explaining stair step prompting yep
1:48:56 little little small prompts
1:48:59 add up to Big results like Vice and
1:49:02 local magazines yep exactly
1:49:08 more you mold chat GPD prompts refine
1:49:11 have others do better give it examples
1:49:13 of your work yep Runo that's that's a
1:49:17 really good point
1:49:18 um
1:49:20 yes think of a chat GPT interaction more
1:49:24 like a conversation than writing the
1:49:26 perfect prompt
1:49:28 and and it it's it is really good like
1:49:31 if you want it to write like you rather
1:49:33 than describing how you write give it a
1:49:36 piece of your writing say here's here's
1:49:39 something that I wrote that captures my
1:49:41 tone perfectly paste
1:49:43 now I'd like you to write a story about
1:49:45 chubby dogs that love meaty goodness in
1:49:50 that style and it'll do it quite well
1:49:52 it's really good at that
1:49:54 and then if it's not if it gives you
1:49:56 something bad you can go no you
1:49:58 completely missed it it's way too wordy
1:50:01 and you're using and you haven't used a
1:50:03 single metaphor and all of my writing
1:50:05 has lots of metaphors in it please
1:50:07 rewrite this to be concise and use Clear
1:50:09 metaphors bang out of little come
1:50:13 like like tell it you like smack it when
1:50:17 when chechi TV's bad chat gbt is bad
1:50:21 smack it and tell it's been a bad boy
1:50:24 bad bad boy and it'll get better
1:50:29 that's badass gpd5 is going to change
1:50:31 his life
1:50:33 I don't know whose life
1:50:35 um
1:50:36 I'd pay for a beginner Workshop okay
1:50:39 um I think we see you're talking uh wait
1:50:43 I think we see what you're talking about
1:50:45 by the middle of next year I don't know
1:50:47 what I was talking about but I just it's
1:50:49 like I can't even imagine where we're
1:50:51 gonna be a year from today so so what's
1:50:53 it going to be July 1st so July 1st 2024
1:50:57 oh holy [ __ ] so a couple of things
1:51:03 check this
1:51:05 the technology is going to be [ __ ]
1:51:07 mind-blowing gpt4 will be probably fully
1:51:10 realized where it'll be multimodal where
1:51:12 it can it can understand images and
1:51:15 you'll be interacting with not only
1:51:16 words but images as inputs
1:51:19 um it will probably be there will
1:51:22 probably be a competitor to gpt4 where
1:51:24 one of these other companies is going to
1:51:26 put something out that kicks the [ __ ]
1:51:28 out of it where all of a sudden we've
1:51:31 now got even way more powerful tools and
1:51:34 July 1st next year
1:51:37 we're going to be
1:51:40 four months out
1:51:42 from the election
1:51:47 there is going to be such insane
1:51:50 misinformation
1:51:52 there's going to be credible
1:51:55 stories from the New York Times that are
1:51:57 just fully fabricated and and no one
1:52:00 discovers it for a week like I it's just
1:52:02 gonna be insane it's gonna be [ __ ]
1:52:05 insane so so so this mark my words this
1:52:09 time next year we're going to be buried
1:52:11 in a mountain of misinformation and deep
1:52:14 fakes
1:52:15 in photos and videos
1:52:17 that you can't you literally can't tell
1:52:20 reality from fiction and then we're
1:52:23 gonna have amazing tools and and we're
1:52:25 probably gonna be sort of fully in the
1:52:28 middle of employment disruption by these
1:52:30 tools
1:52:33 yeah next year this time is going to be
1:52:35 insane
1:52:36 if you think if you're thinking about
1:52:38 taking a long summer vacation next
1:52:40 summer might be the summer take three
1:52:43 months off it just go somewhere without
1:52:45 internet access and just come back when
1:52:47 like the you know everything's on fire
1:52:51 if you see Smoke on the horizon that's
1:52:55 that's the US
1:52:58 oh my god wow that's gonna be crazy
1:53:04 um prompt Engineers could be dinosaurs
1:53:06 by the start of next year yeah so what
1:53:09 Rob what robsie is talking about there
1:53:11 is
1:53:13 right now everyone's talking about
1:53:15 prompt engineering like like here's the
1:53:16 deal I feel like learning how to prompt
1:53:20 and learning
1:53:21 how to get these tools to give you what
1:53:23 you want it's actually really important
1:53:25 to understand
1:53:27 what these things actually are and what
1:53:30 they make possible but within six months
1:53:33 definitely within a year a lot of that
1:53:35 prompting is going to drift to the
1:53:37 background where like prompt perfect
1:53:38 rewrote my shitty prompt into a good
1:53:40 prompt you're gonna have tools where you
1:53:42 just kind of ask it for what you want
1:53:44 and it will go figure the prompting out
1:53:46 you're not gonna have to know all this
1:53:48 sophisticated like if you if if you've
1:53:50 played with mid-journey where if you
1:53:52 want it to be a horizontal image you do
1:53:54 dash dash AR
1:53:57 space 16 colon 9. what
1:54:01 dash dash that means I'm asking you to
1:54:04 do something AR aspect ratio I'm asking
1:54:08 you to do an aspect ratio space 16 colon
1:54:11 nine not 16-9 16 colon nine is a 16 by 9
1:54:15 aspect ratio
1:54:17 you're not gonna have to do that [ __ ]
1:54:19 anymore like like all of this bizarre
1:54:22 like prompt engineering [ __ ] is is going
1:54:25 to drift into the background now if you
1:54:27 want to work building tools that do all
1:54:29 the cool [ __ ] you're gonna need to know
1:54:30 prompt engineering so it's not a bad
1:54:32 skill to learn but uh a lot of the stuff
1:54:36 that's on the surface right now that we
1:54:38 have to learn you're not gonna have to
1:54:40 do so if if you're in a place where you
1:54:43 enjoy geeking out and getting geeky with
1:54:46 it great learn prompt engineering learn
1:54:49 Lang chain learn Google collab if that
1:54:53 kind of [ __ ] makes your head hurt and
1:54:54 you're like I just want to go crawl into
1:54:56 a ball and like suck my thumb and drink
1:54:58 some wine and have pudding then just
1:55:00 wait for six months or a year this
1:55:03 shit's gonna get better
1:55:05 do what you can if I just typing words
1:55:08 into that thing and and don't beat
1:55:10 yourself up if you're not brilliant at
1:55:11 the tech [ __ ]
1:55:13 because the tech [ __ ] is just I don't
1:55:15 know well I mean listen that's gonna
1:55:17 change as well like all these engineers
1:55:21 this shit's gonna do gonna do it for
1:55:23 them a lot of it
1:55:28 I lost my place in my little comments
1:55:30 here
1:55:33 mm-hmm
1:55:35 [Music]
1:55:37 I think I've been in here for like how
1:55:39 long have I been in here two hours when
1:55:40 did I start seven yeah I've been in here
1:55:42 for two hours I I can tell when I've
1:55:45 been in here two hours because my ass
1:55:47 gets numb
1:55:49 I need to I need to get a stand-up desk
1:55:52 at home I have one at work I just don't
1:55:53 have one here
1:55:54 [Laughter]
1:56:00 I gotta sit up straight I gotta move my
1:56:02 shoulders like everything's [ __ ] all
1:56:04 locked up
1:56:06 all right let me see
1:56:12 this dude is legit I'm in my 40s and I'm
1:56:14 tired of seeing only 20 year olds lose
1:56:18 using AI yeah I know I I'm telling you
1:56:20 man if I were gonna hire someone right
1:56:22 now like like uh
1:56:25 late 40s early 50s
1:56:28 creative director that's AI literate
1:56:31 like all day long over someone younger
1:56:34 and not because younger is bad just
1:56:37 because
1:56:39 that person's seen some [ __ ] right and
1:56:42 like older people are just gonna be able
1:56:43 to look at the output of these tools and
1:56:45 and more quickly curate them it's
1:56:49 actually one of the one of the concerns
1:56:51 I have about about sort of the coming
1:56:53 automation of knowledge and creative
1:56:56 work
1:56:57 is the jobs the the functions not the
1:57:01 jobs the functions in companies that are
1:57:03 going to get automated
1:57:05 are the jobs that are historically the
1:57:08 entry-level jobs right so if you're in
1:57:10 an agency in the art Department the
1:57:13 entry level jobs are go iterate these
1:57:17 200 versions of this Photoshop file
1:57:19 right you know go change the aspect
1:57:21 ratios go pixel push right or if you're
1:57:23 a copywriter you know go write us you
1:57:26 know 100 different versions you know
1:57:28 inspired by this uh headline whatever
1:57:33 those are the entry level jobs right and
1:57:36 and that's where they learn the craft
1:57:37 and that's where they learn the politics
1:57:38 and that's where they learn how to
1:57:40 navigate the organization if you
1:57:42 automate all that out where do young
1:57:44 people enter the workforce and I don't
1:57:46 know
1:57:48 like the only thing that I can think is
1:57:50 so if I go back to the scenario about
1:57:53 the the person I'd like to hire right
1:57:55 now is someone who's got 30 years
1:57:57 experience and is AI literate
1:57:59 maybe the entry level job is you put you
1:58:03 pair the the young person with an older
1:58:05 person so it's like a mentor mentee
1:58:08 relationship where they're both AI
1:58:10 literate but the the older person's got
1:58:13 the World experience and then the
1:58:15 younger person goes along and and learns
1:58:17 about curation and taste and and
1:58:21 refinement and you know so so so the
1:58:25 entry so the younger person is learning
1:58:27 kind of I don't know
1:58:30 higher level skills sooner they don't
1:58:32 have to do the [ __ ] soul-crushing
1:58:34 work for four years before they earn the
1:58:37 right to be able to do something
1:58:39 interesting maybe you maybe you pair the
1:58:41 young person with an older person and
1:58:44 then they're both just doing amazing
1:58:46 [ __ ] work because the tools are
1:58:48 remarkable
1:58:50 but I don't know I'm like that that the
1:58:52 entry level thing is is the one thing
1:58:54 that I can't get my head around how
1:58:56 where do young people enter the
1:58:58 workforce if all of the historical
1:59:00 entry-level work is automated I do not
1:59:03 know the answer to that
1:59:04 I've been using AI since 1998 I was a
1:59:06 satellite guy in the Air Force that's
1:59:08 the thing you know for someone like you
1:59:11 someone like me probably drives you
1:59:14 [ __ ] crazy because I've not been in
1:59:16 AI since 1998 I was in the in the world
1:59:18 wide web uh you know I I had one of the
1:59:22 early web agencies back then but again I
1:59:24 wasn't developing things
1:59:27 for the internet right
1:59:30 I was the guy that came along when Tim
1:59:32 berners-lee created the world wide web
1:59:34 and I said oh you can make a page with
1:59:36 little hyperlinks in it that go to other
1:59:38 Pages let me start an agency that just
1:59:40 does that [ __ ]
1:59:41 I wasn't doing protocols and all the
1:59:44 underlying stuff on the internet so
1:59:46 someone who's been doing actual machine
1:59:47 learning and AI it's a thing to remember
1:59:49 this AI stuff is not like a year old
1:59:51 it's like 30 or 40 or 50 years old
1:59:55 but chat GPT just like the world wide
1:59:59 web made it available to the rest of us
2:00:01 right and because a hundred million
2:00:03 people started using that in six weeks
2:00:05 it obviously resonated right so so we've
2:00:08 flipped to we've flipped to the other
2:00:10 side of an inflection point where just
2:00:13 regular people are now have access to
2:00:15 these tools and they're going to use it
2:00:17 in ways you know people like you
2:00:19 wouldn't have imagined you know since
2:00:21 you've been doing it for so long but
2:00:22 it's a thing to remember that AI is is
2:00:24 in almost everything we do it's in every
2:00:27 Google search it's every time you use
2:00:29 your phone like ai's in everything
2:00:33 but AI that we get to play with and [ __ ]
2:00:35 around with that's new
2:00:39 we meaning non-technical
2:00:43 keeps freezing that's because I have
2:00:45 Xfinity and XFinity sucks anyone here
2:00:48 work for Comcast bite my butt
2:00:51 serious guys I'm using chat GPT for some
2:00:54 crazy stuff that's awesome
2:00:56 it's really exciting I know okay I share
2:00:59 your excitement
2:01:00 I mentioned I'm in the military 25 years
2:01:02 I have PTSD because of it yeah I bet
2:01:04 that sucks
2:01:06 created a PTSD program with with chat
2:01:08 EBT it's funny there's a woman in um
2:01:11 there's a woman in the AI Salon that's
2:01:13 creating a game right now
2:01:16 for
2:01:19 um
2:01:20 like like narcis narcissistic abuse like
2:01:24 narcissistic piece PTSD
2:01:27 um so I I think that I think that these
2:01:29 these large language models are going to
2:01:32 be astounding at allowing people to work
2:01:35 through trauma and and
2:01:38 um yeah so awesome awesome use of the
2:01:42 tool
2:01:43 thanks for your service yeah thank you
2:01:45 very much for your service
2:01:48 um
2:01:49 now with the plug-in you say ask me some
2:01:52 questions to personalize the story oh
2:01:54 yeah with the with the stories plug-in
2:01:55 it's pretty slick all right I missed the
2:01:57 name of the plugin can anyone share the
2:01:59 name of the plugin which one
2:02:02 which plugin there was prompt perfect
2:02:04 which I showed there was instacart there
2:02:07 was whatever the shitty one was stories
2:02:10 is the one that makes kids books what
2:02:13 can I do to get all the statistical
2:02:14 facts in a YouTube video presentation
2:02:17 listed what can I do so
2:02:21 I don't know how to get the statistical
2:02:23 facts because if it's only showing on
2:02:26 screen it might not capture those but
2:02:29 there are plugins so if you go to if you
2:02:31 go to the plugins they're a whole [ __ ]
2:02:33 ton of plugins right now that will go
2:02:36 read the transcript of a YouTube video
2:02:38 so it's not actually analyzing what's on
2:02:41 screen that will come like that'll
2:02:43 probably be here within a year
2:02:46 um
2:02:47 but assuming what's on screen was talked
2:02:51 about that'll be in the transcript so
2:02:52 all the plugins right now you can just
2:02:54 go point it to a YouTube URL and say
2:02:57 tell me all the statistics that were
2:02:59 mentioned in this video out it'll come
2:03:02 put them in a table and compare those
2:03:05 statistics with this you know the
2:03:07 current day statistics and create a
2:03:11 third column that you know has
2:03:13 commentary on the implications of the
2:03:16 differences
2:03:18 out it'll come
2:03:19 in a table that you can copy and paste
2:03:24 um
2:03:25 sounds like someone's leaving here I
2:03:27 think I think I have to run but let me
2:03:29 let me go check let me go check
2:03:41 uh
2:03:52 hang up on my life
2:03:54 because he needs me to help carry this
2:03:56 to the car right yeah okay
2:04:03 all right daddy duty calls I gotta go
2:04:06 carry a coffee table out to a car so
2:04:09 maybe I'll come back later but I'm a
2:04:12 little I'm a little crispy my butt's
2:04:13 sore it's probably good for me to get
2:04:14 some exercise so anyway
2:04:18 um yes uh so cavuno said I scolded chat
2:04:21 GPT for being cliche it apologizes like
2:04:24 a good boy exactly tell chat GPT when
2:04:27 it's been bad it will uh it will do you
2:04:30 right
2:04:31 um I hope this has been helpful it's
2:04:33 been a lot of fun for me tonight I
2:04:34 appreciate your comments and uh as
2:04:36 always I really appreciate you hanging
2:04:38 out on a Saturday night why don't you go
2:04:40 out and have some fun now or go there
2:04:43 and have your mind blown have your Kevin
2:04:45 McAllister moment
2:04:48 play with it until you have that moment
2:04:50 and then you'll be hooked all right
2:04:51 peace out everybody