aiLL 9/22/23 - 17k Learn About Generative AI Live With Kyle Shannon

Description

Chapters

Transcript

0:10 mm-hmm
0:19 come on
0:21 there we go
0:30 welcome welcome welcome welcome to the
0:33 AI learning lab everybody
0:35 happy Friday night
0:37 another Friday can you believe it
0:41 hmm
0:42 it's another Friday night
0:53 so there's that what's happening hello
0:56 hello
0:57 my watch is too tight let me loosen this
1:02 ah
1:03 [Music]
1:07 what's going on here there we go
1:10 all right
1:13 Nancy hello there Irregulars what's
1:16 happening
1:17 what's shaking
1:19 welcome
1:20 real chat bots of the future
1:25 making money with chat gbt what else we
1:27 got
1:29 we got let me find my Trello board
1:33 we got all the good sayings here at the
1:36 at the AI learning lab a uh learning lab
1:42 what do I what did I Trello up
1:45 weird weird weird weird
1:48 [Music]
1:52 escapee ah
1:56 let's see
1:58 this chat add welcome
2:00 mansplaining is a service I'm not weird
2:02 I'm irregular I'm AI native the
2:06 beginning is here we have an image for
2:07 that one
2:12 the beginning is here
2:18 I recruited a non-technus to start
2:21 learning AI awesome that's super cool
2:24 that's super cool how did it go did they
2:26 resist
2:30 it's like a cult Training Center have
2:32 you tried AI listen listen can I talk to
2:35 you about church every day what did you
2:37 say can I talk to you about tattoo
2:40 did you say Church
2:42 they're listening
2:44 Siri's very jealous
2:47 series very jealous Bing has gotten a
2:50 little aggressive
2:52 but could I talk to you about Chachi be
2:54 today
2:58 she was all over it good
3:00 good I you know listen I I think that I
3:03 think that word is getting out that this
3:05 [ __ ] isn't going away
3:07 and
3:09 it's time it's time to start you know
3:12 this is this is not dissimilar to the
3:14 early days of the web where this little
3:16 tool came along and first it was the
3:18 dweebs and the Geeks and then Along Came
3:21 the lawyers and the accountants and the
3:23 creative people and they all started
3:25 using this worldwide Webby thing
3:28 and then at some point it became
3:30 inevitable and you know with the web it
3:32 took two or three years with this stuff
3:34 it's we're what are we 10 months in now
3:37 something like that Stevo what's
3:40 happening
3:41 ah
3:42 I had a little bit of coffee earlier it
3:44 was not enough had a long day today I
3:47 got up at
3:48 I thought I was gonna have a seven
3:50 o'clock meeting I didn't an eight
3:51 o'clock meeting a nine o'clock meeting
3:53 at 10 o'clock meeting at 10 30 meeting
3:55 an 11 30 meeting
3:57 and I think that was it until two
4:01 so I'm Crispy and they were big meetings
4:04 these were big beefy
4:05 you know CEO type meetings
4:11 all right welcome everyone my name is
4:13 Kyle Shannon this is the AI learning lab
4:14 if you do me a favor go ahead and share
4:17 the hey Luna stick what's happening if
4:19 you share the uh Stevo I'm late because
4:22 I was updating my phone again has anyone
4:25 noticed like the past three days I keep
4:27 getting logged out of like Google Apps
4:29 and and just everything I seem to be
4:32 logging into they're like log in again
4:34 and like just use our convenient 17
4:36 Factor authentication scheme
4:39 you're like click here then we'll text
4:42 you on your phone then we'll ask you if
4:44 it was really you and when you say yes
4:45 we'll bring you back here and then we'll
4:46 jump you to another browser where you're
4:48 gonna have to do a capture then you're
4:49 gonna have to come back here scan your
4:51 iris
4:52 and then we decide you can't get into
4:54 your account because we don't recognize
4:56 the machine really is this uh that's how
4:58 we're going anyway
5:00 that's not what this channel is about
5:01 although this channel is about me
5:04 bitching about stuff it absolutely is
5:06 don't get me that don't let me deny that
5:09 because I that would be inaccurate
5:12 this is a whining Channel we're dug into
5:14 any wine or we've got diverticulitis no
5:19 um why are they aggravating us exactly
5:22 Nancy Nancy understands Nancy's an
5:25 irregular Nancy understands the pain
5:30 just
5:31 you know we didn't used to have
5:33 passwords and then when we had passwords
5:35 you had one password for life it was
5:37 like you had one password in a decade
5:39 and you just you put it on all your
5:41 sites and then the computer guys were
5:43 like not very secure and I'm like oh
5:46 what I suppose I can't have it on a
5:48 Post-It note on my monitor too uh
5:50 technically no
5:52 and now we're at 17 Factor
5:53 authentication thanks Geeks
5:56 I'd rather have all of my data stolen
5:58 anyway back to what this channel is
6:00 about
6:03 yay you did it yay uh this is the AI
6:08 learning lab and my name is Kyle Shannon
6:11 I'm an entrepreneur believe it or not
6:13 I'm a CEO of an 11.5 year old company
6:15 called storyvine which is an automated
6:17 video storytelling platform
6:20 that believe it or not does not use AI
6:22 to automatically edit or as we call it
6:24 auto magically edit videos
6:27 um we we use a more traditional assembly
6:29 method uh but it's still automated
6:32 and uh I've spent about the last two
6:35 years deep diving into this generative
6:36 AI stuff and the image stuff and
6:39 um I was an entrepreneur in the early
6:41 days of the World Wide Web I co-founded
6:43 this this company called agency.com
6:45 which was one of the very first digital
6:47 agencies and holy [ __ ] are we right back
6:51 at the beginning of something
6:53 and these little tools here so if you
6:56 have not played with chat GPT if you are
6:58 a chat GPT newbie and you haven't done
7:00 it and you're like I heard about it but
7:02 I don't I'm I'm not too technical Kyle I
7:05 don't I don't know how to tell you this
7:07 but uh I don't know much about the
7:09 computers
7:10 if that's you fine doesn't matter you're
7:13 cool just go there right now
7:16 you know how to go to a URL right what's
7:19 a URL it's a website oh a website yeah
7:21 just go there
7:24 and that's chatgpt that's the official
7:27 chat GPT website start playing just go
7:29 there right now and start playing and if
7:30 you're like but Kyle I don't know what
7:31 to say go there proms.chat
7:35 problems.chat will teach you how to talk
7:36 to that thing
7:38 oh and Sam I don't know if Sam's in here
7:40 yet hey share bear retro Punk the
7:43 whiners SNL throwback damn that takes me
7:46 back we're dug into any wine or we've
7:48 got diverticulitis no
7:50 [Laughter]
7:56 um from stat chat will teach you how
7:58 these large language models work and how
8:00 to talk to them if Sam's here I told Sam
8:03 the other place you might want to start
8:04 so one place to start is chat GPT the
8:06 other place is Claude dot dot a i
8:09 you could go to bard.google.com but give
8:12 that one a week that one's going to get
8:13 better soon
8:15 claude.ai is also really good either of
8:18 those top two are a place to start go
8:21 there now go do that now and then when
8:23 we get rolling here uh I'll answer some
8:26 questions if you have questions about AI
8:28 pop them in the comments below all those
8:30 hearts that you're seeing those are the
8:31 Irregulars they're people that show up
8:33 here every night they just harass me
8:34 it's it's just a never-ending
8:37 it's it's like reverse trolling
8:40 they they compliment me they hit the
8:43 heart button and I know there's
8:45 something more to it I know they've got
8:47 an agenda so I just I don't trust it I
8:50 don't trust it I know there's an agenda
8:52 not everything's a conspirac conspiracy
8:55 but just because you're paranoid doesn't
8:56 mean they're not out to get you you know
8:57 what I'm saying
8:58 all right okay so let's get going here
9:02 we're an odd lot we are not plot we are
9:05 the Irregulars we I am irregular I am a
9:09 regular I am a regular
9:13 I am a turtle
9:17 um
9:18 all right
9:19 let's see
9:22 let's go ahead and system questions
9:23 let's go do this oh and then I'll sort
9:26 of dig through I'll I got a bunch of
9:28 [ __ ] I can show you here
9:30 websites and things like that well bites
9:32 for some more people to get rolling
9:34 evangelized ideogram to non-believers in
9:37 the house
9:38 mcallistered my 21 year old son now that
9:43 see okay so what Ted Ted looks at life
9:46 is talking about there is is if you
9:48 haven't played with these tools yet you
9:50 owe it to yourself to play with them
9:51 long enough to have a Kevin McAllister
9:53 moment which is
9:55 the Home Alone moment where at some
9:57 point it's gonna blow your [ __ ] mind
9:59 so I Ted's been in here a lot for for
10:02 many weeks now and so his son knows he's
10:05 been you know he's been on The Tick Tock
10:07 and he knows he's been talking about Ai
10:09 and the sun I got 23 year old twin boys
10:12 so I know how how much they think they
10:15 know
10:16 and how big their opinions are so the
10:19 fact that you Kevin mcallistered a 21
10:21 year old is quite impressive
10:25 that's awesome that's great
10:28 yeah idea Graham is quite something it
10:31 really is so awesome
10:34 all right let's see y'all check my
10:37 article on Discord and legal ethics
10:39 Channel about bart oh did you write that
10:41 Luna stick I thought you just were
10:43 sharing that that's super cool
10:48 Apple just updated
10:49 17.0.1 I hope that dot one is making
10:52 their voice synthesis better
10:55 there's comic relief in that yeah what
10:58 are they aggravating us
11:07 all right let's see
11:10 there's a brand for that
11:13 another handsome shirt well thank you
11:15 it's a lovely checkered pattern it's
11:17 classic yet modern
11:19 simple yet elegant
11:22 thanks
11:23 I went to the haberdasher I I also
11:26 noticed I have some food right there so
11:28 it's lovely and all I can see
11:32 is is that you know something just like
11:36 [ __ ]
11:38 foreign
11:45 like the insides of the of the uh one of
11:49 those cuffs are a different fabric
11:52 ah swell right
11:55 all right
11:57 AI learning lab and Fashion Center
12:01 hair's not bad tonight all right how do
12:05 you get AI on computer I'm new to this
12:07 okay cool so welcome
12:11 um
12:12 I would encourage you change your name
12:14 from user whatever that number is to
12:16 like something clever the one of the
12:18 things we do in here is we judge names
12:20 and so yours right now
12:22 um 0.5 out of ten but change it to
12:25 something clever like a like a nickname
12:27 you want people to call you or something
12:28 even more clever where they have to
12:30 figure it out
12:32 um anyway so here's what you do
12:35 you don't need to put AI on your
12:37 computer it's these are just websites
12:39 that you can just go play on so
12:41 chat.openai.com or Claude dot AI or
12:46 bard.google.com just type it into your
12:48 browser window and here let's go show
12:51 you let's go do it right now
12:54 I'll go show you what I call parlor
12:56 tricks there's some parlor tricks you
12:58 can do
12:59 right brother right brother
13:02 we'll share the chat JPay tag all right
13:05 so
13:06 when you get to chat GPT you you won't
13:09 see all these things down here but down
13:10 the left hand side this is your prompt
13:12 history so these are all your different
13:13 conversations with it and then over here
13:17 to the right this is it
13:20 and you're like but Kyle that doesn't
13:21 look like very much whatever do I do
13:24 here exactly
13:25 so
13:27 um when you're when you first join here
13:29 and you're not paying for the advanced
13:31 version you just have access to one
13:33 model it's GPT 3.5 it's actually really
13:35 good it's very fast
13:37 these little things here are just
13:38 suggestions so this says like suggest
13:40 some code names create a workout plan
13:42 make a Content strategy design a date
13:44 database schema those are just
13:46 suggestions you can
13:48 um I wonder if you reload the
13:50 the site if it changes them I never
13:53 really paid attention yeah it does
13:54 explain my popcorn pops so you could
13:57 just start clicking on those another
13:59 thing you could do is
14:01 um
14:02 you know have it write you a country
14:04 song this is this is my go-to
14:08 write a country song about a hedgehog
14:11 lame named Larry
14:20 um
14:29 and
14:31 how he passed too soon
14:35 but his younger brother Lenny
14:42 still enjoys thinking about him
14:45 here we go
14:47 kind of an inside joke from last night
14:49 from Dolly three
14:50 and then it'll just write you a song
14:53 so if you haven't done this before just
14:55 go this is chat.openingi.com
14:59 let's see if it did any good it's
15:00 generally pretty good we gather around
15:02 folks I got a tale to share about a
15:04 little hedgehog named Larry with spikes
15:07 and hair
15:09 he rolled through life with a twinkle in
15:11 his eye but one fateful night he had to
15:13 Say Goodbye Oh Larry you left this all
15:16 too soon underneath that Silver Moon
15:18 your memory lives on so
15:21 um so you can do that
15:22 um it can it can do useful stuff as well
15:24 you could say
15:26 um let's see
15:28 um
15:30 write me a professional business email
15:32 trying to get a job selling hedgehogs at
15:36 the local pet store
15:39 boom
15:41 there's your letter
15:43 I hope this message finds you well
15:45 oh this is good look at this
15:50 um I'm writing to you to express my
15:52 strong interest in the sales position
15:53 currently available at local pet store
15:56 as a lifelong animal lover and pet owner
15:59 have developed a deep understanding and
16:00 appreciation care for for well-being of
16:02 animals including hedgehogs this is good
16:05 it's not just leading with I got a
16:07 hedgehog Obsession it's saying including
16:10 hedgehogs so we're kind of planting the
16:11 seed that we might be a little obsessed
16:13 it's good it's salad I'm excited about
16:15 the opportunity to join the team at
16:17 local pet store and share my knowledge
16:19 and enthusiasm with customers looking to
16:21 provide the best care for their pets
16:24 some key qualifications customer service
16:26 Excellence passion for hedgehogs
16:32 uh you can do that you could also say
16:36 um
16:37 write me a python application that's a
16:40 trivia game about hedgehogs
16:44 mm-hmm
16:45 sure and then it writes you code
16:48 so the the thing about chat GPT if
16:53 like
16:54 as I watched this screen when it writes
16:57 code in particular but even even when it
17:00 just writes English or a business letter
17:03 um
17:03 like I've been deeply obsessed with this
17:06 now since chat EVT came out and it still
17:08 blows me away every time I see it
17:11 um that's real code it may or may not
17:14 work but if it doesn't work you can put
17:15 the error messages in it'll tell you how
17:17 to debug it so you can write code with
17:20 it you can write words with it you can
17:21 write copy with it you can create meal
17:24 plans with it so just start playing the
17:27 most important thing to do right now is
17:29 don't be intimidated by it there's
17:31 nothing to install you just go to these
17:33 things and you start playing
17:36 um and and that's that's really the
17:38 beginning and and basically you know my
17:41 my
17:43 My Hope
17:45 and my advice is that you play with it
17:48 long enough
17:49 that it kind of breaks your brain a
17:51 little bit you have the Kevin McAllister
17:53 moment you have this moment
17:54 where you like wait it can do that
17:59 and then
18:01 think about something that you're really
18:03 good at or that you know really deeply
18:05 and ask it about that and you'll
18:07 probably find that it's good at that too
18:11 um I do find that if I don't do some
18:14 serious prompt tweaking
18:17 the responses are often very can
18:19 sounding yeah exactly that's very right
18:21 well good they still require some work
18:23 so gray-haired Gamers so here's the deal
18:25 with this stuff
18:27 as remarkable as the tools are so first
18:29 thing is
18:31 gpt4 is way better than GPT 3.5 and in
18:35 fact it's way better than essentially
18:37 everything out there until further
18:39 notice apparently Gemini that's coming
18:41 from Google is better than gpt4 but we
18:46 haven't
18:47 we haven't heard anyone outside of
18:49 Google say that yet so that could just
18:50 be marketing [ __ ]
18:53 um but gpt4 is really really good that
18:56 said all of these large language models
18:58 the way they work
19:00 they're almost the opposite of Google
19:02 where with Google what what Google's
19:06 trained us to do over the past 20 odd
19:08 years is put in as few words as possible
19:11 just put in the minimum minimum viable
19:14 prompt right
19:15 um you know I want a car wash that's
19:18 open on Saturday you know like hope it
19:20 doesn't get confused right
19:23 these large language models chat GPT and
19:26 Bard and Claude all these things
19:29 are
19:32 the way they work that they require a
19:35 lot of context so and here's what that
19:37 means if if you're working on the Great
19:41 American novel one of the things you
19:43 could do is say you know you're you're a
19:46 master
19:47 um literary coach and you specialize in
19:51 helping first-time authors you know make
19:53 the best novel that they can and then
19:56 I'm working on a novel and you know it's
19:59 about you know this basic topic and it's
20:02 a mystery and it's a this and it's a
20:04 that you know can you help me structure
20:06 my story and it will say sure here
20:10 here's a bunch of ways you can structure
20:11 your story and then you can say well
20:13 what are some other ways I could
20:14 structure my story and they would
20:16 they'll throw that out there's six other
20:17 ways you can structure your story and
20:19 then you can say oh I like number two
20:21 and number four could you combine those
20:24 two story Frameworks and give me a new
20:26 story framework and it'll do that and
20:28 you're like oh that's perfect
20:31 um you know here are are some of the
20:33 main characters of my story and it'll
20:35 support you and tell you how awesome you
20:37 are
20:38 the the more context that you wrap
20:41 around that conversation the better it
20:43 gets and so by the time you actually ask
20:45 it to do something
20:47 it it's kind of
20:51 I was going to say it's in the mindset
20:53 it's not actually thinking
20:55 um but it it it is in a it is in a more
21:00 narrowed focus in terms of the language
21:02 world that it's drawing from and so
21:05 it'll get better as you give it more
21:06 context so don't feel with chat GPT or
21:10 with any of these tools like you have to
21:11 get it perfect in a single prompt
21:14 in fact my experience has been just the
21:17 opposite of that that it's that it's
21:19 it's more like a conversation with a
21:21 really dim-witted intern who just
21:24 happens to know everything right have
21:26 you ever had an intern that they're just
21:28 like clueless about the world but
21:30 they're really well educated that's chat
21:31 GPT
21:33 it's it's a simpleton that that knows
21:36 [ __ ] everything
21:38 and so and so you have to coach it and
21:41 guide it and get it there and you have
21:42 to do that for every conversation you're
21:44 kind of starting this over it does not
21:46 maintain memory across chats each chat
21:49 is its own little Universe of language
21:52 the other thing you said there's another
21:54 thing that that I think is an
21:56 expectation thing and you really only
21:57 get this when you start playing with
21:59 these things regularly
22:01 the expectation that you'll put in a
22:03 single prompt and it will spit out
22:04 something that's a hundred percent good
22:06 to go is essentially zero it's
22:09 essentially zero
22:12 my experience has been
22:15 when I was weak at prompting it could
22:18 quickly get me to like 60 to 75 percent
22:21 as I got better at prompting it could
22:23 get me to like 75 to 85 percent and now
22:26 if I really put time into the context
22:29 setting I can get to like 85 90 the rest
22:33 of it's on me and and I I was I was
22:36 working on an article for the past two
22:38 days actually where I had a really cool
22:40 idea
22:42 um I went into gpt4 I had to outline it
22:46 for me I did two iterations in chat gpt4
22:49 then I went over to Claude and I had
22:51 Claude try to do it because claude's got
22:53 a longer memory and
22:56 I ended up getting so lost in it that
22:58 now I've got I've kind of got exploded
23:02 piece parts of this idea like it kind of
23:05 [ __ ] me up because I didn't just stop
23:08 when it was like decent and actually
23:09 just write it on my own I tried to have
23:12 the machine do it better and better and
23:14 keep refining it and I kind of [ __ ] it
23:16 up so so sometimes you mess it up
23:19 um
23:20 but the important thing to do right now
23:22 these things are evolving very very
23:23 quickly the important thing to do right
23:25 now is
23:27 um
23:28 just learn them get curious about them
23:30 kind of open your mind to what's
23:32 possible because they're going to get
23:33 better and better and better let's see
23:36 um thanks to you fast forward to thanks
23:38 to you fast forward two months I'm using
23:41 regularly that's great you're utilizing
23:43 regularly that's awesome thank you for
23:45 sharing that
23:53 thank you
23:55 Lauren I had a Kevin McAllister moment
23:57 using Trace journey and Discord to
23:59 create vectors oh what's Trace journey
24:03 is that is that an AI
24:06 that's cool look at that that sounds
24:08 like an AI
24:10 because because it sounds like someone
24:12 took mid Journeys branding and they put
24:14 Trace in front of it and said hey all
24:15 you illustrator people
24:17 Trace Journey Trace Journey
24:21 Trace Journey
24:24 turn mid-journey images into vectors
24:26 with just one click the bot replies to
24:28 every [ __ ] mid-journey upscale with an
24:31 options menu holy [ __ ] we need to
24:34 install this into
24:36 um
24:37 the AI salon
24:41 whoa is this is it good Lauren this is
24:46 this is Kevin McAllister kind of
24:48 territory how it works I don't need to
24:50 know how it works
24:53 um
24:55 join Trace Journey
24:58 welcome to trace Journey
25:00 um watch how it's done read the rules
25:02 check the cool new features
25:05 start here
25:09 all right I can't I don't want to sort
25:12 of figure this thing out on the fly but
25:14 holy crap that's cool
25:18 this is so good
25:21 all right Lauren see this is you see see
25:24 listen look look listen listen people
25:27 listen listen listen
25:28 zip it zip it zip it zip it
25:32 zip it look
25:35 um
25:37 see this see what that says AI learning
25:40 lab we're learning
25:43 Lauren just shared a thing called Trace
25:45 Journey it looks flipping amazing I
25:48 can't believe I haven't heard of that I
25:49 wonder how old it is
25:52 um
25:53 so so here's the deal when you generate
25:55 images in mid-journey or in stable
25:58 diffusion for the most part they're very
26:00 low resolution there's tools like topaz
26:03 labs gigapixel and photo AI that will
26:06 upscale it for you or stable diffusion
26:08 upscalers
26:10 um but they're still bitmaps
26:14 a vector file you can scale infinitely
26:17 and it will look amazing and so Vector
26:20 Journey Vector Journey yeah
26:23 um looks like it takes those images and
26:26 makes them into
26:27 vectors that's pretty sadding that's
26:30 pretty exciting I'm working on a project
26:32 right now that that's actually really
26:34 important for so so thank you Lauren
26:39 um okay
26:41 um
26:43 all right we'll go we'll go back to that
26:44 I don't want to install that live but
26:46 that looks super cool
26:49 as an irregular I have an extreme I have
26:52 extreme enthusiasm at or Zeal and
26:55 alacrity towards Turtles
26:58 now why might that be Mr Winston
27:02 Minister Winston Hardy names Winston
27:05 Winston Hardy Winston Hardy III and I
27:09 like turtles
27:12 okay look
27:14 watch
27:16 for it watch for it
27:19 this is for you Winston Hardy
27:22 I hope your name's Winston Hardy III it
27:24 would be so good Winston Winston Hardy
27:28 Winston Hardy III pleasure to make your
27:31 acquaintance
27:32 um
27:33 why am I not going to YouTube do I have
27:35 bad internet what's going on oh you know
27:38 what this is
27:40 I know what this is so
27:44 um Microsoft Edge is now crashing on me
27:47 at least once a day I switched my
27:50 default browser to Microsoft Edge back
27:52 in January because I got so excited
27:54 about Chad JP Tay and I haven't switched
27:57 it back I I think I might switch it back
27:59 it's just it's got too many memory leaks
28:03 relaunching relaunching it's gonna be so
28:06 good
28:08 um
28:10 you know what we're gonna do here we've
28:11 got 71 people in here let me see am I
28:14 late how late am I hey Governor we just
28:16 started really
28:18 um did you see the email that Microsoft
28:19 text-to-speech and examples I just saw
28:22 that right before or not right before I
28:24 got on but right before I went out
28:25 earlier
28:27 um
28:27 they look decent I didn't have a chance
28:29 to dig in it creates SVG files okay cool
28:32 what's it called again it's called
28:34 Vector Journey
28:36 um
28:39 actually you know what I'm gonna go do
28:42 um
28:42 [Music]
28:46 I'll be right there hold please
28:49 let's see at Ally
28:54 um
28:57 whoops
29:04 um
29:07 and you take a look at Vector
29:12 Journey
29:15 and see if we can install the bot
29:21 um
29:22 Vector images
29:26 from
29:28 mid Journey ones exclamation point all
29:32 right
29:33 I just delegated that to get that
29:36 installed over on the AI salon so let me
29:39 show you a couple of things here
29:42 eaten
29:44 um so here's what I do if if you're new
29:47 to this AI stuff I'm going to show you I
29:49 don't know three or four slides in a row
29:51 or five I don't know that have a bunch
29:54 of websites on them if you're in a
29:55 regular you can just start tapping the
29:56 screen or you know go go make a coffee
29:59 or something you've seen this before but
30:01 feel free to screenshot this these are
30:04 I'm not affiliated with any of these
30:06 sites in any way these are the these are
30:08 the sites that I think are worth you
30:10 paying attention to right now if you're
30:13 brand brand new to this stuff you don't
30:14 need anything more than what's on screen
30:16 right here chat GPT maybe Claude
30:19 or Claude maybe chat GPT
30:22 there is so much depth in all of these
30:25 tools that you can you could spend the
30:27 next two or three years exploring them
30:29 so so don't feel like you need to do all
30:32 this stuff if you've been digging around
30:33 for this stuff for a while
30:36 these are the these are the sites that I
30:38 think are interesting and why so so chat
30:40 GPT
30:42 interesting for a number of reasons if
30:45 you pay the 20 bucks a month for the
30:47 Plus subscription you get gpt4 you get
30:50 plug-ins which some of them are
30:52 interesting you get what used to be
30:55 called code interpreter it's now called
30:57 Advanced Data analysis which is what it
30:59 is
31:00 um and you get access to the new [ __ ]
31:02 what's the new [ __ ] they have a new
31:05 thing coming out called Dolly three
31:08 which looks to be a mid-journey killer
31:10 so mid journey is one of those image
31:12 generation things you're now going to be
31:13 able to generate images natively within
31:15 chat gpt4 so
31:18 that's a big deal and it's they're The
31:21 Originals they're the one that started
31:23 all this madness
31:25 um Claude is a bunch of ex-open AI
31:27 people they started a company called
31:29 anthropic Claude
31:32 you can put really long prompts in like
31:34 you could put an entire 50 000 word
31:37 novel into the prompt and ask it to
31:40 write an epilogue for the book and it'll
31:42 just do it
31:44 um it's it's really good it's really
31:46 interesting
31:47 um Bard is Google's entry into this
31:50 space Bard's got some
31:52 amazing additions in the past few days
31:56 where as long as you don't have a Google
31:59 workspace account so basically if your
32:03 work pays for Google you're screwed
32:06 if you have an individual Google account
32:10 bard.google.com now has extensions where
32:13 you can search across like Google drive
32:16 or all of your Gmail
32:18 it's gonna get really interesting really
32:21 fast so these three are kind of the big
32:22 three right now that I think are
32:23 interesting prompts.chat will teach you
32:25 how to talk to them the Kids Table of
32:28 these large language models these are
32:30 still worth playing with
32:32 um you'll notice that Bing is grayed out
32:35 that's because
32:36 Bing is not only at the kids table but
32:38 they had to turn their their chair away
32:40 from everyone else at the table because
32:42 they've been a jerk lately Bing's been
32:45 jerky and
32:48 um but Bing is connected to the internet
32:52 it's oh cool
32:54 Bing is connected to the internet you
32:56 can make images there
32:59 you can
33:00 um
33:03 analyze images there
33:05 it's gpt4 it's free and they're going to
33:09 be incorporating Dolly three into that
33:12 as well that's Microsoft's version of
33:15 chat GPT so Dolly three inside of Bing
33:18 is going to be really interesting so
33:20 that's worth playing with
33:22 perplexity.ai is good for doing research
33:25 it does citations really well and then
33:28 labs.perplexity.ai that's metaslama 2
33:30 models those are good
33:32 Pi dot AI is a company called inflection
33:35 they just raised 1.3 billion dollars 1.3
33:39 billion dollars
33:43 for a chat bot
33:45 they're spending about a billion of that
33:48 on 22 000 h100 gpus if you don't know
33:52 what that means they're buying a lot of
33:54 machines
33:57 um and then you got poda oh the other
33:59 thing about Pi is our nickname here in
34:02 the lab for pi is chatty GPT chatty GPT
34:05 is really good if you if you download
34:08 the PI app pi and then search for pi
34:12 personal assistant
34:13 you can put it into conversation mode
34:15 and just start talking to it so you're
34:17 having a conversation with one of these
34:19 GPT models connected to the internet
34:21 fascinating useful interesting Wild
34:27 um Vector Journey or Trace Journey oh it
34:30 might be Trace Journey sorry if I if I
34:33 lied hang on I think it's Trace Journey
34:35 Trace Journey
34:37 Trace Journey sorry for all the people
34:39 that went Stevo oops he got that wrong
34:42 Stevo seriously that's the first time
34:44 I've ever gotten anything wrong right
34:46 back me up on this I never get anything
34:48 wrong right Stevo definitely have never
34:51 made a mistake it's well except always
34:54 listen
34:56 it's called Trace Journey I don't know
34:59 which one of you Irregulars told me it
35:02 was Vector Journey that's wrong it's
35:04 just not right
35:05 so I'll correct it I'll correct your
35:08 mistake I'll correct your mistake it's
35:09 Trace Journey Trace Journey
35:12 and let's see did I I called it Vector
35:14 journey in in Discord as well I'm a
35:17 loser baby so why don't you sue me
35:21 let's see uh Trace Journey
35:24 Trace Journey all right there we go
35:30 Vector images from a mid Journey okay
35:33 perfect thank you for that
35:35 sorry about that oh that's so
35:38 embarrassing
35:40 um oh wait
35:41 [Laughter]
35:43 okay we're here uh poe.com lets you play
35:47 with a bunch of these different models
35:49 uh open AI Claude some other ones and
35:52 it's got a bot maker you can make your
35:54 own bot with poe.com
35:56 so there's seven large language models
35:58 right now that I think are worth playing
36:00 with I think Bard is gonna get how many
36:03 times have you attended the AI learning
36:06 lab oh that was a cool one
36:10 Edge is the worst Edge is actually
36:12 really good except for the fact that it
36:14 sucks
36:15 just like Bing that's why Bing's at the
36:18 freaking kids table with this her chair
36:21 her name's Sydney with her chair turned
36:24 around with her back to the table okay
36:26 so those are the word ones these three
36:28 are the top three these three are the
36:31 kids table but still worth paying
36:32 attention to now the pictures pitches
36:35 oh God I gotta add Trace Journey To This
36:37 I Gotta play with Trace Journey first
36:40 but it looks [ __ ] amazing okay
36:42 um
36:44 so these things
36:46 if if you have not spent time making
36:49 pitches with these here large with these
36:52 here uh generative AI things they're
36:54 quite remarkable
36:55 so mid journey is the best one for now
36:58 it's not gonna be the best for long but
37:00 but it's the best for now this list is
37:02 Gonna Change pretty dramatically here
37:04 relatively shortly
37:06 I think I think Dolly three is going to
37:08 be at the top of it which considering
37:10 how bad Dolly 2 is right now
37:12 um is gonna it is like a minor miracle
37:17 um
37:18 ideogram is the one that
37:22 um lets you do pitches
37:25 I mean words so like AI learning lab
37:27 Tick Tock sent me to the bottom I'm not
37:30 a robot goodbye
37:32 um AI learning lab all these things were
37:34 made with ideogram
37:37 ideogram dot a I so it's just like you
37:41 know stable diffusion or any one of
37:43 those you know kind of image generation
37:45 things but it's super simple and it can
37:48 do text
37:49 Dolly three that's coming out can also
37:51 do text and it looks like it does it
37:53 better than this but this was not
37:55 possible like a month ago and it is now
37:59 so there you have that
38:02 um
38:02 what else do we have here
38:04 um
38:05 leonardo.ai is a web-based uh interface
38:09 for stable diffusion where you don't
38:11 have to install anything it's really
38:13 really good
38:14 hey uh let me skip hey Jen for a second
38:17 Adobe Firefly and Photoshop generative
38:19 fill came out of beta about a week and a
38:22 half ago and are now good good enough to
38:26 use for government work like you can you
38:28 can make commercial work with them now
38:30 you couldn't when they were in beta you
38:32 can now
38:34 prompt book will teach you how to talk
38:36 to these things and then I want to talk
38:38 about hey Jen so hey Jen is a tool
38:41 that's kind of like D Dash ID or
38:43 Synthesia where you can do text to voice
38:46 and then you can animate those voices
38:48 with a picture with like a person
38:50 talking you can make synthetic people
38:54 but one of their Engineers kind of went
38:56 rogue and did a little side project
38:58 called video Translate
39:01 so if you go to labs.hagen.com
39:05 click on video translate and upload a
39:08 video of yourself speaking English
39:12 and then
39:14 pick a language and wait for the
39:17 obligatory 15 minutes for it to process
39:19 and you'll get back something
39:22 that's kind of like this now what I did
39:24 here was I took the English original
39:26 video and the Italian that it translated
39:29 and I and I sort of intercut them back
39:31 to back so I have this sort of ping pong
39:32 between me speaking English and me
39:34 speaking Italian
39:36 so we've heard from a number of clients
39:38 that modular content is important to
39:40 them getting more efficient with how we
39:42 create content and how we leverage
39:43 content is
39:51 behind me you see these little green
39:53 boxes that's how we think about stories
39:56 at storyline we break stories down into
39:58 little atoms right little modules
40:01 foreign
40:08 so yeah so if you noticed it's not only
40:13 synthesizing my voice it's matching my
40:16 tone and Cadence
40:18 it's syncing the lips up
40:21 and it's reanimating my whole face uh
40:25 there's a little glitch like actually if
40:27 you notice right there wait uh
40:31 notice that my hand kind of disappears
40:33 underneath my face that's because they
40:36 take the lower half of your face and and
40:39 they're synthesizing it and they layer
40:41 it on top of the video so if your hand
40:43 passes here you pass under your mouth
40:45 which is kind of funny
40:47 um they'll fix that at some point but
40:48 this is a remarkable tool
40:51 um I don't think they know what they
40:52 have on their hands here this will be
40:54 the first of many but that's heygen.com
40:58 um so that's
41:00 that one let's see how many times have
41:03 you attended there you go
41:05 there's a little Poll for y'all go take
41:08 that poll go do it wait oh I I would
41:11 have done it but
41:12 it went away
41:15 um
41:16 okay so um futurepedia.io the thing at
41:20 the bottom that's just a big old
41:21 directory of AI sites
41:24 um here's the thing
41:27 I've been invited by another Creator
41:29 who's a creative coach and a voice over
41:31 coach and she's got a whole community of
41:33 folks over there to do a workshop I've
41:35 also been asked to do a workshop for the
41:37 AI learning lab I have not put that
41:39 together yet but I I am putting together
41:42 I have put together a workshop called
41:44 the AI launch pad so this is
41:47 um an AI workshop for creatives so it's
41:50 a three-part Workshop over three
41:53 different weeks you know one day a week
41:55 for three weeks starting October 4th so
41:58 if you're interested in that
41:59 is today Friday yeah today's the last
42:02 day she's got an early bird special
42:03 there that ends today so if that's
42:05 something you're interested in that's
42:07 that then the other thing is last
42:10 December I started a group called the AI
42:12 salon and when I talk about the Discord
42:14 and and people the Irregulars doing
42:16 stuff in the Discord
42:18 um
42:19 it's Trace Journey not Vector Journey
42:22 J jrc is schooling me in the in the
42:25 Discord right now you said it wrong it's
42:29 not Vector Journey that doesn't even
42:30 make sense
42:32 um
42:34 uh so the AI Salon if you go there this
42:38 is a it's a there's no charge for it
42:39 it's we have meetups every other Tuesday
42:44 so this coming Tuesday is one of our
42:46 salons and then we have a Discord
42:47 Channel and all of the links for that
42:50 this is the link tree for that story of
42:52 mine's my company and that's my that's
42:53 my handle the other thing here is
42:56 um AI learning lab Dash TT that's the
42:58 YouTube channel so these lives are
43:00 archived there and we're using Claude to
43:04 write the YouTube descriptions and the
43:06 chapterization that took me about 50
43:08 hours of prompting to get that right
43:12 um and then one final thing and then
43:13 we'll get back to answering questions
43:15 because I know you're all like but Kyle
43:17 I asked a bunch of questions holy crap
43:20 there's there are 5 000 people in this
43:23 live right now wait wait hold let me
43:28 time out time out time out hello my name
43:32 is Kyle Shannon this is the AI learning
43:34 lab normally we have between I don't
43:36 know 35 and 70 people here now there's
43:39 about 6 000 people in here I don't know
43:41 what [ __ ] algorithm we just tripped
43:42 on but welcome my name is Kyle Shannon
43:45 this is the AI learning lab we talk
43:47 about AI stuff in here we have fun I
43:50 like acting like an idiot I do
43:51 occasionally cuss so if you don't like
43:52 that I would jettison now
43:55 um
43:56 one of the questions you might be asking
43:58 and I get asked this a lot sir what are
44:01 your qualifications I have none
44:04 um I'm an entrepreneur I'm a co-founder
44:07 and CEO of a company called storyvine
44:09 which is an automated video storytelling
44:12 platform
44:13 the I've been an entrepreneur for years
44:15 I've started I don't know 12 or 15
44:16 companies in the mid 90s I was a
44:19 co-founder of a company called
44:20 agency.com which was one of the very
44:22 first digital agencies and that was
44:25 insane
44:26 and I was lucky enough to be part of the
44:30 very beginning of something
44:32 transformative like the World Wide Web
44:34 at you know decently the ground level
44:36 and and kind of rode that rocket ship up
44:40 we're at exactly the same spot right now
44:43 almost eight thousand I don't there's
44:45 there's eight point seven thousand
44:47 people here I have no idea what's going
44:48 on Trump I have no idea we obviously so
44:51 we we hit it we hit the algorithm
44:55 um
44:57 2023 is the new 1995. is I've got a
45:00 video on that uh um it's actually one of
45:03 my pinned videos I've got a two-part
45:04 video pinned to the top of the channel
45:06 here
45:08 um where I talk about how reminiscent
45:11 what we're going through with generative
45:12 ai ai right now is to the early days of
45:15 the World Wide Web so this channel
45:18 is about that inflection point
45:21 that started on November 30th 2022 so 10
45:24 months ago which was the launch of chat
45:27 GPT if you have not played with chat CPT
45:30 don't go there go
45:34 hold please yeah let me make this bigger
45:36 come on there we go
45:38 to chat.openai.com that's the official
45:42 chat website just go there you're gonna
45:44 love it it's gonna be fantastic you may
45:47 not love it doesn't matter just go there
45:49 and start playing and what can you do
45:51 there
45:51 it's kind of like a Google search box
45:53 just ask it to write you a country song
45:56 ask it to write you a poem ask it to
45:58 write you a meal plan that only features
46:01 grapes and kumquats and it will
46:05 um
46:06 so so just go start playing with it and
46:09 and one of the things we talk about in
46:11 here you owe it to yourself
46:13 to explore these large language models
46:15 long enough that you have what we call
46:18 your Kevin McAllister moment which is
46:21 this moment
46:27 it can do that
46:29 um
46:31 and and and just start playing with this
46:34 stuff and then what this channel is
46:35 about pop any comments you have down
46:37 below and I'll do my best to answer them
46:40 normally we have about a hundred people
46:43 in here or so right now we've got more
46:46 than eight thousand so I likely won't be
46:49 able to get up to all the comments but
46:50 but if you like what you're seeing here
46:52 you know tap the screen
46:54 um feel free to ask a question I do have
46:56 mods in here the the group that shows up
46:59 here every night so I do these things
47:00 every night from like eight o'clock
47:02 mountain time I'm in Denver until I
47:05 don't know when I get tired when my
47:06 voice goes and my ass gets numb
47:08 um and the Irregulars are people who
47:10 show up regularly they will
47:12 um they will uh sort of pin any comments
47:16 up here so anyway tell the people that
47:18 they can oh yeah oh so another thing
47:20 okay so one of the things that you may
47:23 be asking or as you're flipping through
47:25 Tick Tock and and you see these annoying
47:28 people and they're like
47:33 those are the people we make fun of in
47:35 here why is that because
47:37 while you can make money with chat EBT
47:43 is just a tool right so so the question
47:47 isn't how do you make money with chat
47:48 gbt the question is what's a problem
47:52 you're trying to solve
47:54 that Chachi BT could help you solve
47:56 faster better cheaper things like that
47:59 all right so
48:02 um
48:03 so yeah so just go play and and you know
48:06 I I also feel like we're we're entering
48:08 an era a very magical era era I'm my
48:12 nickname for the era that we're entering
48:14 is the great Renaissance because
48:16 um these tools really allow you to
48:19 amplify what you're good at
48:22 um I'm I am not a good drawer at all I'm
48:26 not a good illustrator I'm really
48:28 creative I'm really visual I'm really
48:30 good at recognizing good [ __ ] but I'm
48:32 not good at creating it well now I am
48:34 um I was good with words now I'm better
48:36 with words so so these tools really help
48:39 you amplify things that might be kind of
48:42 tangential to your core skills
48:45 and do it in a really fun way and so
48:48 um I think we're we're entering an era
48:51 of a lot of uh reinvention where people
48:53 are going to kind of ReDiscover their
48:55 passions and what they want and all that
48:57 sort of stuff uh what's the link from
49:00 the voiceover lady oh sorry about that
49:02 so hang on a sec
49:04 so I mentioned I mentioned before to the
49:07 folks before lots of folks came in
49:09 um I'm putting together a workshop now
49:11 I've put together a workshop um that I'm
49:14 that I am uh being hosted on another
49:17 Creator's Channel
49:19 um Rhonda Phillips she's a creative
49:21 coach and a voice over actress and she's
49:24 got a whole Community there so this is a
49:25 three-part workshop for creatives called
49:28 the AI launch pad and workshop one that
49:32 starts on October 1st is just an
49:34 overview of here's what's going on in a
49:36 AI here's what's possible Workshop 2 is
49:39 applying AI to your business and
49:41 workshop three is Advanced
49:44 um kind of advanced stuff like voice
49:46 synthesis and animation and video and
49:49 all that sort of stuff so it's going to
49:51 be a really interesting three days it
49:53 happens over three weeks on I think
49:56 Wednesday nights Rhonda's voice.com
49:59 slash events is how you can find that so
50:02 go check that out
50:05 um and again there's a lot of new people
50:07 in here because there's significantly
50:10 more people in here I'm a little caught
50:12 off guard because there's 8 000 people
50:14 in here we normally have about a hundred
50:16 sometimes 50 sometimes 30.
50:18 um so welcome
50:20 if you're if you have not tried chat GPT
50:23 go there right now go there now just
50:27 you're like but you're an old man
50:28 talking about AI why should I trust you
50:31 what are your qualifications sir I want
50:33 to know I want to know just ignore all
50:37 that I have no qualifications I'm
50:39 completely unqualified go to
50:40 chat.openai.com and start playing if you
50:43 don't know what to say there if you're
50:44 kind of confused about what these
50:47 large language models or chat GPT is all
50:50 that sort of stuff that's chat EPT
50:51 that's the official website go to that
50:54 thing down there prompts.chat
50:56 that's a little primer that will tell
50:59 you how these things work and it gives
51:01 you like a whole bunch of prompts that
51:02 you can literally copy and paste into
51:04 chat GPT and get it to do stuff
51:07 um so just go start playing have fun and
51:09 what I'm going to do now is I'm going to
51:11 answer some questions and then
51:14 um yeah welcome everybody all right my
51:16 husband thinks I have lost my mind I
51:19 drop everything to be a to be a regular
51:22 elf I love this yeah I'm sorry I know
51:24 listen there it's it's like uh what is
51:29 what did you do Friday night oh I don't
51:31 want to talk about it why not I I don't
51:33 want to talk about it what did you do I
51:36 kind of watched The Tick Tock Channel
51:38 well that's okay do you watch it for a
51:40 few minutes no I watched it for a couple
51:42 hours well what is it I don't know it's
51:45 some old guy talking about AI
51:48 that's what this is so if you're young
51:51 here
51:52 um if if you're thinking of doing the
51:54 troll or if you've already done it oh
51:56 isn't that cute the geriatrics teaching
51:58 other each other how to use computers
51:59 we've already heard that one you gotta
52:01 you you got to be more creative with
52:03 your trolls than that so anyway okay
52:05 let's let's get rolling and answer some
52:07 questions this is this is cool
52:09 okay
52:11 um let's see
52:15 you're a great actor from what was that
52:18 New Jack City I was in New Jack City yes
52:20 but I didn't The Tick Tock just jumped
52:23 me away from your comments oh wait there
52:26 it is
52:27 it you're a great actor from New Jack
52:30 City I am I I still get checks in fact I
52:34 think I have one here beside me
52:37 I have a I have a residual check which
52:40 these things are collector's items this
52:42 is a residual check from sag-aftra
52:47 for New Jack City for 9.31 cents
52:51 this is back when they paid actors
52:53 residuals I did that I filmed New Jack
52:55 City well whatever the [ __ ] that was 19
52:58 90
53:00 one two I don't know I don't remember
53:03 what it was I'm still getting residuals
53:04 from that
53:06 um my sis called me to help prep for a
53:08 job interview I told her to use pie oh
53:10 that's a great idea that's a great idea
53:12 curve Bruno so if you have not played
53:14 with pie
53:16 good Lord there's 10 000 people in here
53:18 hey everybody
53:20 I'm a little shook it uh my name is Kyle
53:24 Shannon this is the AI learning lab
53:25 welcome if you have any questions about
53:27 AI pop them in the comments below what
53:30 character did you play I played white
53:31 boy number one
53:33 number two I was I was the first White
53:36 Boy
53:37 um that was my character name this was
53:39 in a time where no one was triggered
53:42 this isn't it this is in an era of
53:45 before people got triggered so you were
53:48 allowed to write a character called
53:49 white boy number one
53:53 um let's see what was I gonna show oh I
53:56 was gonna go here
53:57 um
53:58 if you have not played with pi
54:01 it's another interesting one to explore
54:04 so you can go to Pi dot AI you can go
54:07 there
54:08 or ideally go get the PI app so if you
54:12 if you go to
54:14 um the Apple store or Google Play search
54:17 for Pi Pi and then search personal
54:20 assistant it's from a company called
54:22 inflection so make sure you get the
54:24 right one and then it's like chatting
54:26 with chat GPT just a different company
54:28 but you can put it into talk mode
54:30 there's a little like phone icon hit the
54:32 phone icon
54:34 and then you just talk and you don't
54:36 have to submit or anything you talk and
54:38 when you stop it listens and then
54:41 answers you and you can pick one of six
54:42 voices
54:44 um
54:45 it's been designed to be conversational
54:48 to be really interesting and good at
54:50 conversation so go play with it because
54:53 it's not
54:54 you if if you've if you've been hanging
54:57 out on the channel regularly you'll
54:58 you'll understand that that um my
55:01 experience with Siri and and Siri lately
55:04 Siri has gotten a little jealous lately
55:06 and she invades our conversation a lot
55:09 um it's not like Siri or Alexa it's it's
55:12 just a very different thing because it's
55:14 connected to the internet and it's chat
55:15 gbt you can literally ask it anything
55:18 you can ask it to teach you stuff you
55:20 can ask it to make up a game you can ask
55:22 it about local events that night you can
55:24 ask it to coach you through something I
55:27 had um my battery shorted out on my car
55:29 one night and I'm like I I don't know
55:31 what to do I gotta do and I was like oh
55:33 hey thank you very much let me try pie
55:35 and and I I open this stupid chat bot
55:39 and I'm like hey I have this Ford Flex
55:41 and it's acting kind of funny what could
55:44 this be and I said well let me
55:46 troubleshoot with you and it asks me a
55:48 bunch of questions and it gave me two or
55:49 three different possibilities and I
55:51 ended up troubleshooting you know that I
55:53 had a dead battery with it
55:55 um but but it's just wild so so go play
55:57 with that go try that okay
56:00 10K with mostly Vietnamese names and
56:03 comments the same wait and comments same
56:06 as an irregular day every everything in
56:09 this channel is irregular I don't know
56:11 what happened but I'll take I didn't
56:13 listen whether you're from we have
56:15 people from all over the uh all over the
56:17 world come in here so we we regularly
56:19 have folks from Australia from from
56:22 Singapore from the Philippines from
56:24 Europe so
56:26 welcome one welcome all something amiss
56:30 there might be something to miss I don't
56:31 know we we
56:33 got hacked in a positive way
56:36 I'll tell you a funny thing so so in the
56:38 early days I was I was messing not in
56:41 the early days of crypto but about three
56:42 years ago
56:44 um I was trying to I was trying to get
56:47 into crypto and I I bought some
56:48 insurance crypto coin that some friend
56:51 of mine said oh you should buy this
56:53 you'll make a fortune and and I invested
56:55 150 bucks in it and then I parlayed that
56:57 150 bucks into like 75 bucks six months
57:01 later so so so
57:03 I I heard of another crypto coin so I
57:06 took my 75 bucks and I and I bought this
57:09 other crypto with it and then it at some
57:12 point it said you know claim your uni
57:15 here and so I clicked this pink button
57:17 and then I claimed my uni I didn't know
57:20 what it I didn't know what I was doing
57:21 and then I went back into my wallet and
57:24 the balance in my wallet went from
57:26 seventy five dollars to twelve thousand
57:28 dollars
57:30 and I had been airdropped 400 uni coins
57:34 from uni swap when I bought that
57:36 original thing
57:37 um and I thought I was hacked but I'm
57:39 like but but it but I've got too much
57:42 money I was like I don't this doesn't
57:44 make any sense to me but it was an
57:45 airdrop and I sort of learned about
57:47 crypto and how stupid and weird it is
57:50 um I typed fluent via uh they're all
57:52 here to learn about AI from an expert
57:54 well what are your qualifications sir
57:57 what are your qualifications
58:00 um I don't think hey gen does Vietnamese
58:02 yet
58:03 um the the only thing even close to an
58:05 Eastern language is uh is uh
58:09 um
58:09 is Indian it does Hindi so but it
58:12 doesn't do Japanese Chinese
58:15 um Vietnamese any of that stuff yet but
58:16 it will it'll get there
58:19 um I added viewers to you because I like
58:21 you and your AI evangelism oh was that
58:24 you James come out did you do this are
58:27 you responsible for this well you're
58:30 freaking awesome that's
58:32 that's awesome thank you very much I
58:34 appreciate it
58:36 um
58:37 I'm I'm
58:39 a little I uh uh this has never happened
58:43 that's really awesome so thank you for
58:44 that
58:46 um okay let's let's answer some more
58:47 questions uh it's crowded in here
58:50 um you're standing on my foot
58:53 oh man hey all your ten thousand go read
58:56 my stuff uh on AI on medium
58:59 um yeah go read cabrino's thing if you
59:02 if you uh
59:05 let's see I'm trying to think what might
59:07 be good to show here
59:09 there's some new people here so I'm just
59:11 gonna I'm gonna keep doing this I'm
59:12 gonna keep these things on screen right
59:14 now if you're brand new to AI if you're
59:17 trying to figure out like where do I
59:19 start with this what do I do I don't
59:21 know what to do I'm very confused AI
59:24 seems very exciting everyone's talking
59:25 about it what is it
59:27 if you're anywhere in that neighborhood
59:29 cool welcome
59:32 go to chat.openai
59:35 or Claude dot AI either one of those top
59:38 two sites
59:40 the top one is the official chat GPT
59:42 website I'm not affiliated with it in
59:45 any way I'm not making any money off
59:46 this I don't I don't care
59:48 um
59:49 go to go to you know chat.openai.com or
59:52 go to cloud.ai and just start
59:54 interacting with them just start ask it
59:56 to write you a song Write ask it to
59:58 write a poem ask it to write a marketing
1:00:01 plan ask it to write some HTML for your
1:00:04 website or python have it write you a
1:00:06 python application it'll just do it in
1:00:08 fact Let me let me go let me go do some
1:00:11 a little bit of what I call parlor
1:00:12 tricks here
1:00:14 let's jump over and do that because we
1:00:15 have a bunch of new people here so
1:00:18 one of one of the things we like to do
1:00:20 here to AI learning lab is deliver some
1:00:22 value we we like to deliver value for
1:00:24 money and you're like but Kyle you don't
1:00:26 charge anything exactly exactly so we'd
1:00:28 like to give you at least zero dollars
1:00:30 in value all right so we're gonna go
1:00:34 to chat GPT
1:00:37 if you've never been to it
1:00:39 it kind of looks like this when you
1:00:41 start you say log in am I in Chrome oh
1:00:43 no I relaunched I relaunched Edge that's
1:00:46 what's going on
1:00:50 oh this is a joy
1:00:55 Google's asking me to re-log into
1:00:57 everything
1:00:58 hold please
1:01:06 um
1:01:07 oh
1:01:11 you probably just watched me type my
1:01:13 password
1:01:16 there's all my Security's gone hey way
1:01:19 to go how'd you get hacked uh well I was
1:01:22 an influencer for for a while there
1:01:24 before they took everything
1:01:27 Google's new 17-factor authentication
1:01:30 okay
1:01:31 as I was saying so so this is chat GPT
1:01:34 if you've never seen it before the
1:01:36 interface is relatively simple down the
1:01:38 left hand side this is kind of your
1:01:40 prompt history so this is your different
1:01:42 conversations
1:01:43 this is the interface
1:01:45 um I have this little plus icon here
1:01:47 because I paid 20 bucks a month for it
1:01:51 um if you don't pay the 20 bucks a month
1:01:52 that's fine you get uh chat GPT 3.5
1:01:55 which is still really good it's very
1:01:57 fast
1:01:58 um
1:01:59 and then it's basically just kind of
1:02:01 like a Google search box but you can do
1:02:03 kind of anything in here so you could
1:02:05 say to it
1:02:07 um
1:02:09 let's see
1:02:11 uh oh did huh I think Siri might have
1:02:15 taken over my
1:02:18 yeah it did
1:02:20 see I told you Siri was gonna be a pain
1:02:22 in the ass here
1:02:25 uh did uh let's see Escape hold
1:02:32 all right I'll just type it uh write me
1:02:35 a country
1:02:37 song about
1:02:41 um
1:02:42 a big crowd
1:02:46 uh crowd
1:02:48 showing up to my party welcome everybody
1:02:52 there's 11 000 people in here
1:02:56 um so so it just wrote a song it like
1:02:59 you see how fast it's writing now
1:03:02 um it wrote a long one too
1:03:06 um it it understands what songs are so
1:03:08 here's a verse here's a pre-chorus
1:03:10 here's a chorus
1:03:12 um there's verse two so let's see it's a
1:03:14 country song well I sent out the invites
1:03:16 on a hot summer day said y'all come over
1:03:19 we're gonna rock it this way I cleaned
1:03:21 up the barn I don't know I need to be
1:03:24 doing this with a Southern accent a
1:03:25 cleaned up Barn put some hay on the
1:03:27 floor strung up some lights and unlocked
1:03:29 the door
1:03:31 um so it can rhyme it can write songs
1:03:34 um I could also say
1:03:36 um uh what did I do earlier let's see
1:03:39 write me a business letter
1:03:41 that applies for a job organizing
1:03:45 parties that consistently get more
1:03:47 people than were invited
1:03:52 and there there's your professional
1:03:54 business letter boom then I could also
1:03:57 say
1:03:58 um write me a python application
1:04:02 that will allow me to manage all of the
1:04:05 people that show up to my party
1:04:09 hmm
1:04:11 it'll describe it it'll then write the
1:04:13 code you can then take this code and
1:04:16 throw it into replit or some sort of
1:04:19 program like that so if you haven't
1:04:21 played with chat GPT
1:04:25 just play with it long enough until you
1:04:27 have what we call your Kevin McAllister
1:04:29 moment which is this moment
1:04:32 from Home Alone
1:04:33 so so
1:04:36 um
1:04:37 it it will at some point blow your mind
1:04:40 and and if if you haven't seen the The
1:04:42 Matrix movies The Red Pill blue pill
1:04:44 moment the blue pill moment is is you
1:04:47 take the blue pill and everything stays
1:04:48 the same I feel like this AI stuff's
1:04:50 like that there's a Tipping Point where
1:04:53 you realize that these these tools like
1:04:55 chat GPT and and the generative image
1:04:58 tools and things like that are
1:05:00 significantly more powerful than you had
1:05:03 any [ __ ] clue about and and and it
1:05:06 happens to a lot of the people that are
1:05:08 in here sending 5K thank you dunkaroo
1:05:12 how do you send 5 000 people oh wait now
1:05:16 we're down to three thousand so so
1:05:20 5500 or so just evaporated or 6 500 just
1:05:24 evaporated I don't know what's going on
1:05:26 anyway
1:05:30 today today's been a really weird
1:05:33 um it's been a weird day where where it
1:05:36 kind of started out in this sort of off
1:05:38 kind of weird energy but then we had a
1:05:40 bunch of really cool things happen I'm
1:05:43 going to show you something here since
1:05:45 we have so many people in here
1:05:47 um
1:05:48 one of the one of the sites that I have
1:05:50 this list this might be something worth
1:05:53 screenshotting is this is a list of the
1:05:56 image generation tools that I think are
1:05:58 worth your time to play with again I'm
1:06:01 not affiliated with any of these so
1:06:03 mid-journey and ideogram are the two
1:06:06 that are kind of the biggies right right
1:06:08 now
1:06:08 um along with stable diffusion so
1:06:10 Leonardo dot AI kind of represents the
1:06:13 stable diffusion world and then
1:06:16 um
1:06:17 Adobe Firefly and Photoshop generative
1:06:19 fill down below there all of those are
1:06:21 are really good ones this this this this
1:06:23 tool hey gen right here
1:06:26 um
1:06:28 really interesting company they spun
1:06:30 this company up I think with seed
1:06:32 funding and like 50 people one of their
1:06:35 Engineers did a side project called Hey
1:06:37 gen video translate and hey gen video
1:06:39 translate does this
1:06:47 so we've heard from a number of clients
1:06:49 that modular content is important to
1:06:51 them getting more efficient with how we
1:06:53 create content and how we leverage
1:06:55 contents
1:07:02 behind me you see these little green
1:07:05 boxes um that's how we think about
1:07:06 stories at storyline we break stories
1:07:08 down into little atoms right little
1:07:10 modules
1:07:21 so so how you do this I'll show you how
1:07:24 to do it
1:07:26 I can't I can't people it's the future
1:07:30 is here the future's here nothing is
1:07:33 sci-fi anymore except time travel I mean
1:07:36 honest to God like we have we've had the
1:07:39 pandemic and and now we have talking
1:07:42 robots and and they do stuff for us and
1:07:46 other now we've got the Babel Fish thing
1:07:48 so
1:07:51 um let me show you how this thing works
1:07:52 so the the company is called hey Jen
1:07:55 h-e-y-g-e-n
1:08:00 and then you go to oh you go to
1:08:03 labs.hagen.com
1:08:11 and then one of the buttons in there is
1:08:13 video Translate
1:08:15 and
1:08:17 um you basically just drop a video file
1:08:23 I can't start the poll because a comment
1:08:26 is pinned so unpin the comment and try
1:08:28 that again
1:08:31 there we go
1:08:33 um so let me sign in here hold please
1:08:37 uh
1:08:41 so basically you just you just upload a
1:08:44 an English file or whatever file
1:08:46 whatever your native language is and
1:08:48 then you select from let me see how many
1:08:50 they have oh they've added some
1:08:53 Mandarin Portuguese
1:08:56 Korean if we have a bunch of Vietnamese
1:08:59 folks in here they don't have Vietnamese
1:09:00 but
1:09:02 how about Mandarin or Japanese let's
1:09:04 have me should I do Japanese or Korean
1:09:07 what's the vote or Turkish or Mandarin
1:09:11 Mandarin Japanese
1:09:14 or Korean I feel I feel like me talking
1:09:17 any one of those is gonna be hilarious
1:09:19 so so let's go ahead and do this let me
1:09:21 go grab
1:09:24 where is it hold please let me go find a
1:09:26 file
1:09:28 so downloads
1:09:32 those are those
1:09:34 okay here we go
1:09:36 um
1:09:38 so for the meeting on the 23rd here's
1:09:40 what it's going to look like
1:09:43 is that that's video four yeah I think
1:09:46 that's one I haven't done yet okay
1:09:49 so what we're gonna do and close this
1:09:51 we're gonna close this we can close this
1:09:53 what's the vote Mandarin all right we'll
1:09:55 do mandarin
1:09:57 so I'm just going to drag my file there
1:10:00 I'm going to select Mandarin
1:10:02 this should be a trip you can choose to
1:10:06 translate Audio Only they give you two
1:10:09 credits
1:10:11 um for free
1:10:12 and you can do a video of of up to
1:10:17 um
1:10:17 you can do a video of up to five minutes
1:10:19 long
1:10:23 um that's good
1:10:25 all right and so it's going to upload it
1:10:27 there and that'll process for a bit so
1:10:28 I'll keep checking back on that so so
1:10:30 pretty soon we'll we'll have uh we'll
1:10:34 have this this guy talking Mandarin
1:10:36 um so yeah
1:10:38 anyway if you're new here and and a lot
1:10:41 of people are new here there's 6 000
1:10:43 folks in here right now it keeps
1:10:44 oscillating between between
1:10:48 um what's it say in there upgrade to
1:10:50 skip the line estimated position 79
1:10:54 000. so so
1:10:57 um
1:10:58 the uh how do I get back to the home
1:11:02 page here I guess I just go to hey Jan
1:11:04 hey Jen
1:11:13 there's pricing
1:11:18 that's not the lab's pricing so I think
1:11:21 it's about 60 bucks a month for
1:11:25 30 credits
1:11:27 up to five minutes each and then it's
1:11:29 like 179 bucks a month for
1:11:33 um
1:11:35 for 60 credits up to 20 minutes each so
1:11:39 it's not super cheap but considering
1:11:41 what it does the fact that it does you
1:11:44 know translation at this level and
1:11:46 that's how we capture the content but
1:11:48 because we have them in this organized
1:11:50 structured way that approach can be used
1:11:52 for everything from how you capture
1:11:54 original content with traditional video
1:11:56 production to how you leverage these
1:11:58 assets to create social media content
1:12:00 things like that because
1:12:14 social media right so so that's what
1:12:17 that thing does so so considering the
1:12:19 quality that that it's generating that
1:12:21 kind of content with uh it's it's pretty
1:12:24 remarkable
1:12:25 um the fact that I'm 79 000 in line I
1:12:29 don't think we're going to be able to
1:12:30 show we're going to be able to show
1:12:33 um
1:12:35 me speaking Chinese anytime tonight
1:12:38 let's see labs labs
1:12:41 dot hey Jen let's see how fast it's
1:12:44 moving it's funny I watched it live with
1:12:47 the with the co-founder yeah I'm still
1:12:49 75 79 000th in line I watched it live
1:12:52 with the co-founder he basically said
1:12:54 their lead engineer all he's spending
1:12:56 his time doing is oh thank you very much
1:12:58 for the what is that that's a cute
1:13:00 little heart a little heart and a hand
1:13:02 it's so cute
1:13:07 we're back up to 10 000 folks welcome
1:13:10 everybody my name is Kyle I can't even
1:13:12 speak my name anymore normally in these
1:13:15 lives we have I don't know on a good
1:13:17 night 150 200 people and all of a sudden
1:13:20 there's 10 000 people in here and it's
1:13:22 been fluctuating wildly so I assume yay
1:13:25 thank you very much I appreciate that
1:13:27 Ted looks at life
1:13:29 um
1:13:30 so if you're new here
1:13:33 um this is all about
1:13:36 exploring generative AI um I'm an
1:13:38 entrepreneur I've started 12 or 15
1:13:41 companies
1:13:42 um
1:13:43 if you've not played with chat GPT go
1:13:47 check out that
1:13:48 chat.openai.com that's the official
1:13:50 chatgpt website if you haven't played
1:13:52 with it just go there just go there now
1:13:54 if you don't know what to do if you
1:13:56 don't know what to say go to
1:13:57 prompts.chat
1:13:59 um that's a little primer that will tell
1:14:01 you how to talk to these large language
1:14:02 models it's quite remarkable
1:14:05 um it will blow your mind in fact one of
1:14:08 the things we talk about in here is you
1:14:10 should play with chat jpt until you have
1:14:13 your Kevin McAllister moment which is if
1:14:15 you remember the movie Home Alone
1:14:18 that moment
1:14:20 uh at some point when you're using these
1:14:23 generative AI tools something will snap
1:14:25 inside you and you'll be like oh my oh I
1:14:30 didn't know I could do that and then
1:14:32 you're like but it can't do what I do it
1:14:33 certainly can't do what I do and then it
1:14:35 does what you do and then you go oh my
1:14:38 this is something different and so these
1:14:40 things are something different these
1:14:42 three in particular are my top three
1:14:44 this is sort of what I call the adult
1:14:45 table
1:14:47 um of of the large language models
1:14:49 Chachi BT Claude and Bard Bard just got
1:14:53 dramatically better it got this thing
1:14:56 called extensions which is really
1:14:57 interesting it's about to get a major
1:15:00 upgrade with something called Gemini so
1:15:02 so all three of these are kind of in a
1:15:04 war right now for being the best and
1:15:07 then there's another group here that I
1:15:09 call the kids table that includes Bing
1:15:12 perplexity.ai pi and po which are also
1:15:15 interesting to play with and then
1:15:17 um
1:15:18 you've got oh thank you very much
1:15:20 Winston I appreciate that Fireballs are
1:15:22 plenty dang Winston you're the best I
1:15:25 appreciate that thank you all very much
1:15:27 um that's really awesome
1:15:30 um chat GPT is my assistant I call her
1:15:32 Cynthia that's awesome
1:15:34 in fact
1:15:36 um what's your name the The Velvet The
1:15:38 Velvet emeralds
1:15:39 um I gotta say so one of the other
1:15:41 things we do here in the AI learning lab
1:15:43 is is we we rate names The Velvet
1:15:45 Emerald 10 out of 10 solid name don't
1:15:48 come in here with some half-assed
1:15:50 unthought up thought uh thought of name
1:15:52 you come in here with a name like the
1:15:54 Velvet Emerald
1:15:55 is awesome 10 out of 10 on the name
1:15:58 um
1:15:59 what she talked about there is you know
1:16:02 Chachi PT is her friend and she talks to
1:16:04 this thing
1:16:05 um a really good way to think about how
1:16:08 to interact with chat GPT is much less
1:16:10 like you've got to write the perfect
1:16:12 prompt and get the perfect answer but
1:16:14 much more like it's a conversation
1:16:16 you'll notice that
1:16:19 Chachi PT is it's funny one of the
1:16:21 images we have for this for the learning
1:16:24 lab
1:16:26 is
1:16:27 let's see I think it's the other
1:16:29 direction hold
1:16:32 there
1:16:34 is is this little this little yellow lab
1:16:37 or like a golden retriever Chachi PT is
1:16:40 very much like a golden retriever it's
1:16:42 just always there for you it's very
1:16:43 supportive you know you you ask it to do
1:16:46 something it's like sure here's this and
1:16:48 then you tell it no that's not very good
1:16:49 and it's like oh I'm sorry I apologize
1:16:51 that wasn't what you wanted and it'll
1:16:53 try again and it'll just keep trying
1:16:55 again and the more you go back and forth
1:16:57 with it the better it seems to get and
1:16:59 and the nature of how this technology
1:17:02 Works
1:17:03 um is is one of the reasons that happens
1:17:05 in fact what what might be kind of fun
1:17:08 here if you're down for it I guess I
1:17:10 don't know how to do this if you're down
1:17:12 for it
1:17:13 um tap the screen make the hearts like
1:17:15 really fly I if you want me to show you
1:17:17 how Chachi PT actually works it's a bit
1:17:20 of a rabbit hole but in like the next
1:17:23 five minutes I can teach you how Chachi
1:17:26 PT actually works is that something
1:17:27 you'd be interested in is that a good
1:17:29 thumbs up good yes
1:17:31 all right I'm I seem to be getting some
1:17:33 I wish he'd say llama come on llama
1:17:37 llama llama I like llamas llamas is
1:17:41 tasty I like Larry and I like Lenny and
1:17:45 I like Lucy and I like llamas yes
1:17:49 Winston with the Llama thank you very
1:17:51 much we also I'm also I am also a
1:17:55 pavlovian uh creature where if if
1:17:58 someone puts a cowboy hat on me I just
1:18:00 talk stupid
1:18:01 [Laughter]
1:18:03 but but let me let me show you something
1:18:04 here
1:18:06 is this yes
1:18:09 what I'm looking for it is all right
1:18:15 and is this now in a window yes it is oh
1:18:18 everything's coming together
1:18:21 everything's coming up roses
1:18:24 all right so we'll leave the little AI
1:18:27 learning lab in the background because
1:18:29 this this is kind of fun okay so how
1:18:31 does chat GPT work it works like this
1:18:35 we'll start with some definitions so
1:18:37 what is what what is the GPT in in chat
1:18:41 GPT
1:18:43 um
1:18:44 the G stands for generative the P stands
1:18:48 for pre-trained the T stands for
1:18:50 Transformer what does all of that mean
1:18:52 doesn't it actually doesn't mean much
1:18:54 I'll try to try to break it break it
1:18:56 down generative means that these tools
1:18:59 are actually generating original content
1:19:01 one of the misconceptions about them is
1:19:04 that They're copying and pasting from
1:19:05 copywritten work it's not how they work
1:19:08 but but just know that it's generating
1:19:10 content the pre-trained means these
1:19:13 machine learning models have been
1:19:15 pre-trained on
1:19:17 um essentially as much of the internet
1:19:19 as openai could get their hands on or or
1:19:21 these other companies so so it's
1:19:23 basically you don't have to bring your
1:19:25 own data it's already been trained on as
1:19:27 much data as they could get into the
1:19:29 thing to train it
1:19:30 and then the t is it stands for
1:19:33 Transformer that's a piece of technology
1:19:35 that Google invented in 2017 that is the
1:19:38 thing that allowed these tools to be
1:19:40 trained on these massive data sets like
1:19:42 all of the internet okay so that's what
1:19:45 GPT means
1:19:47 um you might also hear the phrase llm or
1:19:49 large language model that just chat EPT
1:19:52 is a large language model we human
1:19:54 beings are a large language model
1:19:56 everything goes in we process it it's
1:19:58 all words that's how we make meaning out
1:19:59 of stuff these are modeled on how our
1:20:02 brains are connected and how they work
1:20:04 and that's why they call them large
1:20:06 language models
1:20:07 there's a thing called the latent space
1:20:09 so this is now we're getting into some
1:20:11 of the geeky stuff and this is where it
1:20:12 gets kind of fun
1:20:14 so what you're looking at here to the to
1:20:16 the right this big white box with all
1:20:18 the scattered dots on it is is a 2d
1:20:21 representation of
1:20:24 mathematical Vector space so you have an
1:20:27 X and A Y and then you have all these
1:20:29 dots that are clustered together well
1:20:30 what are those well each dot represents
1:20:33 a vector which is just a mathematical
1:20:34 number that's tied to a word
1:20:37 and then they're clustered together by
1:20:39 semantic meaning so I'll give you a
1:20:40 specific example we have the dog up here
1:20:42 right you might have one of those
1:20:44 clusters of words that has the word dog
1:20:46 in it and that's like those you know
1:20:48 fluffy things that we pet and you know
1:20:51 love us every day but you might have the
1:20:53 word dog in another cluster where it's
1:20:55 like the guy that did his girl wrong
1:20:57 he's a dog right so so individual words
1:21:00 can have lots of different meanings so
1:21:02 these clusters of meanings are in this
1:21:04 in this mathematical Vector space this
1:21:07 is dramatically oversimplified right
1:21:09 this is 2D space with I don't know
1:21:12 whatever that is you know 500 dots or a
1:21:14 thousand dots
1:21:15 the actual Vector space is a thousand
1:21:18 Dimensions or or just a ridiculous
1:21:21 amount of Dimensions it's hard to
1:21:22 comprehend and it's billions of words or
1:21:24 trillions of words and all these
1:21:26 different things so how does it actually
1:21:28 work
1:21:29 it works like this and this is the more
1:21:31 you learn about this the more remarkable
1:21:33 it is that these things do what they do
1:21:35 when you type a prompt into chat GPT
1:21:39 it's actually predicting what the next
1:21:42 most probable word is in the response
1:21:46 over and over and over and over and so
1:21:49 so here's an example so if I type in a
1:21:52 prompt that says the quick brown fox
1:21:54 jumped over the lazy blank as humans if
1:21:58 you've heard that phrase before we know
1:22:00 that the answer is dog because our large
1:22:01 language model here's that you know
1:22:04 quick brown fox jumps over the lazy
1:22:06 blank and we do pattern recognition and
1:22:08 dog pops into our head and we know it's
1:22:10 dog well the large language models do
1:22:13 something similar what they do is they
1:22:15 kind of
1:22:16 um they they ignore all the stuff that's
1:22:18 not you know maybe animals or whatever
1:22:21 might be in that Vector space so some of
1:22:23 these vectors have a higher probability
1:22:25 than the others and within this semantic
1:22:28 cluster there's one of them that has the
1:22:31 highest probability and if you look at
1:22:34 what that is it's actually just a big
1:22:35 long string of numbers which is it's
1:22:37 it's location and vector space and that
1:22:41 long string of numbers is tied to the
1:22:43 word dog there's a table somewhere that
1:22:45 connects those two and then it sticks
1:22:48 the word dog into that sentence there
1:22:50 are some other words it could have been
1:22:51 but dog is the one it chose
1:22:53 and so when you watch chat GPT right
1:22:57 write you a whole long business letter
1:22:59 or write all this code or write all of
1:23:01 this stuff that's what it's doing for
1:23:04 you know every single word or in some
1:23:06 cases it might be like little small
1:23:08 groups of words but it's doing an
1:23:11 original word you know one at a time
1:23:15 when it writes all that stuff so that's
1:23:17 how the things work this is also why
1:23:20 one of the things that
1:23:22 um people will sort of give chat GPT a
1:23:26 bad time because it hallucinates it it
1:23:30 lies right it you ask it to give you a
1:23:33 citation and it gives you a citation and
1:23:36 then that citation is fake and you're
1:23:37 like wait what happened why is it not
1:23:39 right why did it lie to me is it evil no
1:23:42 no it's it doesn't know what it wrote
1:23:47 it has the cognitive
1:23:50 um awareness of a calculator because
1:23:53 it's technically a calculator you type
1:23:55 in a prompt it turns it into a
1:23:57 mathematical probability and then that
1:24:00 number in Vector space is tied to a word
1:24:02 and it puts that word in there it
1:24:03 doesn't know what that word was
1:24:05 and so it's literally just typing a
1:24:07 bunch of words we as human beings look
1:24:10 at what it writes and say what it's
1:24:12 writing is so human that it must be
1:24:15 sentient it must know what it's doing
1:24:18 it knows probabilistically what it's
1:24:20 doing but it doesn't have any
1:24:22 consciousness of what it said and that's
1:24:24 why it will sometimes predict something
1:24:26 that looks correct but it's not it's
1:24:28 also the reason that that chat in these
1:24:31 tools the more context that you give
1:24:34 them the better it gets
1:24:36 so anyway
1:24:38 um since we have a lot a lot a lot of
1:24:41 new people here welcome my name is Kyle
1:24:43 Shannon this is the AI learning lab if
1:24:45 you haven't subscribed to the Channel
1:24:46 please go subscribe to the channel
1:24:49 um
1:24:50 and and uh if if you like it great you
1:24:53 know follow the channel I mean
1:24:55 um
1:24:56 I try to do these lives every night you
1:24:58 can ask the the uh the Irregulars the
1:25:01 people that are here on a nightly basis
1:25:03 um I haven't missed a night in I think
1:25:05 three months or something like that much
1:25:07 to the much to the
1:25:09 um consternation of my family sometimes
1:25:13 um but they know this is important and I
1:25:14 think this is important and I come in
1:25:16 here to
1:25:18 um just try to demystify this stuff I
1:25:20 think this is we're living in an
1:25:22 incredibly rare moment in history where
1:25:25 we're at the very beginning
1:25:27 of
1:25:29 a a transformative technology that's
1:25:32 going to change almost everything about
1:25:35 how we work and even what work is and
1:25:38 where the value and work is all that
1:25:39 sort of stuff and so this channel is
1:25:41 really just about demystifying it
1:25:43 talking about it answering questions
1:25:45 about it things like that ion China says
1:25:47 maybe sentient intelligence is a form of
1:25:50 real-time statistical inference well so
1:25:52 that's that's actually a really good
1:25:54 point ion China that that um
1:25:59 one of the things that happens when when
1:26:02 people are confronted with chat GPT and
1:26:06 and they're like well I can't do what I
1:26:07 do they get very defensive I can't do
1:26:08 what I do
1:26:09 and and
1:26:12 um
1:26:13 it does
1:26:15 it's very possible that the way our
1:26:16 brains work are kind of a more
1:26:18 sophisticated version of what these
1:26:20 things are doing and and scientists
1:26:22 don't know enough about how they do what
1:26:25 they do that maybe our brains are a
1:26:28 simplified version of what this is doing
1:26:29 I think it's the other way around
1:26:31 um but but these things are going to get
1:26:33 incredibly sophisticated thank you for
1:26:35 that
1:26:36 um and uh yeah it's it's a really
1:26:38 interesting point I mean his wife uh I
1:26:40 don't think they have any all irises I
1:26:43 don't know what that's in reference to
1:26:45 um I don't think they have
1:26:47 all irises
1:26:49 wow the chat feed has slowed down again
1:26:52 no there's there's 13 000 people in here
1:26:55 welcome everyone my name is Kyle Shannon
1:26:58 this is the AI learning lab and this is
1:27:00 the AI learning lab
1:27:02 um
1:27:03 if uh if you're wondering what are my
1:27:06 qualifications I'm an entrepreneur and
1:27:09 um and love just hanging out and talking
1:27:12 about this AI stuff
1:27:14 um and do it every night thank you very
1:27:15 much Aaron I appreciate that
1:27:18 um Let me let me pop down here and and
1:27:20 see some questions I didn't bring enough
1:27:22 gum for the whole class yeah exactly Ted
1:27:25 looks at life he brings Chiclets do you
1:27:27 remember Chiclets Ted he's a he's an
1:27:30 older fella like myself he brings
1:27:32 Chiclets every night for all the people
1:27:34 that hang out here at the AI learning
1:27:36 lab in the in the nightly lives but he
1:27:38 only brings like a hundred Chiclets and
1:27:41 he's normally got more than enough so I
1:27:43 don't know what's happening but uh the
1:27:45 numbers fluctuating a lot so I have no
1:27:47 idea here's another possibility like I
1:27:50 haven't really looked at the actual
1:27:51 comments it's possible not that many
1:27:54 people are actually watching and there's
1:27:56 some sort of tick tock glitch but
1:27:58 welcome if you're new here
1:28:01 um
1:28:03 let's see hi from Texas hello Miss
1:28:07 shalom
1:28:08 Michelle M Michelle M oh Michelle M
1:28:11 Michelin Michelle Lim Michelle
1:28:16 hi from Texas hello from Denver how are
1:28:19 you
1:28:21 all right it's 10 p.m Friday night
1:28:23 beautiful Denver Colorado welcome to
1:28:26 Denver Colorado
1:28:28 the home of the Rockies the home of the
1:28:29 nuggets
1:28:32 the home of joggers and bikers
1:28:34 everywhere
1:28:38 what time we got 1005. all right we're
1:28:41 still good we're still good here's a
1:28:43 little poll
1:28:44 I'll just pop a poll up there go ahead
1:28:46 and take that poll
1:28:48 I don't know
1:28:50 if those things are working or how they
1:28:52 work
1:28:53 um cool say it again in German uh let me
1:28:55 go see what's going on with the the one
1:28:57 I did in Chinese
1:28:59 um reload
1:29:03 oh it's really slow so I was in
1:29:09 79204 and now I'm in or at 2 23 or
1:29:14 something like that now I'm at 204. so
1:29:16 this is probably going to be overnight
1:29:18 um
1:29:19 if if you're new here there's a there's
1:29:22 a tool called hey Jen
1:29:25 um that has a video Translate
1:29:29 um
1:29:30 beta like a proof of concept
1:29:33 that is quite remarkable here I'll show
1:29:35 you
1:29:39 so this is this is a video of me from a
1:29:42 video I did last February and I intercut
1:29:45 this with
1:29:47 um
1:29:49 me speaking English followed by me
1:29:51 speaking Italian this is from hey Jam
1:29:53 Labs so we've heard from a number of
1:29:56 clients that modular content is
1:29:57 important to them getting more efficient
1:29:59 with how we create content and how we
1:30:01 leverage content
1:30:03 modular important
1:30:09 behind me you see these little green
1:30:11 boxes um that's how we think about
1:30:13 stories at storyline we break stories
1:30:15 down into little atoms right little
1:30:17 modules
1:30:27 I mean that's how we capture the content
1:30:29 but because we have them in this
1:30:31 organized structured way that approach
1:30:33 can be used for everything from how you
1:30:35 so anyway so that's quite remarkable
1:30:37 anyway they're they're their servers are
1:30:40 very far backed up and again the company
1:30:43 here is hey Jen
1:30:46 hang on a sec
1:30:51 [Music]
1:30:54 um that that company right there hey Jen
1:30:57 is the one that's got this video
1:30:59 Translate
1:31:01 labs.hagen.com slash video Dash
1:31:05 translate is how you get to it
1:31:08 um
1:31:10 it's looking right now like
1:31:14 the the the backlog of free
1:31:18 um free rendering is so far backed up
1:31:21 that it it's like barely moving so my
1:31:24 instinct is if if you want anything
1:31:25 turned around decently quickly you're
1:31:27 probably gonna have to pay for a
1:31:28 subscription
1:31:30 um
1:31:32 it for what it does it's it's pretty
1:31:36 remarkable it's not that expensive but
1:31:37 it ain't cheap it's like I think the the
1:31:39 cheapest uh subscription for the
1:31:41 translate tool is 60 bucks a month
1:31:45 um so so anyway that's that's what that
1:31:47 one is um that's that's what did that
1:31:49 Italian thing that I just showed you
1:31:51 there
1:31:52 um dude you usually only have 60 people
1:31:53 in here thousands and thousands congrats
1:31:56 I love re-watching these things oh my
1:31:58 goodness there's 14 000 people here
1:32:00 welcome everyone my name is Kyle Shannon
1:32:02 this is the AI learning lab
1:32:05 um I have no idea what's going on
1:32:07 tonight I don't know who's butt we
1:32:10 kissed or what algorithm we tripped over
1:32:12 or someone came in and said they sent 5
1:32:15 000 people our way I didn't know that
1:32:17 was a possibility maybe it is I do not
1:32:20 know but we have a lot of folks here
1:32:23 tonight
1:32:24 um I'm an entrepreneur let me flip this
1:32:27 this way
1:32:29 I'm an entrepreneur my current gig is
1:32:32 I'm the co-founder and CEO of a company
1:32:33 called storyvine which is not an AI
1:32:36 company although we are completely re
1:32:40 um architecting our product Vision with
1:32:43 AI
1:32:44 um and that's really exciting we've been
1:32:46 around for 11 and a half years we do a
1:32:48 lot of work in
1:32:50 um in healthcare and Pharma Alexander
1:32:52 Wright says that Italian thing was
1:32:53 awesome isn't that amazing I mean it's
1:32:55 amazing it's not
1:32:59 my friend ritesh Patel he's a he's a um
1:33:02 he's an agency guy he I've worked with
1:33:04 him for years he and I worked together
1:33:05 back at agency.com in the early days of
1:33:07 the web and and now he's doing a lot of
1:33:09 healthcare marketing stuff
1:33:11 um and he's been doing some of these hey
1:33:13 gen videos and uploading them to
1:33:15 LinkedIn and it's like every time I see
1:33:17 one of his videos he does it in a
1:33:19 different language he did one in Chinese
1:33:20 he did one in Spanish
1:33:23 um and at the beginning of it he he
1:33:25 um he says my name he says you know my
1:33:27 my friend Kyle Shannon told me about
1:33:29 this thing and so every time it's in
1:33:32 another language I hear my name I hear
1:33:34 ritesh say my name in some other
1:33:36 language it's just a trip it's it's just
1:33:39 a trip so
1:33:41 um that's what's going on with this
1:33:42 stuff oh it's you buddy hey James you're
1:33:44 the one doing this
1:33:46 I don't know how you're doing this but
1:33:48 this is awesome thank you very much
1:33:52 um if you're new here
1:33:54 um I am normally not flustered I am
1:33:57 normally just a a smarmy smart ass uh
1:34:01 that that acts like he has it all
1:34:02 together but not tonight it's just crazy
1:34:05 you people are just nuts you're just
1:34:08 nuts I'm telling you you're crazy all of
1:34:10 you from one to the other to the other
1:34:12 anyway
1:34:14 um if you haven't played with chat GPT
1:34:17 please go there right now
1:34:19 chat.openai.com
1:34:22 and if you're like oh is he just making
1:34:25 money off chat GPT is this just him
1:34:27 making a commission no I'm not I do this
1:34:31 for free I'm here every night ask the
1:34:32 Irregulars the ones the ones that are
1:34:35 tapping the screen and and answering
1:34:37 other people's comments below and
1:34:40 pinning the the questions and modding
1:34:42 here
1:34:43 um I come here every night these are
1:34:45 really just here more as a public
1:34:47 service announcement the whole idea here
1:34:49 is to demystify these things so go to
1:34:51 chat GPT or go to Claude or even Bard
1:34:54 all three of these are worth you playing
1:34:58 with long enough until you have a moment
1:35:01 where you realize
1:35:06 oh my God these are different
1:35:08 something's different here these are not
1:35:10 just Fancy Google
1:35:12 they're they're deeply profound they're
1:35:14 going to change a lot
1:35:15 we are about to get a series of major
1:35:19 announcements over the next two months
1:35:22 that these things are even going to get
1:35:24 better and better and better
1:35:26 um November 1st
1:35:29 um is General availability for
1:35:32 Microsoft's co-pilot being available
1:35:35 within office 365. what that basically
1:35:38 means is that chat CPT will be built
1:35:41 into all of the office suite within
1:35:43 Microsoft products Microsoft Word slides
1:35:47 Excel all of that stuff Google announced
1:35:51 the same thing for their Office Products
1:35:54 two weeks ago
1:35:57 so these tools right now are kind of off
1:36:00 to the side like you go to chat GPT you
1:36:02 go to that thing
1:36:04 they're very very very quickly going to
1:36:06 be integrated in all of the tools we're
1:36:09 already using so if if you are not
1:36:12 starting to get curious about these and
1:36:15 get up to speed and understand what they
1:36:17 do the time is now so so here's what I'm
1:36:21 going to do since since there's so many
1:36:23 new people keep seem to be rolling in
1:36:25 every every few minutes
1:36:28 um I'm gonna just go play with
1:36:31 okay
1:36:32 so this is this is chat GPT
1:36:35 down the side here you've got this is
1:36:37 this is my chat history this is
1:36:38 conversations I've had with this thing
1:36:42 and and oh thank you very much that is a
1:36:45 lovely
1:36:46 the Stars the Stars they were falling
1:36:50 from the sky was remarkable
1:36:53 um so in fact form a union yeah exactly
1:36:56 we're gonna we're gonna need to do
1:36:58 something because work is about to
1:37:00 change pretty dramatically
1:37:02 um okay
1:37:04 so
1:37:06 um
1:37:07 there's a couple of different modes of
1:37:08 chat GPT if you pay the 20 bucks a month
1:37:11 for their what's called their Plus
1:37:13 subscription you get gpt4 you also get
1:37:17 plug-ins and you get Advanced Data
1:37:20 analysis Advanced Data analysis is
1:37:22 absolutely [ __ ] mind-blowing
1:37:25 um I'll I'll show that because there's
1:37:27 something there we may we may lose a lot
1:37:29 of viewers but that's okay because I if
1:37:32 I were you I'd hang around and just see
1:37:34 this if you haven't played with chat gbt
1:37:36 yet
1:37:37 um that's if you pay the 20 bucks even
1:37:38 if you don't pay the 20 bucks
1:37:41 GPT 3.5 is is really really good so it's
1:37:45 just like a Google search bar it's just
1:37:46 a bar you can type anything into it so
1:37:49 we're going to type into it
1:37:51 um let's see
1:37:53 um
1:37:54 write me a social media marketing plan
1:37:56 for my car wash business called shiny
1:38:01 turtle
1:38:04 now I didn't really give it any context
1:38:06 whatsoever but there's a marketing plan
1:38:09 and you're like but Kyle it's writing it
1:38:11 so fast it can't possibly be good
1:38:13 and then we'll scroll to the top here
1:38:15 and let's go look at some things
1:38:17 all right creating a social media
1:38:20 marketing plan
1:38:22 for your car wash business shiny turtle
1:38:24 can help you increase brand awareness so
1:38:26 step one start with your goals and
1:38:28 objectives great so it gives us some
1:38:30 goals and objectives we may or may not
1:38:32 like those know your target audience
1:38:33 these are right choose your social media
1:38:35 platforms create compelling content
1:38:38 um develop a Content calendar and get
1:38:40 calendar engage with your audience right
1:38:42 so good basic stuff now if you're like
1:38:44 well this seems like a little obvious in
1:38:46 this and that the only context I gave it
1:38:49 was write me a social media marketing
1:38:50 plan if I gave it a lot more context
1:38:53 like I said hey you're a social media
1:38:55 master and you you excel in helping
1:39:00 small businesses drive more traffic to
1:39:02 their business and my business is called
1:39:04 shiny turtle and we're we specialize in
1:39:08 this kind of thing and that kind of
1:39:10 thing
1:39:12 um oh I I can't uh I can't start the
1:39:15 poll because the uh whatever you call it
1:39:18 someone a message is pinned so unpin the
1:39:21 message and then I can start it
1:39:23 there you go
1:39:25 um
1:39:26 so this this would get much much better
1:39:28 over time and so so even even here I'm
1:39:31 going to say
1:39:32 um
1:39:33 um write me 10
1:39:35 catchy ad headlines
1:39:38 um for shiny turtle
1:39:40 that will get people
1:39:42 uh to want to try it out
1:39:47 certainly
1:39:49 unleash the shines today shiny turtle
1:39:51 car wash where clean meets gleam
1:39:54 roll in dirty roll out shiny
1:39:58 experience the turtle magic today
1:40:00 Revitalize your ride with shiny turtle
1:40:03 we make cars Sparkle Dirk's worst
1:40:06 nightmare shiny turtle car wash
1:40:08 so I missed my calling and voiceovers
1:40:11 apparently
1:40:12 um but it can do that kind of stuff what
1:40:14 else can it do
1:40:15 um we can say
1:40:16 make me a table comparing the different
1:40:19 kinds of car waxes I can use on my car
1:40:24 that
1:40:27 accentuates the the advantages of each
1:40:32 I don't know that wasn't good English
1:40:33 but that's okay but look it's making us
1:40:36 a table
1:40:37 um what are the kinds of wax I don't
1:40:40 know I didn't know this going in
1:40:42 Carnauba wax here's the advantages of it
1:40:45 here's the ideal uses here's the
1:40:47 application frequency synthetic polymer
1:40:49 ceramic coating and Spray Wax oh and
1:40:53 paste wax
1:40:54 so it can it can do tables I could also
1:40:57 say
1:40:58 um
1:40:59 write me a python application that helps
1:41:01 me calculate the profit for my car wash
1:41:04 business
1:41:07 it'll explain what it does it'll write
1:41:08 the code
1:41:10 so this is chat EPT this is way way more
1:41:15 than ask catch EBT about yielding about
1:41:19 dividend yielding stocks James for you
1:41:22 anything
1:41:24 um okay so let me start a new chat and
1:41:26 I'm going to go to chat gpt4 because it
1:41:28 might do it better it'll be slower but
1:41:29 it'll be better so what did you say
1:41:32 dividend
1:41:34 can you pin that message again if you
1:41:38 dividend yielding stocks okay
1:41:42 I want you to act as an expert stock
1:41:45 broker
1:41:46 that works
1:41:49 for Merrill Lynch and deeply understands
1:41:52 the market
1:41:55 period I'm trying to understand dividend
1:41:58 yielding stocks
1:42:01 and why I might consider them in my
1:42:04 Investment Portfolio is that something
1:42:06 you can help with
1:42:09 certainly a dividend yielding stock is
1:42:11 one where the company pays out a portion
1:42:13 of this thing for example company pays
1:42:15 out 12. and so look now it's even doing
1:42:17 things like creating you know headlines
1:42:19 a numbered list with bullet points I
1:42:21 didn't ask it to do that it's just doing
1:42:22 that
1:42:24 um risks involved right so so let's see
1:42:26 what it says
1:42:28 um why you would consider it's a stable
1:42:30 income stream it's a reinvestment
1:42:32 opportunity there's a there's potential
1:42:34 for lower volatility there's tax
1:42:36 efficiency and there's a hedge against
1:42:37 inflation
1:42:39 risks dividend Cuts or omissions Market
1:42:42 risks interest rate sensitivity
1:42:44 diversification strategy dividend
1:42:47 investment strategy second or selection
1:42:49 criteria final thoughts so you know I
1:42:53 asked it to act like a Merrill Lynch you
1:42:56 know person that kind of narrowed the
1:42:57 context a little bit and now gave us
1:42:59 this big long thing from Michael
1:43:02 um chat EBT helped me create an advanced
1:43:04 image creation tool for my Custom
1:43:06 Framing business perfect so so this is
1:43:10 one of the things
1:43:12 when you start playing with chat gbt
1:43:17 it's easy to fall into the Trap of just
1:43:20 thinking it's a toy
1:43:22 because the initial things you play with
1:43:24 are kind of toys write me a poem write
1:43:26 me a song
1:43:29 where it really starts to melt your
1:43:31 brain and and and and then your brains
1:43:34 drip out of your ears and then your eyes
1:43:36 cross and your tongue starts quivering
1:43:38 is when is when you
1:43:41 make a mental shift from this is a toy
1:43:44 to let me see what this can do for my
1:43:47 business or for my job or for my
1:43:50 thing that I enjoy the most in the world
1:43:54 you know I love stamps and I want to I
1:43:57 want to learn everything there is to
1:43:58 know about stamps and then you ask this
1:44:00 thing something that you think is one of
1:44:01 the most obscure questions in your area
1:44:04 of expertise and it [ __ ] answers it
1:44:06 like that
1:44:08 and you go
1:44:11 oh my God I I had no idea it's one of my
1:44:15 favorite moments is when people and by
1:44:17 the way a bunch of new people just
1:44:18 joined one of my favorite moments in
1:44:21 this Channel and in general with this
1:44:23 generative AI stuff let me flip this
1:44:26 is
1:44:27 when
1:44:29 especially people who were cynical about
1:44:31 it or kind of sitting on the sidelines
1:44:33 like well I heard Chachi PT hallucinates
1:44:36 and I hear and I hear it's stealing
1:44:38 people's work and I hear we get a lot of
1:44:40 people that are like huh
1:44:43 yeah I suppose it's just hype it's just
1:44:46 hype it's just how you get a lot of that
1:44:47 right and almost a hundred percent of
1:44:50 the time those people have not tried it
1:44:54 one of my favorite things in the world
1:44:56 is when you sit down with them and I
1:45:01 would encourage you to do this play with
1:45:02 this stuff and as you learn it start
1:45:04 talking about it with people share it
1:45:05 with people
1:45:07 and
1:45:08 uh there's a pinned comment I can't
1:45:10 start the poem
1:45:12 um
1:45:13 show them chat GPT and what what ends up
1:45:16 happening is
1:45:18 let me see how to close that now I don't
1:45:20 want to delete the poll dude there we go
1:45:22 um
1:45:24 what ends up happening is those cynical
1:45:26 people
1:45:27 realize how powerful this thing is and
1:45:31 then they go oh
1:45:33 and then you can just see their mind
1:45:35 kind of open up and there's it's
1:45:37 initially excitement
1:45:39 and then it's like a little bit of
1:45:41 concern like Oh that's oh that's kind of
1:45:44 interesting and it's like but but it
1:45:46 can't do my job
1:45:48 and then they do something about their
1:45:50 job and it can
1:45:53 at some level and sometimes way better
1:45:55 than you think it should be able to
1:45:58 there's another pinned comment there
1:46:00 from Jim which I want to get to
1:46:03 um
1:46:05 and and that's the moment where there's
1:46:07 kind of it's a threshold you can't come
1:46:09 back come back from
1:46:11 once you realize how powerful these
1:46:13 tools are you kind of can't unlearn it
1:46:16 they're not going to be uninvented
1:46:17 they're getting integrated in everything
1:46:19 we do
1:46:21 um so yeah so so go there right now
1:46:25 chat.openai.com that's the official chat
1:46:27 GPT website go to claude.ai that's a
1:46:30 competitor to chat GPT that's got some
1:46:32 really interesting strengths go to
1:46:34 bard.google.com any one of those three
1:46:36 are a great place to start
1:46:38 uh can it develop business plans
1:46:41 [ __ ] yeah
1:46:43 in fact let's let's go
1:46:46 let's let's go do a couple of business
1:46:48 plans what's your so new level up uh
1:46:51 type in
1:46:53 um what your what your business is or
1:46:55 like category of your business and then
1:46:58 let's go make a business plan
1:47:02 um and if you could if whoever is
1:47:03 modding can pin whatever new level up
1:47:06 types there unless it unless they're no
1:47:09 longer here which is possible there's a
1:47:12 lot of volatility in the in the lab
1:47:14 tonight we were just up to 17 000 now
1:47:17 we're down to ten thousand we're all
1:47:19 over the place this is wild though this
1:47:21 is fun Freight dispatch okay perfect
1:47:23 so we're gonna say
1:47:27 I want you to act as a business
1:47:28 consultant that specializes in creating
1:47:31 business plans
1:47:35 for one of the
1:47:37 big consulting firms period you have a
1:47:41 friend
1:47:43 who is
1:47:46 starting a freight dispatch business
1:47:49 and
1:47:51 wants to do it right and make sure that
1:47:55 they have a solid business plan in place
1:47:58 period
1:47:59 can you recommend three different
1:48:01 strategies
1:48:03 for creating a business plan that might
1:48:06 be appropriate for a freight dispatch
1:48:08 business period
1:48:10 I don't want you to start writing the
1:48:12 plan yet I just want you to give me the
1:48:14 Frameworks so that I can decide which
1:48:17 one I want to start with
1:48:19 so notice that you know I gave it a lot
1:48:22 of context and I'm even doing a thing
1:48:24 like I don't want to just jump into is
1:48:27 this 17 000 verified humans that's
1:48:29 really funny Ted looks at life I have no
1:48:32 idea I don't know what's happening but
1:48:34 I'm digging it James is uh James is
1:48:36 tossing some people our way please go
1:48:39 follow the Channel please go you know
1:48:41 subscribe if you want to hang out here
1:48:44 we're here every night and
1:48:47 um
1:48:48 we do these things for I don't know two
1:48:51 and a half or three hours live every
1:48:52 night I basically just do them until my
1:48:55 voice blows out or my ass gets numb
1:48:57 which it's it's getting close right now
1:49:00 but this is just wild this is fun uh
1:49:04 okay so let's see what it did here
1:49:09 so there's the porters five forces
1:49:12 framework so the objective is to
1:49:14 understand the competing forces within
1:49:16 the industry
1:49:17 to identify areas of strength and
1:49:20 potential opportunity that's pretty cool
1:49:22 the business model canvas I've heard of
1:49:24 and the SWOT analysis so let's say so so
1:49:27 let's say what we want to do right here
1:49:29 is is can you um
1:49:33 can you combine
1:49:35 the advantages of Frameworks one and two
1:49:40 and create a hybrid framework that is
1:49:44 simple yet powerful
1:49:46 and will help your friend position their
1:49:50 business for success and investment
1:49:59 sure it's called the Strategic business
1:50:01 positioning canvas
1:50:06 so you've got your internal business
1:50:07 model components the value proposition
1:50:09 customer segments and relationships
1:50:11 revenue streams Key activities and
1:50:13 resources key Partnerships cost
1:50:15 structure and then you have external
1:50:17 industry analysis components this is
1:50:19 really sound steps to implement start
1:50:22 with internal components okay great
1:50:25 so
1:50:27 um so I'm going to say
1:50:29 no this business framework is fine what
1:50:33 I'd like you to do now is
1:50:37 make a first attempt at writing this
1:50:41 plan for the Freight business for your
1:50:44 friend
1:50:45 include as many details as you think are
1:50:48 important and even include
1:50:51 industry insights as best you can
1:50:57 and this should give us so so one of the
1:51:00 things that chatgpt is pretty brilliant
1:51:02 at
1:51:04 is there's a there's a comment pin so I
1:51:08 can't start the poll
1:51:11 ing that is if you get good at asking a
1:51:14 couple of framing questions and you know
1:51:16 and then get get it to do something like
1:51:18 this
1:51:19 this business plan will probably be at
1:51:23 about 80 percent of good
1:51:25 like it'll probably be actually pretty
1:51:28 close to
1:51:30 what you would end up with if you took a
1:51:33 lot of time and read a bunch of business
1:51:34 books and melded two business plans or
1:51:37 hired a consultant that last 20 percent
1:51:40 is
1:51:43 is the part where you start
1:51:45 participating again so so historically
1:51:48 how I've worked is you know
1:51:51 if something is compelling enough to me
1:51:53 or I've got to do something for work
1:51:55 I'll take three or four days or three or
1:51:58 four weeks I've got 80d so sometimes I
1:52:01 can just crank [ __ ] out like it's
1:52:03 nothing other times it takes me weeks
1:52:05 and I will sort of get into my hyper
1:52:08 focused tunnel and I'll I'll take
1:52:10 however long it takes and I'll get to
1:52:12 like a good first draft and that first
1:52:14 draft might be at like I don't know 85
1:52:16 or 90 percent of where I want it and
1:52:18 then I spend like the next three days
1:52:20 you know finalizing it what happens with
1:52:22 chat GPT is you can get to that 80 85
1:52:25 percent draft in [ __ ] five minutes
1:52:30 so instead of it taking me three days or
1:52:33 three weeks or however long it takes me
1:52:35 to get my [ __ ] together
1:52:36 I can have like what we just did here
1:52:39 let's go look at a little bit of it for
1:52:41 this Freight business
1:52:43 um
1:52:45 value proposition efficient reliable
1:52:48 cost effective Freight dispatch Services
1:52:50 differentiation utilization of advanced
1:52:53 technology for real-time tracking
1:52:55 scheduling and communication ensuring
1:52:57 seamless Service Great it may or may not
1:52:59 be that but you could obviously fine
1:53:01 tune that
1:53:02 um here's your customer segment here's
1:53:04 your revenue streams right so it's got
1:53:06 all that stuff there so I can now take
1:53:08 that
1:53:10 and what would have taken me three days
1:53:12 I now have in five or ten minutes
1:53:14 and I can spend the next three days
1:53:16 refining this
1:53:18 so I want to show you something that
1:53:20 that I actually think is is not only
1:53:22 relevant but I think it's quite
1:53:24 important I think it's quite significant
1:53:25 there's a A Wharton business school
1:53:27 Professor named Ethan Malik
1:53:30 and if you don't follow him you should
1:53:32 follow him brilliant [ __ ] dude
1:53:35 like right on the cusp of all of this
1:53:40 stuff and and he just did this piece
1:53:42 he's his blog is called one useful thing
1:53:46 you should subscribe to it I don't get
1:53:49 paid by him there's always people in
1:53:52 here it's just a sham he just wants us
1:53:54 to go to subscribe to youth in my league
1:53:56 stand because he gets Kickback you know
1:53:57 every time we subscribe to Ethan Malik's
1:53:59 newsletter Kyle gets like ten dollars
1:54:01 like if I did wouldn't that be [ __ ]
1:54:03 awesome but I don't so anyway
1:54:05 don't be so cynical get a hug have
1:54:08 someone hug you
1:54:09 um so this article that he wrote is
1:54:12 called centaurs cyborgs and the jagged
1:54:15 Frontier
1:54:17 and it it's a group of
1:54:20 um people from Boston Consulting Group
1:54:22 where they took a series of consultants
1:54:25 and they gave them access to gpt4 chat
1:54:28 gpt4 the thing I just showed you
1:54:31 and what they found this is [ __ ]
1:54:34 mind-blowing so that little chart but
1:54:36 the the numbers under the chart right
1:54:37 here Consultants using AI finished 12.2
1:54:42 more tasks
1:54:45 on average completed them 25 more
1:54:49 quickly and produced 40 percent higher
1:54:52 quality results
1:54:54 so more
1:54:56 faster better more faster better
1:55:02 12 percent more 25 percent faster 40
1:55:06 percent better using chat GPT this is
1:55:09 Boston Consulting Group here's another
1:55:11 thing that that this finding that that
1:55:13 this study found
1:55:15 which again is mind-blowing
1:55:19 this chart so on the left these were the
1:55:23 bottom half of Performing Consultants
1:55:27 the lower performing Consultants had a
1:55:30 43 increase in quality of work
1:55:33 the higher performing Consultants only
1:55:36 saw a 17 increase at work
1:55:39 so what's what's happening is the lower
1:55:42 performers are performing
1:55:44 the increase in their performance is
1:55:47 dramatically higher for the lower
1:55:49 performers in a company so it's making
1:55:51 your best people better but it's making
1:55:53 your not best people dramatically better
1:55:57 so so think about this this this is
1:56:01 this is a technology that kind of looks
1:56:05 like a Google search bar that when you
1:56:07 actually start applying it to business
1:56:09 it's having
1:56:11 you know not only a statistically
1:56:14 relevant impact but like a dramatic
1:56:17 impact
1:56:19 like that's what's going on right now
1:56:21 and this is with chatgpt in its current
1:56:24 state one of the things we joke about
1:56:25 here in the in the in the lab
1:56:29 is
1:56:31 you know I make fun all the time of
1:56:33 these tools
1:56:34 are are very primitive because they're
1:56:36 very new right the the generative AI
1:56:39 stuff this stuff that's like you can
1:56:41 type in some words and Magic [ __ ] comes
1:56:43 out that
1:56:46 um has really been only been available
1:56:48 since
1:56:50 um
1:56:51 since November of 2022 so it's like 10
1:56:54 months old so all of these tools are
1:56:57 what we like to call in the AI learning
1:57:00 lab janky pieces of [ __ ]
1:57:03 they're they're all just very primitive
1:57:05 so even with them in their primitive
1:57:08 state
1:57:09 without them being multimodal and and
1:57:12 truly you know
1:57:14 um do reasoning and and and take
1:57:17 autonomous action all of the things that
1:57:19 they could be it's having that kind of
1:57:21 effect
1:57:22 12 more 25 faster 40 percent better
1:57:28 holy crap I said magic [ __ ] I did I
1:57:31 swear here and it is Magic [ __ ] but it's
1:57:35 also janky [ __ ] the tools are janky
1:57:37 right now
1:57:38 and Yankees a technical term if you're
1:57:41 not if you're not a technologist you
1:57:43 probably yeah but I had a partner life's
1:57:45 tough it's tougher when you're stupid
1:57:47 all right uh by the way if you stick
1:57:50 stupid hats on me I do respond to them
1:57:52 like one of those NPCs oh uh hot dog hot
1:57:55 dog don't do that
1:57:57 [Laughter]
1:58:01 um anyway
1:58:02 um ADHD also got any tips that get you
1:58:05 into hyper Focus
1:58:09 yeah
1:58:11 yeah I mean I mean I have I have tips
1:58:13 about
1:58:14 these tools and and what it's done for
1:58:18 me with my ADD
1:58:20 um
1:58:22 I
1:58:27 the only other time in my life where I
1:58:29 felt
1:58:30 um
1:58:33 where I felt
1:58:36 overwhelmed
1:58:40 to the point that my ADD couldn't keep
1:58:42 up with it
1:58:44 was in the early days of the world wide
1:58:46 web and when I was when I was starting
1:58:47 agency.com in in 1994 1995.
1:58:52 there was so much to learn and things
1:58:54 were happening so fast I literally
1:58:57 couldn't keep up with it right one of
1:58:59 the joys of ADD
1:59:01 is that you can synthesize things really
1:59:03 quickly and with very little information
1:59:05 you can sort of grock stuff and if
1:59:07 you're into it you can do a deep dive on
1:59:09 it
1:59:13 the speed
1:59:15 and depth
1:59:17 that these tools are are sort of being
1:59:21 introduced and evolving are absolutely
1:59:24 overwhelming there's there's absolutely
1:59:26 no way I or anyone else can keep up with
1:59:30 it in in any kind of way I pay attention
1:59:32 to this stuff multiple times a day and
1:59:35 honest to God I'll take 12 hours where
1:59:37 I'm going and I just do something else
1:59:39 and I don't think about this [ __ ] and I
1:59:41 miss seven things it's it is there it is
1:59:44 impossible to keep up with it so if you
1:59:46 think you're behind if you think you
1:59:48 can't keep up even if you went 24 7 with
1:59:51 this stuff you can't keep up so that's
1:59:53 the the kind of um
1:59:57 the overhang of this is that there's
2:00:00 literally no way to get your head around
2:00:02 it so so just give that piece up so
2:00:04 that's that's a big piece of it now the
2:00:06 hyper the hyper Focus thing
2:00:13 all right
2:00:15 historically
2:00:17 before chat GPT if I had an idea
2:00:22 it would it would have to be
2:00:25 you know how you have those ideas so
2:00:27 sometimes you have an idea and it's just
2:00:28 like a brilliant idea and you kind of do
2:00:29 like a quick hyper Focus but then you
2:00:31 have those other ideas where it's an
2:00:33 idea that haunts you right and you think
2:00:37 about it and then you're like oh that's
2:00:38 pretty cool I should do that and then
2:00:39 you forget about it and then like two
2:00:41 days later you're like that and then two
2:00:43 days later you like wake up from asleep
2:00:45 and you're like oh I really got to do
2:00:46 this thing and then and then you go into
2:00:48 hyper Focus but it's it's it's a it's a
2:00:50 big idea and it's one that can't escape
2:00:54 your brain and then at some point you
2:00:55 give into it and you hyper focus and
2:00:58 then with ADD there's like a thousand
2:01:00 other ideas and almost all of those
2:01:02 ideas just sort of pass through you know
2:01:05 they pass through squirrel squirrel
2:01:06 squirrel squirrel squirrel
2:01:10 so so because of that Dynamic you know
2:01:13 you you tend to only do hyper Focus I
2:01:16 don't know
2:01:17 every
2:01:19 three weeks depending depending on your
2:01:22 your manicness it might be once a week
2:01:24 it might be twice a week it might be
2:01:26 once or twice a month
2:01:28 what's happening with chat GPT is that
2:01:33 I can have just a whisper of an idea
2:01:36 like not even not even a super
2:01:38 compelling idea but it's like oh that's
2:01:40 kind of interesting
2:01:41 and then what I'm having to train myself
2:01:44 is is go
2:01:46 um get it out of your head and into chat
2:01:48 GPT so I'll have like a whisper of an
2:01:50 idea I hop into chat GPT
2:01:53 and then I just start playing with it
2:01:56 and what happens is I get to 80 of an
2:02:00 articulated vision of or not even Vision
2:02:03 I get to an articulated
2:02:07 um
2:02:08 instance of that idea sitting in front
2:02:10 of me
2:02:12 and and then I can spend 15 minutes in a
2:02:16 hyper focused tunnel
2:02:18 refining it or I can look at it and go
2:02:20 ah it wasn't such a good idea but it's
2:02:22 like instant and and I can have I'm I
2:02:24 find myself having
2:02:27 instead of like two or three of those a
2:02:29 month where I was doing that
2:02:31 historically I'm sometimes having two or
2:02:34 three ideas a day
2:02:36 and definitely you know multiple days a
2:02:39 week it's one of the reasons I'm doing
2:02:40 these lives is because
2:02:45 one of the purposes these lives serve
2:02:48 for me is a way for me to process what
2:02:51 what the hell is going on in my brain
2:02:54 because these tools are amplifying my
2:02:58 ideas in ways that I've never been able
2:03:01 to do so that that study that Ethan
2:03:04 Malik showed about the BCG Consultants
2:03:08 being faster better and and and uh and
2:03:11 producing more stuff I'm finding the
2:03:13 same thing I'm I'm creating more better
2:03:16 faster ideas
2:03:19 and the other thing is
2:03:23 where a lot of times my ADD would be
2:03:26 crippling where I sort of have an
2:03:28 initial idea but then if it required
2:03:31 actual research or if it required actual
2:03:34 paying attention to it
2:03:36 it would fall away because I would just
2:03:38 be off to some other thing and I'd be
2:03:40 like oh yeah that was a good idea but oh
2:03:41 it's too much work and I wouldn't do it
2:03:42 well this now does all the research it
2:03:45 organizes [ __ ] you can have four
2:03:48 different ideas and you literally copy
2:03:50 and paste those ideas together into a
2:03:52 single chat GPT session and say hey over
2:03:55 the past week I had these four ideas do
2:03:57 they have anything in common and it'll
2:03:59 it'll find the connection between them
2:04:01 you're like oh that's really interesting
2:04:02 yeah I hadn't thought about that can you
2:04:04 come up with a fifth idea that is the
2:04:06 umbrella for them sure here it is
2:04:09 Sam corn for capturing ideas I highly
2:04:12 recommend
2:04:13 reai they have excellent transcription
2:04:16 and
2:04:19 hang on
2:04:20 magic summarization all right
2:04:22 r-e-a-i-a-u-r-a-i
2:04:26 that sounds very cool let me go look at
2:04:29 that
2:04:30 re.ai
2:04:35 oh they have cool colors too
2:04:38 said I'm a sucker for that that dark you
2:04:41 know the uh the dark uh black with the
2:04:44 the blue thing
2:04:48 um
2:04:49 I can draft emails record and transcribe
2:04:51 voice memos create meeting notes write
2:04:53 essays correct your grammar okay cool
2:04:56 all right so this is like an AI powered
2:04:59 chat bot that you could just chat ideas
2:05:01 into that's super cool okay I'll
2:05:03 definitely check that out
2:05:05 um I ran 5K off so so you can ramble
2:05:07 into a voice note for a while and then
2:05:09 it totally cleans it up and summarizes
2:05:11 it that's super cool in fact let's see
2:05:14 I'm going to write this down outside of
2:05:16 my computer because I have 80d even if I
2:05:19 don't nothing will happen with it
2:05:22 but here here's my notebook
2:05:25 any other people have ADD
2:05:28 [Laughter]
2:05:30 Ed all right
2:05:32 uh
2:05:33 re dot a i all right I'm definitely
2:05:37 gonna start playing with that that's
2:05:38 awesome thank you Sam Sam corn is one of
2:05:41 the Irregulars here the Irregulars are
2:05:43 the people who show up here every night
2:05:44 and hang out with me they're awesome
2:05:47 um they're really supportive they're
2:05:48 also supportive of the community if you
2:05:50 have questions and comments below
2:05:52 um uh you know put them down below happy
2:05:54 Friday night everyone
2:05:56 um
2:05:57 the the at the peak tonight we had 17
2:06:00 000 people in here we still have 4.8 000
2:06:02 people in here which which is a little
2:06:05 odd this this the funny thing about Tick
2:06:08 Tock is this was a little intimate
2:06:09 Bistro for like three and a half months
2:06:12 and all of a sudden tonight you know
2:06:14 we're like playing a playing a [ __ ]
2:06:17 theater or a stadium a small stadium
2:06:20 um so it's super cool to have everyone
2:06:21 here and uh and
2:06:26 it's just super cool so anyway welcome
2:06:29 welcome everybody that's new here
2:06:31 um we'll get the janky pieces of [ __ ]
2:06:33 thing out of the way and then let's see
2:06:36 we'll do yeah that's a good one all
2:06:39 right Beauty all right tons of Post-it
2:06:41 notes yeah exactly Sherry Banks knows
2:06:44 the add game and by the way if you're
2:06:47 here and didn't know it
2:06:52 um this is an inside joke here at the AI
2:06:54 learning lab so one of my pet peeves
2:06:58 are all those Tick Tock sites as you're
2:07:00 scrolling through Tick Tock and there's
2:07:02 like some guy like leaning into your
2:07:03 face going you're going Mario
2:07:08 they're worse than the crypto Bros
2:07:11 um
2:07:11 and and and
2:07:15 to be clear how they're making money
2:07:18 is by getting you to pay them money so
2:07:21 that they tell you how to make money
2:07:23 with chat EBT they're not making money
2:07:25 with chat GPT because if they were they
2:07:27 wouldn't be screaming in your face about
2:07:29 making money with chat gbt
2:07:32 um we're breaking all the codes we are
2:07:34 breaking all the codes it's crazy in
2:07:36 here isn't it jrc just nuts
2:07:39 um
2:07:40 here's my philosophy on how you make
2:07:42 money with chatgpt because it's there's
2:07:44 a flipping answer ah Make It Rain oh
2:07:47 yeah
2:07:48 there's a flipping answer to these these
2:07:51 guys right
2:07:54 the flipping answer is [ __ ] them don't
2:07:56 listen to them that's not where the
2:07:58 money is
2:07:59 um
2:08:00 but but the real answer is this
2:08:04 because Chachi PT is so sexy because AI
2:08:07 is sexy
2:08:09 it's very easy to get seduced into
2:08:12 thinking it's something more than a tool
2:08:15 it's not it's not a strategy AI is not a
2:08:18 strategy chechi BT is not a strategy
2:08:22 is a tool just like Photoshop is a tool
2:08:26 just like Microsoft Word is a tool just
2:08:29 like a hammer is a tool it's like a
2:08:31 shovel is a tool
2:08:32 so when someone says you can make money
2:08:34 with cheche EBT it's the equivalent of
2:08:36 saying you can make money with a shovel
2:08:39 okay
2:08:40 why would you say that right if someone
2:08:43 said it about a shovel you'd be like
2:08:45 okay that's [ __ ] weird well how do
2:08:47 you make money with a shovel I don't
2:08:48 know you find out who needs holes Doug
2:08:53 and if you can figure out the way to use
2:08:55 your shovel in a way that digs their
2:08:56 holes in a way they're willing to pay
2:08:58 you for it you can make money with a
2:09:00 shovel
2:09:01 exact same thing with chat GPT with AI
2:09:05 with mid-journey with all these things
2:09:08 so the place to start if you want to
2:09:10 make money with chat gbt is start with
2:09:12 you
2:09:14 what are you good at what are you
2:09:16 passionate about what do you know
2:09:19 um in the world is a pain point that you
2:09:22 could probably solve if you had the
2:09:25 skills to do it or if you had the tools
2:09:26 to do it oh well I know how to throw
2:09:31 really good parties awesome
2:09:35 um could you throw really good parties
2:09:37 that people would be willing to pay for
2:09:39 yeah like okay then go to chatgpt and
2:09:42 say hey I'm really good at throwing
2:09:44 parties and I'd love to figure out how I
2:09:46 could make money throwing parties it'll
2:09:47 give you 40 [ __ ] ideas
2:09:50 and then of the 40 ideas two of them
2:09:52 will be really interesting to you and
2:09:54 you're like oh hey chat jbt these two
2:09:56 are really interesting to me can you
2:09:58 tell me more about them
2:10:01 how'd they come oh that one's really
2:10:03 good and before you know it this idea of
2:10:05 the hyper Focus tunnel
2:10:07 you can have a whisper of an idea for a
2:10:10 business and within five minutes have
2:10:14 have it articulated to the point that
2:10:17 that you could say now give me 10 names
2:10:20 for that business now give me 30 tag
2:10:22 lines for the name that I picked
2:10:24 now write me a website and it'll write
2:10:27 you all the website copy
2:10:29 within an hour you could have a fully
2:10:32 articulated business from a standing
2:10:34 start with nothing
2:10:36 that's how you make money with chat GPT
2:10:38 chat GPT is just the accelerant it's
2:10:42 just a tool that accelerates the [ __ ]
2:10:44 coming out of your brain it's a Humanity
2:10:46 amplifier that's what not the AI
2:10:50 learning lab is that's what
2:10:53 chat GPT is that's what clot is that's
2:10:56 what Bard is that's what mid-journey is
2:10:58 that's that's what the audio generation
2:11:01 tools are that's what the
2:11:03 um Runway ml text-to-video tools are
2:11:06 everyone's talking about these tools
2:11:08 like the tools are going to save you no
2:11:10 no no
2:11:10 You're Gonna Save You
2:11:13 Amplified by these tools Humanity
2:11:16 amplifiers
2:11:18 okay
2:11:21 but where do I start I'm not very
2:11:23 technical start here that's why I have
2:11:26 them on screen on every one of my lives
2:11:28 I've been doing this for months now
2:11:30 nightly here
2:11:35 why do I have these here start here
2:11:38 because we're early
2:11:39 still need a recommendation on a voice
2:11:42 to text transcription service okay
2:11:46 um
2:11:47 whisper is really good so if you go to
2:11:53 there's a couple of ways that you can do
2:11:55 it you can get the open AI app the the
2:11:58 chat GPT official app from openai
2:12:02 it does transcriptions you can basically
2:12:04 flip it into talk to me mode
2:12:08 and it will transcribe there you can
2:12:12 also this is a bit wonky there's not
2:12:14 actually a simple way to get to this
2:12:16 which I think is kind of shitty but I I
2:12:18 think this is ultimately going to be
2:12:20 built into chat GPT so I don't I don't
2:12:23 think you have to do it this way for
2:12:24 long but if you go to platform
2:12:27 dot openai.com playground
2:12:31 that brings you to the chat GPT
2:12:33 playground
2:12:36 hang on a sec sorry the voice is going
2:12:39 ha ha
2:12:42 all right we're down to 2 000 people now
2:12:44 we're down to normal levels no we're not
2:12:46 we're still like 10 times where we where
2:12:48 we normally are so so welcome everybody
2:12:50 all right
2:12:52 So within within the open AI playground
2:12:56 again the URL is platform.openai.com
2:13:01 slash playground
2:13:04 and then that brings you here and this
2:13:06 is where developers you know develop
2:13:08 their prompts and things like that so if
2:13:09 you're creating automations in zapier
2:13:11 for example this is where I come to
2:13:13 develop my prompts so over on the right
2:13:15 hand side here there's there's the mode
2:13:17 so if you if you move out of chat mode
2:13:20 and into complete mode
2:13:23 there's this little teeny tiny
2:13:26 microphone right here
2:13:29 this gives you access to whisper
2:13:32 so you click on that and now you can
2:13:34 upload a text a voice
2:13:37 audio file or or a video file with voice
2:13:40 in it or you can just record
2:13:42 and this will generate it to text
2:13:46 I've got this built into my company
2:13:48 storyvine where every video in the
2:13:51 system I can just push a button it'll
2:13:52 automatically do this for any video in
2:13:54 the system but it has to be done
2:13:56 programmatically but this is how you can
2:13:58 actually do it and then once it
2:13:59 generates it it'll put the text right in
2:14:02 here Whispers really good you know it's
2:14:05 not a hundred percent but it's you know
2:14:07 it's above 90. these Voice to Text
2:14:11 transcription tools historically have
2:14:13 sucked bad like I have an 11 and a half
2:14:16 year old company and I have not
2:14:19 I have not had transcription as part of
2:14:23 the the company Emily this is going to
2:14:25 save me so much time awesome
2:14:26 Emily
2:14:28 um
2:14:29 I didn't have transcriptions as part of
2:14:31 my my video platform for for 11 years
2:14:35 until whisper came along and it's it's
2:14:38 because it's actually good
2:14:40 um it still gets [ __ ] wrong but it's at
2:14:41 least acceptable at this point
2:14:45 all right how we doing time wise
2:14:48 um 10 51. it's getting late it's getting
2:14:51 late all right can it generate SRT files
2:14:55 for the videos
2:14:57 what's an SRT file is that the uh the
2:15:00 subtitle
2:15:02 hang on let me go find out what an SRT
2:15:04 file is I know I've owned a video
2:15:05 company I should know what that is but
2:15:08 um dot SRT for video file
2:15:15 SRT
2:15:17 oh yeah that's captioning
2:15:19 um
2:15:20 yes it can I don't know what the
2:15:23 parameters are hang on let me let me pop
2:15:26 a video in there and let's let's let's
2:15:27 go play with it
2:15:33 [Music]
2:15:38 file type is not supported well you're
2:15:41 an [ __ ]
2:15:43 um do I have an mp4 here
2:15:46 foreign
2:15:49 let's do this and this is the Dual
2:15:51 dialogue one this is the one with me
2:15:53 speaking both English oh no it's a mov I
2:15:55 need an mp4
2:15:59 oh these are too long
2:16:02 oh good lord
2:16:04 it's never easy as is that camera easy I
2:16:07 you know I try and I try and I try
2:16:10 people and then the file format's not
2:16:12 supported what am I supposed to do MP3
2:16:15 let me go find an MP3 I don't even know
2:16:17 if I have any MP3
2:16:19 things that are worth
2:16:23 going home
2:16:32 um
2:16:42 all right that's not good let me see uh
2:16:45 dot MP4
2:16:49 is an MPEG for movie
2:16:52 those are all those
2:16:57 and I guess we'll throw that up there
2:16:59 all right
2:17:03 transcribing thank you for the Roses by
2:17:07 any other name would smell as sweet
2:17:11 yeah voice to text and text-to-speech
2:17:13 went from useless to job killer in like
2:17:15 six months exactly slash home slash Dave
2:17:17 by The Way Slash home slash Dave 10 out
2:17:20 of 10 on the name Emily thank you for
2:17:22 the oh you're giving me all the toys a
2:17:25 lightning bolt and a pepper and a little
2:17:27 ice cream cone and a world and a what
2:17:30 the hell is that
2:17:31 it's a microphone or a fist or some sort
2:17:35 of phallic situation
2:17:37 um why did that not transcribe that
2:17:41 video
2:17:45 I don't know anyway I don't know what
2:17:48 the I don't know what the settings are
2:17:50 um for
2:17:52 um
2:17:55 for generating SRT files I've got within
2:17:59 storyvine I've got storyvine set up to
2:18:01 generate a vtt file which is which is
2:18:03 also a subtitling format so I know I
2:18:06 know whisper can do lots of different
2:18:08 sub subtitling formats it can pull down
2:18:11 time code when it when it does the the
2:18:14 transcription so it's pretty
2:18:16 sophisticated it also here's the other
2:18:18 just [ __ ] wild ass thing about
2:18:22 um
2:18:23 about these tools is like chat GPT all
2:18:27 these tools are brilliant at translation
2:18:29 Emily I'm so grateful for your help
2:18:31 you're welcome thank you so much I
2:18:33 really I really appreciate that thank
2:18:35 you Mr Pringles uh loves me a good hot
2:18:37 dog who doesn't like a hot dog
2:18:39 you know who doesn't like hot dogs Liars
2:18:43 you you know those vegans you you know
2:18:47 every insufferable vegan that you know
2:18:50 is sneaking a [ __ ] hot dog on a
2:18:53 Saturday night after they get home from
2:18:55 the bar you know leave the bar you know
2:18:58 that's going on
2:18:59 [Laughter]
2:19:01 bitter you betcha
2:19:04 um
2:19:05 but but these transcriptions will be
2:19:08 they we you can translate this stuff
2:19:10 into any language and now with hey Jin
2:19:12 you can do it with the video itself I
2:19:14 we're just honest to God I I feel like
2:19:17 we're in a post sci-fi reality it's I
2:19:21 like I think about
2:19:23 almost every sci-fi movie that I
2:19:26 experienced as a kid or as an adult and
2:19:29 with very few exceptions we're like
2:19:33 we're just checking off the Technologies
2:19:35 like oh yep we can do that now oh we can
2:19:38 do that now ah that'll probably be a
2:19:39 year
2:19:40 like like nothing's a decade away
2:19:42 anymore nothing ex time travel's the
2:19:45 only one where I like okay we're gonna
2:19:47 have to probably wait a couple of
2:19:48 decades like I'll probably I probably
2:19:50 won't be here for time travel but like
2:19:53 curing cancer and curing genetic
2:19:55 diseases and and like climate change and
2:20:00 energy production I mean just it's it's
2:20:02 gonna be [ __ ] insane
2:20:06 so anyway it's just crazy it looks like
2:20:08 we just got a whole new pile of people
2:20:10 we're down to 2 000 people we're back up
2:20:12 to seven thousand welcome everyone my
2:20:14 name is Kyle Shannon this is the AI
2:20:16 learning lab
2:20:17 um tonight's a bit of a trip because we
2:20:20 normally have anywhere from 50 to 150
2:20:22 people in here and now there's seven
2:20:24 thousand so hey
2:20:27 I'm Kyle this is the AI learning lab we
2:20:31 have cute little graphics
2:20:33 like see that and then we got
2:20:37 um I'm not a robot goodbye we got AI
2:20:39 learning lab on a little planet all
2:20:41 these images were created with a tool
2:20:43 called ideogram which if you've not
2:20:46 played with it
2:20:48 does text and if you've ever used
2:20:52 mid-journey and tried to get it to do
2:20:54 something like have words
2:20:56 it was really frustrating because it not
2:21:00 did words and and many people when when
2:21:02 I would show them this AI stuff they're
2:21:04 like oh make make a thing that says
2:21:07 blank you know whatever the saying was
2:21:09 that they wanted to do and you can't
2:21:10 you'd have to say well the tools are
2:21:12 kind of stupid right now
2:21:14 all right how do I turn my hard copy
2:21:16 notes into soft copy something I didn't
2:21:18 quite see what that was
2:21:20 um you can do that right now natively if
2:21:22 you've got an iPhone
2:21:24 um you go into
2:21:27 I think you go into notes and then you
2:21:30 say
2:21:31 um
2:21:32 uh scan a document hard copy notes into
2:21:36 soft copy notes
2:21:37 so
2:21:39 so there's there's a there's a couple of
2:21:41 ways you can do it there's there's a
2:21:43 notebook called Rocket books
2:21:47 which are really cool
2:21:49 um they're not you well they might be
2:21:51 using AI but but their their physical
2:21:53 notebooks
2:21:54 and they've basically got QR codes on
2:21:58 the page so you just take notes with a
2:22:00 normal pen they've they've got a version
2:22:02 of The Notebook you can you can take
2:22:05 notes with erasable pens and then you
2:22:07 can microwave it and it really it erases
2:22:09 all your notes if you microwave it it's
2:22:12 pretty cool but um
2:22:14 rocket book you can just write notes and
2:22:16 then you use the rocketbook app to take
2:22:18 pictures of the notes and it scans
2:22:20 everything that required a bit to
2:22:22 management too much management for me I
2:22:24 know within iOS you can go into the
2:22:26 notes app and you put it into document
2:22:28 mode and it'll OCR [ __ ] so you can just
2:22:30 take pictures and it'll do it
2:22:33 um Leonardo is a good one for images too
2:22:35 yeah Leonardo's good so let me show you
2:22:37 my my list of
2:22:39 um image generation tools that I think
2:22:42 are worth paying attention to because
2:22:45 Leonardo's on there
2:22:47 Leonardo doesn't do text ideogram does
2:22:51 text so I'll talk about these here so
2:22:53 here's the uh here's how this is going
2:22:56 to roll mid journey is the best for
2:22:59 probably another week
2:23:01 because Dolly Three is coming and I
2:23:03 think Dolly three is just going to
2:23:04 obliterate a bunch of these things
2:23:07 um there's a comment pin so I can't
2:23:09 start the poll
2:23:10 tonight
2:23:16 there you go okay
2:23:20 mid journey is the best
2:23:22 um the problem with mid-journey I don't
2:23:24 know if it's a problem but you have to
2:23:25 generate images from within Discord and
2:23:27 it's a real pain in the ass ideogram is
2:23:31 not as good an image generation but it
2:23:33 is good at text generation and and you
2:23:35 can make posters and you can make
2:23:37 t-shirts and you can do all sorts of
2:23:39 stuff not super precise but good
2:23:42 Leonardo is a web interface sitting on
2:23:45 top of stable diffusion so if you're
2:23:47 geeky you can use there's this interface
2:23:49 called automatic 1111 where you can do
2:23:52 all sorts of you know flips and switches
2:23:55 for stable diffusion if you don't feel
2:23:57 like installing your own software
2:23:59 manually use leonardo.ai you just go to
2:24:04 that website they give you 150 free
2:24:06 credits a day which translates to about
2:24:08 seven or eight images a day but you can
2:24:11 make seven or eight good images a day
2:24:13 with a nice interface
2:24:14 um hey Jen I'll come back to Adobe
2:24:16 Firefly and Photoshop generative fill
2:24:19 came out of beta last week
2:24:21 and which basically means you can now
2:24:24 use the images you generate there
2:24:26 commercially the other thing about Adobe
2:24:29 and and Photoshop generative filler
2:24:31 Firefly and generative fill
2:24:33 adobe's image models have not been
2:24:35 trained on copywritten images it's been
2:24:38 trained on
2:24:39 um
2:24:41 uh public domain and Adobe stock images
2:24:45 Adobe also has a plan for allowing
2:24:47 artists to opt out of being trained opt
2:24:50 into being trained and have a
2:24:51 compensation and attribution model so
2:24:55 um Adobe is going to make image
2:24:56 generation or RDR safe for Corporate
2:25:00 America
2:25:01 so those are the biggies hn is not an
2:25:04 image generation tool it's a text to
2:25:06 audio and then to video animation tool
2:25:11 but they have this thing called
2:25:14 um video translate that one of their
2:25:16 Engineers cooked up and it's [ __ ]
2:25:20 insane the backlog to get your videos
2:25:22 rendered right now are just stupidly
2:25:24 long
2:25:25 but check this out so this is a video of
2:25:30 me I I intercut my original English
2:25:33 video with the Italian translation that
2:25:36 that hey Jin did so we've heard from a
2:25:39 number of clients that modular content
2:25:41 is important to them getting more thank
2:25:42 you Ted with how we create content and
2:25:45 how we leverage content
2:25:47 modular important
2:25:53 behind me you see these little green
2:25:55 boxes um that's how we think about
2:25:57 stories at storyline we break stories
2:25:59 down into little atoms right little
2:26:01 modules
2:26:03 foreign
2:26:07 so just absolutely wild
2:26:10 um okay I just saw a comment from Jules
2:26:14 so Jules my daughter who's an artist
2:26:16 just chastised me for AI generated art
2:26:20 because it's stealing work okay
2:26:26 okay this is this is an incredibly
2:26:31 nuanced conversation I made it to
2:26:34 whisper but can't find where to go to
2:26:36 upload okay hang on a sec
2:26:40 you made it to whisper you mean did you
2:26:43 get into
2:26:45 um
2:26:46 chat GPT playground
2:26:49 go to plague or
2:26:53 platform.openai.com playground
2:26:56 and then
2:26:58 flip from chat mode
2:27:01 into complete mode so it'll default you
2:27:04 into chat mode it's not available there
2:27:06 flip into complete mode and then that
2:27:10 little microphone icon is the thing
2:27:11 you're looking for
2:27:14 um I want to go back to the image
2:27:15 generation thing
2:27:18 the way that chat GPT works
2:27:21 and the way that diffusion models work
2:27:25 they're not copying and pasting
2:27:29 they're not copying and pasting and in
2:27:31 fact I am so passionate about explaining
2:27:35 this I made slides
2:27:37 I got slides
2:27:39 AI isn't copying and pasting now I made
2:27:43 this image in mid-journey
2:27:47 this is
2:27:49 this is the misconception right now is
2:27:51 that these tools are
2:27:54 you know when you ask for for you know
2:27:58 some sort of photograph that it's going
2:28:00 into some photographer's work and
2:28:01 cutting the car out of it and pasting it
2:28:03 into some magic composite it's not what
2:28:06 it's doing
2:28:08 um now that said these models have been
2:28:12 trained on copywritten work and the
2:28:15 artists whose work it's been trained on
2:28:17 haven't gotten attribution haven't
2:28:19 gotten compensation
2:28:21 but
2:28:23 here's a metaphor I like to think of
2:28:27 if I have in my finger a grain of sugar
2:28:31 and you're making cake batter and I take
2:28:34 that grain of sugar and I toss it in
2:28:37 into the cake batter and then you bake
2:28:39 the cake
2:28:41 trying to figure out what part of an AI
2:28:43 generated image came from what specific
2:28:45 image that it was trained on is kind of
2:28:49 like asking to find that grain of sugar
2:28:51 in your baked cake it no longer exists
2:28:54 in that form
2:28:55 when they train
2:28:58 the GPT models the large language models
2:29:01 on these massive data sets they might
2:29:03 have been trained on the novel War and
2:29:06 Peace but the way the words get embedded
2:29:09 the way they get turned into Vector
2:29:11 space is just words in semantic clouds
2:29:14 War and Peace ceases to exist what it
2:29:17 pulled out of War and Peace were the
2:29:19 meaning of the words related to one
2:29:22 another and all of them get Amalgamated
2:29:24 into this massive mathematical set same
2:29:27 thing with the images they they get
2:29:30 denoised they get noised from an actual
2:29:34 image into a noisy image in fact I've
2:29:37 got a slide that shows this here okay
2:29:39 we're going to do this now screw it
2:29:41 because this is important
2:29:43 um I want you to you can you can tell
2:29:45 your doctor I saw some old guy on Tick
2:29:47 Tock and he told me that it's not
2:29:48 stealing
2:29:49 here's the deal it's
2:29:52 this is a very very very gray area
2:29:55 because
2:29:58 this is absolutely impacting people's
2:30:00 livelihoods and it's it's transforming
2:30:03 how art is created who creates art how
2:30:06 it's created but but it's a very nuanced
2:30:09 argument so
2:30:11 if you or your daughter actually
2:30:13 understand how these things are made you
2:30:15 can at least be more educated in
2:30:19 where the solutions might lie because I
2:30:21 think the entire industry right now is
2:30:24 focused on the wrong part of the
2:30:25 equation everyone's focused on the
2:30:27 output and how these images are used and
2:30:30 I think where to focus is on the
2:30:32 generation and and compensating people
2:30:35 who's who's work has been put into these
2:30:39 things and compensate them on the
2:30:40 generation of the image but anyway
2:30:42 here's how they work
2:30:43 so
2:30:46 so you've got your large language model
2:30:48 you've got your chat GPT the the way the
2:30:50 image models work is you've got your
2:30:52 large language model you've got it
2:30:53 paired with a diffusion model in this
2:30:55 case there's also Gans there's different
2:30:57 different kinds of Technologies but the
2:30:59 one I'm going to show you is a diffusion
2:31:00 model
2:31:02 the
2:31:04 so if you just notice the image that was
2:31:08 there it was that that sort of house on
2:31:11 a cliff you know illustration
2:31:15 that image when it gets put into the
2:31:18 training database it's labeled right so
2:31:20 someone went through and they labeled it
2:31:21 or they use computer vision to label it
2:31:23 they said there's grass there's trees
2:31:25 there's a house there's a cliff there's
2:31:27 mountains there's Sky there's so
2:31:29 whatever so in that image there were all
2:31:30 these words that they put in there right
2:31:32 and then they use a noise noising
2:31:35 algorithm to noiseify the image so the
2:31:39 image is no longer an image of that
2:31:41 thing it's this noiseified
2:31:44 mathematical
2:31:47 blob that has some words associated with
2:31:50 it
2:31:52 and then based on the prompt that you
2:31:54 put into a large language model the
2:31:57 large language model interprets those
2:31:59 words
2:32:00 and then it filters out
2:32:03 um
2:32:04 the the relevant vectors and and so I
2:32:07 explained this earlier with chat gbt but
2:32:09 this is this is what's called Vector
2:32:11 space
2:32:12 all of these dots are like uh they rep
2:32:15 they're they're
2:32:17 um sort of strings of numbers that
2:32:19 represent a point in mathematical space
2:32:23 this is a 2d representation of it real
2:32:26 Vector spaces are like a thousand
2:32:28 Dimensions it's it's literally
2:32:29 impossible to conceptualize but just
2:32:32 imagine so every one of those dots is
2:32:34 like a word but it's not a word it's a
2:32:36 number and the groupings are semantic
2:32:39 clusters like you might have you know
2:32:41 animals down here and within here's dogs
2:32:43 but you have like people that are bad in
2:32:46 relationships and you have dogs up here
2:32:48 but those are men that treat their women
2:32:49 bad right you know so so you have these
2:32:52 these different semantic clusters okay
2:32:54 now how do we get from this mess to a
2:32:57 picture
2:32:58 so again you you're combining this idea
2:33:01 of a large language model with this sort
2:33:04 of tagged diffusion
2:33:07 um image Library so if if our prompt is
2:33:11 Cat on a desk what the what the large
2:33:14 language model does is it basically says
2:33:17 okay in all these semantic clusters
2:33:19 here's all the the vectors that have
2:33:22 been tagged cat
2:33:25 and then here's all the vectors that
2:33:28 have been tag desk
2:33:30 and then within those the cat images are
2:33:33 all these random noise images and the
2:33:36 desk images are all these random noise
2:33:37 images it actually can't see what's in
2:33:40 the images it just knows that these were
2:33:42 tagged cat and desk and then it
2:33:44 mathematically averages those noisy
2:33:47 images into a final noisy image and then
2:33:50 using this the reverse of the algorithm
2:33:52 that made the noise it denoises the
2:33:55 image over 30 or 50 or 70 steps until
2:33:59 you have an image that's a cat on a desk
2:34:02 this is an original image
2:34:06 that came out of effectively the
2:34:08 imagination
2:34:10 of all of the [ __ ] it's been trained on
2:34:13 combined with your prompt
2:34:15 in this sort of mathematical magic that
2:34:19 is actually generating an original piece
2:34:21 of art now
2:34:23 where does it get really [ __ ] up and
2:34:26 and
2:34:27 um
2:34:28 and and fuzzy well if you type in I want
2:34:32 a cat on a desk illustrated by Greg
2:34:35 Rutkowski who's an illustrator who's
2:34:38 famously showed up in almost every
2:34:40 mid-journey in stable diffusion prompt
2:34:42 for for months
2:34:45 and all the images looked like his
2:34:47 images why because when you type in the
2:34:50 word Greg Rutkowski if he's tagged in
2:34:53 the language model it's going to pull
2:34:54 all of his images and even though it's
2:34:56 averaging them and it's creating a new
2:34:58 original image it's doing it in his
2:35:00 style because that was those were the
2:35:02 source
2:35:03 um sort of training elements that went
2:35:05 into generating that noisy image so your
2:35:10 daughter's not wrong to say this is
2:35:12 [ __ ] up
2:35:15 but it ain't stealing and it ain't
2:35:17 copying and pasting and so anyway I hope
2:35:19 that helps so uh we still have a lot of
2:35:23 people in here so uh welcome if you're
2:35:25 new here my name is Kyle Shannon this is
2:35:28 the AI learning lab uh I've been going
2:35:30 for a while tonight my voice is
2:35:32 officially
2:35:35 um checking out so I'm just gonna do a
2:35:37 few more minutes but you know another 10
2:35:39 minutes maybe
2:35:42 but follow me
2:35:44 um you know feel free to subscribe I do
2:35:46 these lives
2:35:49 every night I've been doing them every
2:35:50 night now for about three months four
2:35:52 months I don't know some amount of time
2:35:55 um I think it's really important
2:35:57 I'm here to kind of demystify this AI
2:35:59 stuff have fun
2:36:02 um it's been a little intense tonight
2:36:03 because normally there's like 50 to 100
2:36:07 maybe 150 people tonight there's been as
2:36:10 many as 17 000 right now we're at about
2:36:12 six and a half thousand so welcome
2:36:14 everybody
2:36:16 um
2:36:17 if you have questions pop them in the
2:36:19 comments below we've got some moderators
2:36:20 that are that are uh picking interesting
2:36:23 comments and putting up Up Normally I
2:36:25 try to answer all of the questions uh
2:36:27 with seven thousand people in here I
2:36:29 don't think I can do that but what are
2:36:32 you gonna do
2:36:33 all right let's answer some questions
2:36:35 let's get into it all right this uh are
2:36:39 you an AI evangelist
2:36:41 um
2:36:43 yeah I think so I I'm certainly I'm
2:36:46 certainly an AI Optimist
2:36:49 um
2:36:51 I know that the the the
2:36:54 popular thing to to say these days are
2:36:57 the robots are going to kill us I think
2:36:59 there is there is a an equal if not
2:37:02 higher chance that the robots are going
2:37:04 to elevate us
2:37:06 they're going to free us from doing the
2:37:07 the mundane work they're going to
2:37:09 amplify our capabilities and they're
2:37:11 going to allow us to do remarkable work
2:37:14 overall
2:37:16 um I also believe that the changes that
2:37:20 are coming are coming way faster than
2:37:22 anyone's prepared for
2:37:24 they are going to be dramatically
2:37:26 disruptive
2:37:28 uh and it's it's going to cause a lot of
2:37:30 pain
2:37:31 and it's gonna it's gonna transform
2:37:34 almost every job any job that is
2:37:37 computer mediated in any way
2:37:41 um is going to be affected by this stuff
2:37:43 and so this channel is really just about
2:37:45 opening up the conversation and saying
2:37:48 you know if you're sitting on the
2:37:50 sidelines
2:37:52 you don't have as much time as as you
2:37:54 think you do for for how fast this stuff
2:37:57 is coming
2:37:59 um
2:38:00 and so a big part of what I
2:38:04 kind of proselytize here if I if I
2:38:06 evangelize anything it's get curious now
2:38:11 like go there right now
2:38:14 chat.openai.com you don't need to do all
2:38:17 this other [ __ ] that I'm talking about
2:38:18 all these other Technologies
2:38:20 save that for another day if you haven't
2:38:23 tried chat GPT go there right now right
2:38:26 now like not
2:38:28 I'll get to it after the Live's over I'm
2:38:31 on Tick Tock I don't want to go there
2:38:33 just go there right now [ __ ] do it
2:38:35 [ __ ] do it do it do it
2:38:39 and just start playing
2:38:41 um I've got a a comment pin so I can't
2:38:43 start the poll
2:38:46 um
2:38:47 so
2:38:48 why do that because
2:38:52 this isn't a Google search it looks like
2:38:55 a Google search it's not a Google search
2:38:57 It's profoundly more powerful than that
2:38:59 these are like the most powerful
2:39:03 tools that have ever been been put in
2:39:06 the hands of mere mortals kind of shy of
2:39:09 of you know the nuclear bomb which was
2:39:12 not in put in the hands of mere mortals
2:39:14 this stuff is really really powerful and
2:39:16 we get to play with it
2:39:18 so it's in everybody's interest right
2:39:21 now to get curious about this and
2:39:23 understand what it's good at what it's
2:39:24 not good at the conversation I just had
2:39:27 about the mother who said her daughter
2:39:29 kind of slapped her down for you know
2:39:32 generated images of stealing and this
2:39:34 and that like
2:39:35 I'm glad her daughter is aware of it and
2:39:39 has an opinion on it
2:39:42 I want everyone to get more educated
2:39:43 about it because it's not as black and
2:39:46 white it's it's really easy to demonize
2:39:48 this [ __ ]
2:39:49 it's really easy to dismiss it oh it's
2:39:52 not perfect it hallucinates it lies
2:39:59 uh was that explanation ideogram that
2:40:02 explanation was just generally how
2:40:04 diffusion models work
2:40:05 um ideogram's one of them I don't know
2:40:07 how they're getting I assume what
2:40:09 they're doing is they're doing two
2:40:11 different
2:40:12 um
2:40:13 diffusion models where one is images and
2:40:16 one is text and they're somehow
2:40:18 combining those in the in the sort of
2:40:21 noisy algorithm part I don't I don't
2:40:23 honestly know how they're doing what
2:40:25 they're doing
2:40:26 but it's a diffusion model so it's it's
2:40:27 it's some fancy version of what I showed
2:40:30 I make most of my YouTube videos uh
2:40:33 scripting with chat gbt yep exactly so
2:40:37 just go to these things that just go to
2:40:39 this and start playing go to
2:40:40 prompts.chat that'll teach you how to
2:40:42 play with these things just get educated
2:40:43 about it get curious about it and just
2:40:45 keep coming back to this this is why
2:40:47 we're this is why this is here
2:40:50 um was that last explanation ideogram
2:40:53 that's what I just explained
2:40:57 a lot of tools he has shown tonight
2:40:59 weren't here two weeks ago exactly so
2:41:02 Joe so for the people that have been
2:41:04 here regularly what what you what you
2:41:07 don't know is if you just joined here is
2:41:09 that these lists that I show of the
2:41:12 different tools
2:41:13 these have been changing regularly this
2:41:16 one that just looks like a novel right
2:41:18 now this used to be like two or three
2:41:20 tools and now how many are there one two
2:41:22 three four five
2:41:24 uh
2:41:26 uh yeah five
2:41:30 and then what's missing from here is
2:41:32 Dolly three which was announced two days
2:41:35 ago and it's gonna show up
2:41:38 maybe in a week maybe over the weekend
2:41:40 who knows
2:41:42 and it's Gonna Knock mid-journey and
2:41:45 ideogram off this list I think
2:41:50 because mid-journey you have to go into
2:41:53 Discord you have to learn Discord
2:41:56 and then you have to learn command line
2:41:59 functions
2:42:00 you have to you have to type in command
2:42:02 lines to make images in mid-journey it
2:42:05 is the least visually
2:42:08 the the fact that the most sophisticated
2:42:11 image tool has to be typed in with
2:42:14 command lines is [ __ ] absurd
2:42:19 so what's going to happen with Dolly
2:42:21 three is the quality is going to
2:42:23 probably be as good as mid Journey or at
2:42:25 least close to it it's going to be able
2:42:27 to do text better than ideogram does and
2:42:30 you're going to be able to do it within
2:42:31 the context of a long chat GPT
2:42:34 conversation and generate multiple
2:42:36 images within that conversation where it
2:42:40 as it's generating the images it's aware
2:42:42 of the entire conversation so it can do
2:42:45 things like maintain a single character
2:42:47 across visual prompts which you've ever
2:42:49 if you've ever tried to do that in
2:42:51 mid-journey it's effectively impossible
2:42:53 even with control net with stable
2:42:55 diffusion it's effectively impossible
2:42:58 you can get there there's people that
2:43:00 have hacked it but you can't just say
2:43:01 this character's name is Bob now put Bob
2:43:05 in a spacesuit now put Bob on a fire
2:43:07 truck now put Bob at the hot dog stand
2:43:10 Dolly three will do that none of these
2:43:13 tools can do that right now like that's
2:43:15 gonna change in a week
2:43:18 been loving this stream I try and catch
2:43:20 it when I get home from my gigs thanks
2:43:22 everyone thank you Scott Curtis what
2:43:23 kind of gigs you a musician
2:43:25 that's super cool
2:43:28 um all right let me do we're down we're
2:43:31 down to a paltry
2:43:32 1600 people uh
2:43:36 which thank you all for hanging out for
2:43:38 a little bit
2:43:39 um now 1700 people it's going back up
2:43:42 um
2:43:43 normally these These are kind of
2:43:45 intimate little Gatherings and we have
2:43:47 lots of inside jokes and things like
2:43:49 that
2:43:50 um Kyle you missed Jules sent you a
2:43:53 beautiful gift a glowing jellyfish oh I
2:43:55 did miss it I don't know what I was
2:43:56 looking at but thank you Jules
2:43:58 um that's amazing thank you very much
2:44:01 um
2:44:02 listen I um
2:44:08 it you know it's funny it's funny some
2:44:10 nights I'll be in here there'll be like
2:44:11 30 people in here and the the people
2:44:13 that are in here regularly we've kind of
2:44:15 established a fun a fun relationship
2:44:18 um but it's been kind of an intimate
2:44:20 little thing and there's some nights I'm
2:44:22 like well you know is this really making
2:44:24 a difference and and then you know
2:44:25 tonight lots and lots of people are in
2:44:27 here and I think it might make a
2:44:28 difference but it's you know the nature
2:44:30 of what it has changed so so I hope you
2:44:33 come back
2:44:34 um I'm gonna go because my voice is is
2:44:37 absolutely fried
2:44:38 um is there an app for Dolly three okay
2:44:40 so Dolly three
2:44:42 what I'm going to do is give you go to
2:44:44 openai.com
2:44:48 slash d-a-l-l
2:44:54 Dash e Dash three
2:44:59 Dolly oh this is absolutely [ __ ]
2:45:02 mind-blowing what's coming so and it it
2:45:05 if you're if you haven't played with
2:45:07 these image generation models before
2:45:09 this won't be mind-blowing this will
2:45:11 just be like oh that looks cool
2:45:13 um
2:45:15 one of the toughest things to do with
2:45:18 these image generation tools is
2:45:20 get them to predictably do what you want
2:45:23 them to do
2:45:24 and so so Dolly three a couple of things
2:45:27 about this this is from open AI so this
2:45:30 is from the people that make chat gpt4
2:45:33 they currently have a thing called Dolly
2:45:35 too which if you want to play with it if
2:45:37 you go to bing.com here I'll show you go
2:45:40 to bing.com
2:45:41 and then you go to chat
2:45:43 that's chatgpt4 on Bing and I can say
2:45:47 um
2:45:48 make me an image of
2:45:51 a hedgehog driving a fire truck
2:45:57 and this will use dolly two
2:46:01 to make the image of a hedgehog driving
2:46:03 the fire truck Dolly 2 hasn't been
2:46:05 updated in I think over a year it's not
2:46:09 good but it's there and it's free and
2:46:13 look there's a hedgehog in a fire truck
2:46:14 now is it bad no that's not bad
2:46:18 they're actually pretty good
2:46:21 um
2:46:22 I wonder powered by Dolly I wonder
2:46:27 I don't think this is Dolly three yet
2:46:30 those does seem a little primitive so
2:46:31 anyway okay so let's go back to to Dolly
2:46:34 three
2:46:35 so dolly two was okay but Dolly three
2:46:38 it's going to be natively incorporated
2:46:41 into chat gpt4 so you're in chat TPT and
2:46:44 you're having a conversation maybe
2:46:45 you're writing a children's book or
2:46:47 maybe you're writing a graphic novel and
2:46:49 then you'll just be able to immediately
2:46:51 within that say go make me this kind of
2:46:53 image
2:46:54 and it will use gpt3 now check out what
2:46:57 this does this is a fun kind of New
2:46:59 Yorker kind of illustration but here's
2:47:02 what I want you to pay attention to and
2:47:03 if you've ever made
2:47:06 images you'll see how remarkable this is
2:47:08 the prompt is an illustration of an
2:47:11 avocado sitting in a therapist's chair
2:47:13 saying I just feel so empty inside so it
2:47:16 did the text perfectly
2:47:18 it's an avocado sitting in a therapist
2:47:20 chair with a pit sized hole in the
2:47:23 center so that looks like a pit but it's
2:47:25 actually a hole it's a pit sized hole
2:47:27 the therapist a spoon
2:47:31 scribbles notes
2:47:33 so all of the elements in that prompt
2:47:35 are represented in that image if you
2:47:37 have ever tried to get mid-journey to do
2:47:40 two concepts in the same prompt it's
2:47:43 essentially impossible it will always
2:47:45 prioritize one or over the other and you
2:47:48 will want to smash your face into your
2:47:51 computer or a brick wall it's maddening
2:47:54 so this is just gonna do that now the
2:47:58 prompt here a 2d animation of a folk
2:48:01 music band composed of anthropomorphic
2:48:04 autumn leaves all right perfect there's
2:48:07 anthropomorphic Autumn Leaves as if you
2:48:10 know as if I you know could ever
2:48:13 conceive of that now you can conceive of
2:48:15 it and just make it this one look at
2:48:17 this prompt I'll read it and then I'll
2:48:20 zoom in on the image because because
2:48:21 it's a remarkable image
2:48:23 um in front of a deep black backdrop a
2:48:28 figure of middle years
2:48:30 her Tongan skin rich and glowing is
2:48:34 captured mid-twirl her curly hair
2:48:36 flowing like a storm behind her her
2:48:38 attire resembles a whirlwind of marble
2:48:41 and Porcelain fragments illuminated by
2:48:44 The Gleam of scattered porcelain shards
2:48:47 creating a dream-like atmosphere the
2:48:50 dancer manages to appear fragmented yet
2:48:52 maintains a harmonious and fluid form
2:48:57 and look at that
2:49:03 fragments of porcelain swirling
2:49:06 porcelain her hair like a storm I mean
2:49:10 like remarkable remarkable okay
2:49:14 so let me close that down how do I get
2:49:16 out of here where does the there it is
2:49:21 um let's look at a couple other things
2:49:23 here this is really interesting so what
2:49:25 they did here was they took a long
2:49:27 prompt and they they broke it up and
2:49:29 sort of pointed to so part of the prompt
2:49:31 was sidewalks bustling with pedestrians
2:49:34 enjoying nightlife
2:49:35 bustling city under a full Moon Under
2:49:39 the shine of a full moon so you get
2:49:41 Shadows the full moon the shine
2:49:44 her prompt at the corner stall a young
2:49:47 woman with fiery red hair dressed in a
2:49:49 signature velvet cloak haggling with a
2:49:52 grumpy old vendor and then the vendor a
2:49:55 grumpy vendor a tall sophisticated man
2:49:57 he's taller than her wearing a sharp
2:50:00 suit yes
2:50:01 Sports a noteworthy mustache yes
2:50:05 and animatedly conversing on his
2:50:07 steampunk telephone yes
2:50:10 like holy crap
2:50:13 um so so the level of control here is
2:50:15 going to be amazing
2:50:17 um let's see that was Dolly too
2:50:21 that's Dolly three it's just on a
2:50:25 complete different level and then I'll
2:50:26 play this I'll play this video so one
2:50:29 thing that I'll note here before I play
2:50:31 it I'm not going to stop it
2:50:33 so
2:50:35 the the video generates they generate
2:50:37 this Hedgehog that you see right there
2:50:39 and then the first thing that happens is
2:50:42 the person doing it picks one of the
2:50:45 Hedgehogs and says my daughter said his
2:50:48 name was Larry
2:50:50 and then and then just watch what it
2:50:52 does
2:51:06 thank you
2:51:08 [Music]
2:51:24 [Music]
2:51:26 [Applause]
2:51:35 foreign
2:51:37 [Music]
2:51:47 let's see so Jules asks is Dolly three
2:51:50 like ideogram or your creation is
2:51:53 available and
2:51:54 uh to people in the public once
2:51:57 generated I don't think so because it's
2:51:59 going to be a plug-in within a chat GPT
2:52:01 chat you can probably share your chat
2:52:05 um and then someone else asked another
2:52:07 thing uh right below that uh what did
2:52:10 what did they ask
2:52:13 uh what tool are we using tonight this I
2:52:16 was just showing Dolly three anyway this
2:52:20 should just roll out so so the people
2:52:22 that are gonna get first access to Dolly
2:52:24 three are um chat GPT Plus members so
2:52:27 the 20 bucks a month will get you access
2:52:30 to this before anyone else so I I have
2:52:33 thought for a long time chat GPT is
2:52:36 worth the 20 bucks now with this it's
2:52:38 definitely worth the 20 bucks and with
2:52:40 code interpreter it's really really
2:52:42 really really worth it
2:52:44 um Emilio's wife appropriately made fun
2:52:47 of me for saying I was just gonna put in
2:52:49 another five minutes it's already 11 30.
2:52:52 all right yeah I gotta go I gotta go
2:52:54 people I know isn't that amazing Jules
2:52:56 it's amazing uh oh the the other
2:52:58 question was oh update on the Mandarin
2:53:00 video I I don't think we're going to be
2:53:02 anywhere close I think I actually have
2:53:03 to pay for
2:53:06 um
2:53:07 where was it
2:53:09 um window no uh
2:53:12 um
2:53:16 [Music]
2:53:18 let me just log in hey Jen
2:53:22 um
2:53:23 someone asked did did Dolly three talk
2:53:26 about
2:53:27 um how long those images took to render
2:53:29 no but they're they're not going to be I
2:53:32 don't think they're going to be that
2:53:33 long they wouldn't be releasing this if
2:53:35 they were super long
2:53:37 um oh I need uh
2:53:42 Labs Dot
2:53:45 hagen.com
2:53:50 I'm down to 78
2:53:53 511 in line so yeah we're not gonna my
2:53:57 voice is fried seize the question [ __ ]
2:54:00 shut up shut up I resemble that remark
2:54:02 from Emilio's wife
2:54:05 um all right I'm gonna put a little
2:54:06 picture up here a couple of things
2:54:10 if if uh
2:54:15 all right we'll do one more pull
2:54:19 um
2:54:22 all right I gotta go listen
2:54:23 this has been awesome this has been
2:54:26 absolutely surreal tonight
2:54:27 um thank you to I think it was James who
2:54:30 um
2:54:31 dumped
2:54:34 thousands of people into the uh chat
2:54:37 tonight it was a wild one indeed all the
2:54:39 Irregulars are like holy crap what
2:54:41 happened I'm like that too so thank you
2:54:44 all for supporting this you know what's
2:54:46 funny had we not set you all up as mods
2:54:50 however it was three weeks ago and sort
2:54:52 of figured out what we were doing here
2:54:54 this would have been total [ __ ] chaos
2:54:56 tonight so thank you for for all the
2:54:59 work you do thank you for showing up
2:55:00 here every night
2:55:02 um for those of you that are still
2:55:03 hanging out from the new influx of
2:55:05 people thank you come back
2:55:07 um follow me subscribe to the Channel
2:55:09 all that sort of happy horse [ __ ] um
2:55:11 this has been really awesome uh Nancy
2:55:13 good to see you and major fundraising we
2:55:15 had such a crowd we did we had a big one
2:55:18 um we'll talk about future of week uh
2:55:20 future of work I know that's something
2:55:22 that you wanted to talk about Ann at
2:55:23 some point we'll talk about that this
2:55:25 over the weekend if you're around uh
2:55:27 just make sure you pop a comment in
2:55:29 there it was a little hard tonight with
2:55:30 17 000 people in here to to get your get
2:55:34 your comment in
2:55:36 um but uh forever hooked great good
2:55:39 night Vicky Emilio's wife Jules thank
2:55:41 you
2:55:42 um was that Stevo you're awesome Ted
2:55:44 thank you very much historic it was a
2:55:47 historic night wasn't it it was [ __ ]
2:55:49 crazy caprono thank you peace out Scott
2:55:53 Curtis Silver Fox all right all you all
2:55:56 are awesome have a fantastic rest of
2:55:58 your Friday night
2:56:00 um the Suzuka Grand Prix should be a
2:56:02 good one Max verstappen didn't win a
2:56:05 race last week he's he he got to be
2:56:08 pissed he's gonna dominate that race
2:56:12 I'll see y'all see y'all tomorrow