aiLL 10/23/22 - ChatGPT Levels Up with DALL-E 3 - The AI Renaissance Is Here

Description

Chapters

Transcript

0:25 [Music]
0:29 do all
0:31 right it's Monday night I have Monday
0:36 attitude finally on time for ones yeah
0:39 you're right you're the first one in or
0:40 second one in will um oh look all your
0:44 regulars piling in Clint here Robert
0:47 Rossy what's shaking Be not Afraid
0:50 what's happening what's happening people
0:52 what's happening good people good people
0:54 of the AI learning
0:58 lab anyone else have a Monday that was
1:00 just like it was not bad but you're just
1:02 in a shitty
1:03 [Laughter]
1:08 mood H this will be better this is
1:11 always the highlight of my day so hello
1:13 family hey Brandon with
1:18 shaking Robert Rossy welcome welcome
1:22 welcome oh got a little sting in the
1:27 eye allergies I'm guessing
1:32 allergies this is a Sherry D image very
1:36 cool image there then I think that one
1:40 who did that
1:41 one forget who did that
1:46 one that one oh those are cool who's
1:49 doing
1:50 that experimenting with text a few came
1:53 out without edits a
1:56 few look these are from uh
2:01 from
2:02 Marduk these are pretty
2:07 slick
2:10 nice is that spelled right yeah that's
2:13 spelled right it's actually a pretty
2:16 slick
2:21 logo wasn't nice
2:23 dude or D that I don't don't know if
2:26 you're a man or a woman I
2:29 assume that one's pretty cool too isn't
2:36 it that's
2:38 solid with the robot and the graduation
2:42 cap that's pretty
2:44 good I'm telling you man like when
2:48 when when people with creative minds and
2:52 not even necessarily creative skills
2:54 with creative minds with with creative
2:56 eyes get their hands on these tools and
2:59 really start digging deep on them we're
3:01 going to see such a cool Renaissance in
3:05 stuff so anyway welcome everyone my name
3:08 is Kyle Shannon this is the AI learning
3:10 lab as evidenced by the little sign
3:14 here and um we talk about this
3:17 generative AI stuff in here we kind of
3:19 go nuts it's like a little it's like a
3:22 little support group for for weirdos and
3:25 Irregulars who have decided that yeah
3:29 this is the thing and and we got to
3:30 learn it and it doesn't matter if you're
3:33 young or if you're
3:35 old or if you're Gen X or if you're a
3:38 [ __ ] Boomer if you're a
3:40 millennial so long as you're curious
3:42 about this stuff that's what this
3:43 channel is about it's about
3:45 exploring um I completely redid my
3:49 recommendation slides when I when I
3:51 start these things off so if you don't
3:53 know where to begin with this stuff get
3:55 ready to do some screenshotting because
3:58 I think I now have seven slides of
4:01 recommendations in different categories
4:04 and this is not like recommendations
4:06 like
4:09 um well a they're not
4:11 paid as someone said last night are you
4:14 getting paid for this
4:16 advertisement for chat
4:19 GPT I was like no I mean you know maybe
4:23 if you give me a gift I get paid I'll
4:25 get a
4:25 nickel
4:28 um but um they're there because they're
4:31 thing they're they're tools
4:35 that as I've been digging into this
4:38 stuff are ones that I feel like are
4:40 worth your
4:41 time because there are thousands of
4:44 tools in fact if you want to see
4:46 thousands of them go to futur
4:53 pedia.com
4:57 we get started with that since I got
5:00 since I got new stuff to show why don't
5:02 we go show it well why don't you show it
5:04 Kyle because all you're doing now is
5:06 talking maybe you shouldn't talk so much
5:09 you should do a little bit more do you
5:10 ever think about that I have thought
5:13 about that so here's the little
5:16 convention that I came up with first of
5:18 all I simplified them so they're easier
5:19 to read because they were a little they
5:21 were a little rough to read I also um
5:26 you may think that those are
5:28 fingerprints on my screen they're not or
5:30 not I actually um took a diamond uh
5:34 etcher and I etched I I hand created
5:37 fingerprints in the glass of the screen
5:39 so they're there permanently so I hope
5:41 you enjoy those all right here's the
5:43 convention on each of these slides I
5:46 highlighted one in yellow which is the
5:48 one that I'm particularly enamored of of
5:51 that page okay so these are the big four
5:54 these are the important llms if you
5:56 don't know that term llm it's it stands
5:59 for large language model the most
6:00 important one is chat GPT by far worth
6:03 the 20 bucks to to pay for chat GPT
6:07 plus Claude is a close second because
6:10 you can it's got a really big what
6:11 called context window you can put in
6:13 really big prompts Bard's about to get
6:15 really good and Bing just pisses me off
6:17 because it's safety guard rails are too
6:19 much and then prompt. chat um if you
6:22 don't if you've never played with chat
6:23 jpt before go to prompt. chat I'll kind
6:25 of teach you how to talk to these things
6:27 and give you some examples of some
6:28 prompts you can play with okay so that's
6:31 that's the big four this this is the
6:33 adults table prompt. chat will teach you
6:35 these are the big four if if you only
6:38 play with one play with chat GPT okay
6:40 got it now just get ready just get ready
6:42 to screenshot because I'm not going to
6:44 take a ton of time on these other
6:46 important llms Pi is the is the one of
6:49 the of the Kids Table um of the other
6:52 llms that I think is the one that's
6:54 that's the most interesting right now
6:56 and it's it's actually the IOS app of Pi
7:00 um and why it's interesting is it does
7:03 this so you launch it oh wait well you
7:07 can't really see it ah well anyway it
7:09 doesn't matter I'm going to put it into
7:11 talkie
7:14 mode I'm trying to figure out which llms
7:17 I should pay attention
7:22 [Music]
7:25 to ah the big old AI Showdown here are a
7:28 few llms large language models worth
7:31 paying attention to open AI gpt3
7:34 generative pre-train Transformer 3 it's
7:37 not right
7:39 so it's it's conversational Ai and and
7:43 of of all the conversational AIS there's
7:45 only a handful of them that one that
7:47 one's actually really interesting
7:48 perplexity is good for research po lets
7:51 you play with lots of different models
7:53 and it's got a bot maker you can make a
7:55 bot you can use other people's Bots and
7:57 then uh llama 2 you can play with that
8:00 at po you can play with that at labs.
8:02 perplexity so that's that
8:04 one image creation Dolly 3 is the new
8:07 darling so dolly3 now is available to
8:10 any chat GPT plus user I'm pretty sure
8:13 it's rolled out wide to all plus and
8:16 Enterprise users now it's remarkable
8:19 being able to create images in the
8:21 context of a chat in a you know in a
8:23 conversation is remarkable mid journey
8:26 is a close second because the the
8:28 quality is good the problem with mid
8:30 journey is you have to make images in
8:33 Discord and if you're not Discord
8:35 friendly then it's not friendly Leonardo
8:39 is a really powerful tool B built on top
8:41 of stable diffusion the Adobe tools are
8:43 are quite good and especially if you
8:45 work in a company they're important and
8:47 then prompt book the thing at the bottom
8:49 there that's going to teach you how to
8:50 talk to these things okay um video
8:54 creation this one got bigger I added
8:57 three things to this so um hey gen the
9:01 the reason that one's there why why I
9:03 think
9:04 Heen is is the highlighted one is
9:06 because you can do video translations
9:10 into I think it's 13 different languages
9:13 where it synthesizes your voice
9:15 translates the language and resyncs your
9:18 lips to match the new language it's
9:20 astounding um Runway ml Moon Valley and
9:23 pabs are all text to video creation
9:27 tools and then hey Jen and
9:32 D- uh um are voice synthesists and like
9:36 Avatar makers and Visa is one it's kind
9:40 of interesting I put it here because
9:42 it's you know sometimes you need to make
9:44 just a quick video what visa does is you
9:47 put in a topic it writes the script
9:51 narrates it with a voice puts background
9:53 music in and then chose chooses a bunch
9:55 of stock video like you know from stock
9:58 houses and it cuts together you a
10:01 complete video with with subtitles and
10:04 voice over and music and it does it in a
10:07 few minutes um it's not generating video
10:09 but it's stitching together video all
10:11 right so that's that one so screenshot
10:13 that go
10:18 go um here's people to follow Ethan
10:22 mullik he's he's the one of all these I
10:24 it's a tossup between Ethan mullik and
10:26 David Shapiro and and quite frankly all
10:28 four of these people are worth following
10:30 Ethan mik David Shapiro on YouTube Paul
10:32 rer at the marketing AI Institute and
10:34 here on Tik Tok ra Rachel Woods at the
10:36 AI
10:37 exchange um and then uh the directories
10:40 future pedia and there's an AI for that
10:42 are both that's where you can go find
10:44 lots of
10:45 tools here's music or or audio um 11
10:49 Labs will let you clone your voice
10:51 refusion and sound draw are two
10:54 different approaches to music I don't
10:56 have a fourth one there right now I
10:58 haven't been digging deep in the music
10:59 thing so if anyone has any thoughts on
11:01 that I'll I'll look um those are good
11:04 there um 3D worlds and objects blockade
11:07 La labs and wonder Dynamics if you've
11:09 not seen Wonder Dynamics it's amazing
11:13 you you basically take action of
11:16 yourself as a human in a scene or or
11:18 whatever video of a person in a scene
11:20 and you select that person and pick
11:22 which robot you want replaced into the
11:24 scene and it goes off for an hour or two
11:27 and does its processing and then you
11:29 have like a fully replaced the human is
11:32 fully replaced with the robot using the
11:34 human's motion it's pretty amazing um
11:37 all of these things the the the uh the
11:40 the 3D stuff right now is very very new
11:42 and and not great and then here are um
11:46 the AI learning lab resources so we have
11:48 a YouTube channel which has got um
11:50 recordings of these lives at AI learning
11:53 lab dtt the AI Salon I'll refer to the
11:57 Discord uh tonight likely um it's a
12:00 community that I created uh last
12:02 December we're coming up on our oneyear
12:04 anniversary and we meet every other
12:06 Tuesday tomorrow night's the next
12:07 meeting so go sign up there sign up for
12:10 the Meetup for the meeting and then
12:12 there's a link also to the Discord go
12:14 join the Discord and then storyvine is
12:16 my company and then tinyurl.com kyes AI
12:20 tools is a Google sheet that um that
12:25 Thelonius uh put together every time I
12:28 mentioned the tool for like the past 4
12:30 months he's been adding it to this uh
12:32 this spreadsheet so it's just a big long
12:34 list of tools um based on what I've been
12:37 babbling about so anyway how cool is
12:39 that and then one final thing if you
12:42 subscribe to this channel if you follow
12:45 it um make sure you're not following the
12:47 one that has two L's at the beginning of
12:49 Learning the the one with the stutter AI
12:53 learning lab is a scammer so unfollow
12:56 that person report them and put comments
12:59 in the videos there saying that they're
13:02 an
13:02 [ __ ] the real AI learning lab knows
13:05 how to spell and doesn't stutter okay so
13:09 so there's that bitter you betcha all
13:13 right and then what we do here is pop
13:14 any comments you have in in the
13:16 questions or comments in the in the
13:18 comments below and we will Dive Right In
13:22 we'll Dive Right In Kyle what are we
13:23 going to do tonight what are we going to
13:25 do I don't know I don't know I don't
13:28 know what we're going to do
13:31 somebody took this picture outside of my
13:33 house that's my uh that's my van where I
13:35 pick up
13:38 unsuspecting AI curious people I got
13:41 free candy will you help me find my
13:43 puppy
13:46 [Laughter]
13:52 please all right how long till the
13:57 Meetup which Meetup the Meetup for the
13:59 this or the Meetup for the salon the
14:01 salon Meetup is tomorrow
14:04 night 5: to 7:00 p.m. tomorrow the
14:06 Meetup here the the uh I think I think
14:10 we'll do the
14:11 Meetup in Vegas when I when I move the
14:14 uh when I move the AI learning lab from
14:17 Tik Tock to the sphere then I think
14:19 that's when we do the Meetup because it
14:21 it I mean the problem with the sphere is
14:23 it only seats 25,000 people and we're
14:25 going to have way more than
14:27 that I mean you too can barely fill the
14:30 place up right so like I mean they're
14:33 they're old tired right so they can't
14:35 fill it up so what I'm probably going to
14:38 have to do is I'm probably going to have
14:39 to do a residency at the sphere and then
14:42 we'll do the AI learning lab I don't
14:44 know maybe two times a day for like for
14:48 like a month to just be able to get
14:50 enough people in there I don't know I
14:52 mean it Fame is hard people you know
14:54 people tell you they're like oh but Kyle
14:56 it must be so nice to be famous I'm like
14:58 hey it's rough
15:00 you don't you don't know I walk down the
15:02 street and absolutely no one recognizes
15:04 me and I'm like whoa back off
15:07 people do they think I'm weird
15:10 sure all right carrying on all right
15:14 that's metic well well looks like Kyle's
15:17 lost his ever living mind isn't that
15:19 terrific wonder what could have happened
15:21 there oh well maybe who could that have
15:23 been who could have been assisting in
15:24 his mental state oh maybe it was Satan
15:27 well well well if if you want to do a
15:30 tag me in the
15:32 Discord maybe wait if you want to do tag
15:36 me in the Discord maybe like an hour
15:40 before can we sick a hacker on that
15:43 impostor feel free Mary
15:45 Mary I know nothing I know nothing I do
15:49 not know what you speak of the hackers
15:52 going after the scammer account I do not
15:54 know what I do know is that when I went
15:57 to Tik Tok to say hey I got scammers
16:00 that stole all my videos that stole my
16:02 name they're trying to scam my people
16:04 and Tik Tok said we don't care that's so
16:08 cute what you do but if you say a bad
16:10 word we'll kick you off here for seven
16:12 days okay that's our little agreement I
16:14 hope you're good with that
16:18 okay yeah so you
16:20 know would love to see vials in action
16:24 tonight if possible please
16:27 vials V
16:29 vs veils vials what you forever hooks
16:34 what are you talking
16:35 about vials I got Tootsie
16:40 Rolls can you tell it's close to
16:45 Halloween the kids are going to get here
16:47 on Halloween I'm be like yeah we got
16:49 nothing left I don't know what happened
16:50 to
16:52 it forever hooked I'm happy to show
16:56 vials in action if I knew what vials in
16:58 action actually he was I got no idea
17:00 what you're talking about oh it looks
17:03 like the it looks like the Irregulars
17:04 are on their their heart Fest so so one
17:06 of the things we do in here sometimes is
17:08 we'll just go nuts on the on the the
17:10 liking of the channel hey do me a favor
17:13 since we're just getting this thing spun
17:14 up um if you're willing to share the
17:17 live with your friends um for two
17:19 reasons one is I'd like to get more
17:21 people in here and you know educate more
17:24 people but also um they'll think you're
17:26 a
17:27 weirdo
17:30 and it'll be fun they'll be like Jane
17:33 why why did you send me that old man
17:35 talking about AI that was disturbing and
17:39 you'll be like I
17:42 know don't you like messing with people
17:45 I sure
17:46 do how can I dress up like chat GPT for
17:50 Halloween that's a rough one I don't
17:52 think it's quite mimy enough yet
17:57 um
18:01 I mean the only thing you could maybe do
18:03 is you could maybe do like
18:05 um you know it would obviously it would
18:08 be a big board and if you if you did the
18:10 do thing where it draws images like
18:13 where it makes images you could have
18:15 like a crazy Halloween prompt like make
18:17 me a scary image for Halloween and it
18:19 would have them down there but and then
18:21 you would have typos on it and you would
18:23 have mangled hands make sure you get
18:25 mangled hands yeah that that would be a
18:27 good one is have it be like a chat GPT
18:30 window with a funny prompt and then four
18:33 images that are all [ __ ] up the hands
18:34 are like this and someone's got three
18:37 eyes and the Hat's on weird
18:39 and the horse has six
18:42 legs and what's cool about doll is
18:45 because it's good at prompt coherence
18:46 like you could give it [ __ ] up things
18:49 to do and it would put them in the in
18:50 the image that might be
18:52 fun I don't know I don't know you could
18:55 go as you can make money with j
18:59 you could go as a you can make money
19:01 with chbt
19:03 button are there privacy concerns with
19:06 the data you input yes ping 357 there
19:09 are data concerns with the privacy
19:11 concerns with the data you input you
19:13 should not put in uh sensitive or
19:15 private information into these things
19:18 unless um you're running a like an open
19:21 source large language model locally or
19:24 unless you have uh chat GPT Enterprise
19:27 which I think at this point you have to
19:29 be spending 10 or 15 grand a month with
19:31 them for that to be a reality so that's
19:33 probably not the case
19:36 um yeah so I would I would just you know
19:38 anonymize and de depersonalize any data
19:41 you're going to put in there um you
19:43 don't you don't like Here's my thought
19:46 like I don't give a [ __ ] if like the
19:49 only thing I'm concerned about is I
19:51 don't want client data getting out like
19:53 but I do a lot of my work um I send it
19:57 to chat GPT um you can also um rather
20:00 than using chat GPT itself you can use
20:03 the GPT playground so platform.
20:07 open.com
20:09 playground um that's kind of the
20:10 development environment for um for
20:13 that's what chpt is built on that's the
20:15 tools that they use to build it and
20:17 that's what a lot of developers use um
20:19 that's still being sent to open AI so
20:22 they're their Agreements are a little
20:24 shitty when it comes to privacy they
20:26 basically reserve the right to use your
20:27 [ __ ] um but I feel like the developer
20:30 side it's less egregious than chat GPT
20:32 itself so if you're if you're using chat
20:35 GPT then that's that's that's a
20:37 something the other thing that I I need
20:39 to research that I I don't know is if
20:42 you're using Advanced Data analysis I'm
20:45 pretty sure most of that data analysis
20:47 is happening inside a container in your
20:50 chat session so I don't think when
20:53 you're doing Advanced Data analysis it's
20:55 sending data back to open AI but you
20:58 know what it is when it's when it
20:59 actually does the the large language
21:02 model query yeah never mind just assume
21:05 that your data is Visa visla oh Visa oh
21:08 you want me to show visla okay yeah
21:10 happy to show that vile it's like can
21:14 you show vile I'm like I could show vile
21:16 for forever hooked you want to see
21:20 Visa all right so
21:22 here
21:25 us and we will go why do I not oh I quit
21:29 it because it was it tanked my memory
21:31 like it always
21:44 does oh hello Visa oh hey Mark
21:52 hi now I don't think I have a oh no I
21:56 don't think you need it I don't have a
21:58 uh
22:00 a paid account here so I think the only
22:03 thing that it gets [ __ ] up is if I
22:07 want to generate a non-watermarked
22:11 video
22:12 then the free video stock footage is
22:16 really
22:18 shitty okay so create video with AI next
22:23 upload cut extractor more screen record
22:27 okay so there's a bunch of [ __ ] you can
22:29 do there but let's do a creative video
22:32 so basically this is this is just like
22:34 chat
22:37 GPT it says choose what you'd like to
22:39 start with an idea a
22:42 text so if you have a script you can put
22:44 your own script in you can start with a
22:46 Blog just give it a
22:50 URL you could start with audio you could
22:52 start with images and video so so this
22:55 is cool this is actually pretty slick so
22:56 so what it's giving you is
22:58 um bunch of different options to
23:02 start and then let's see let's go over
23:05 to chat
23:07 J you can
23:14 make all right so we're going to go to
23:17 chat
23:18 gp4 I'm G to say
23:27 um
23:33 you know what I'm not going to I'm not
23:34 going to have it right let's let's let
23:36 visa do
23:38 it so what I'm going to do I'm GNA do
23:41 something
23:42 stupid Please make a video about a llama
23:48 Hedgehog and turtle
23:50 sanctuary in Southern New
23:55 Hampshire so I'm just gonna do something
23:58 kind of silly and and ridiculous and
24:00 that's basically it so so what it's
24:02 doing now it's writing a
24:04 script it's synthesizing the voice over
24:09 and then it's
24:11 selecting stock video
24:13 footage and then and then I'm pretty
24:15 sure you can go in like I haven't played
24:16 with this a ton but I'm pretty sure that
24:18 you can go in and you can swap out like
24:20 if it didn't get the video segments
24:22 right you can swap out other videos you
24:24 know you could you know if you wanted to
24:25 you could generate video and put it in
24:27 there but I I I mean I think this is
24:29 kind of a slick
24:32 um a slick tool in that sense that
24:35 because the the text to video tools
24:37 right now are kind of
24:39 janky um and and you know the quality is
24:43 you have to be you actually have to be a
24:45 very good Storyteller and quite a clever
24:48 prompter to get things like mid journey
24:51 and pabs and Moon Valley to give you um
24:55 video that's not just weird right right
24:59 so what this is doing is it's saying hey
25:01 let's just what what we'll we'll do all
25:03 the same basic stuff we'll write a
25:05 script we'll synthesize the voice but
25:08 rather than generating the video we'll
25:09 just go find real footage which is
25:12 pretty slick
25:14 right and so if you wanted to do I mean
25:16 this is the kind of thing where you
25:17 could very easily automate a Tik Tok
25:20 channel for example right so here's our
25:22 script so let's see welcome to the Llama
25:25 Hedgehog and turtle sanctuary in
25:27 Southern New Hampshire
25:29 here these adorable animals find Safe
25:31 Haven to thrive and be cared for okay so
25:34 let's go let's go watch our dumb
25:37 video we'll make it big can I make it
25:40 not this
25:42 big well
25:48 whatever welcome to the Llama Hedgehog
25:51 and turtle sanctuary in Southern New
25:54 Hampshire here these adorable animals
25:56 find a safe haven to thrive and be cared
25:59 for our Sanctuary is nestled in the
26:01 picturesque Landscapes of Southern New
26:03 Hampshire llamas roam freely their
26:06 gentle nature adding a touch of
26:08 Tranquility to the
26:09 sanctuary hedgehogs find comfort in
26:12 their cozy burrow pretty sure that's an
26:13 iguana or chameleon Turtles bask in the
26:17 Sun and serenity of their natural
26:20 habitat our dedicated team of caretakers
26:23 ensures the well-being of these unique
26:25 creatures visitors can learn about the
26:28 fascinating lives of llamas hedgehogs
26:30 and turtles come and experience the joy
26:33 of being surrounded by these incredible
26:36 animals visit our sanctuary in Southern
26:38 New Hampshire and embark on a memorable
26:42 Journey there you go so that's it now
26:46 you know like all these tools is it a
26:48 great script no but you could you could
26:51 create your own script or you could
26:52 start with a blog or you could start
26:54 with whatever and then uh and then
26:56 you're off to the racist so that's vla
26:58 Visa could be a miscellaneous M wait
27:01 could be a miscellaneous mistake images
27:03 Halloween special like a horse with more
27:05 legs win yeah exactly yeah exactly the
27:08 Halloween costume the chat GPT Halloween
27:12 costume free AI
27:14 inside come on you want some free
27:17 candy help me find my
27:22 puppy
27:26 uh uh uh uh
27:29 making it weird one prompt at a time you
27:32 can make money with Chach
27:34 B all right all right good evening good
27:39 evening good day to you sir good day to
27:41 you sir good day it's a fine
27:46 day all right what questions do we have
27:54 here just a reminder that AI ain't free
27:59 yeah what Pate
28:01 said I don't know what you mean
28:04 Pate some of it's free oh all those
28:07 video things are not free like video and
28:10 audio and animation all of those seem to
28:12 cost money chat GPT image
28:16 generation and the large language models
28:18 seem to be mostly free but you know the
28:22 plus version of chat GPT is 20 bucks POS
28:25 20
28:26 bucks um what's it called claws 20 bucks
28:30 right so those are those the uh the
28:32 animation ones like the ha genen the
28:34 ones that do the video the voice
28:36 translation and video translation those
28:38 are those are expensive it's like two
28:39 bucks a minute to do translations but
28:43 the quality of it's quite good and it's
28:47 uh considering the alternative like
28:50 hiring an actor or you know doing voice
28:53 dubbing or whatever you know exp
28:56 expensive
28:59 can Visa make a video from a comic
29:03 illustration I'm kind of doubting it
29:06 jeie I don't know Visa very well at
29:09 all um Runway ml can pabs can and uh
29:14 Moon Valley can where you start with an
29:16 image so you could go to Mid Journey or
29:18 Del create a comic IM image and then use
29:21 that as your first frame for video
29:25 generation free the AI free the AI we
29:29 need free AI if you can code some
29:33 resources are free Kyle yeah well
29:35 exactly so so Zippy to pin heads got it
29:38 right there here's the deal so what I
29:40 talk about mostly on this channel is
29:42 [ __ ] that assumes that you don't want to
29:45 be a coder and you don't want to be a
29:47 mathematician and you don't want to F
29:49 spin up a instance of Google collab and
29:52 figure out what a collab notebook is and
29:54 Learn Python if you do want to do that
29:56 [ __ ] the the the world is your oyster
29:59 right now there are so many projects
30:01 from stable diffusion to the open source
30:04 large language models to Lang chain
30:06 there's like all sorts of cool [ __ ] out
30:07 there to play with if you're geeky so if
30:10 you're
30:10 geeky like my my actual real challenge
30:16 with with all this stuff is it's
30:19 impossible enough to keep up with the
30:22 commercial tools and and the major
30:24 commercial tools like I'm not even
30:27 delving in [ __ ] like Jasper and copy Ai
30:30 and all the really tactical tools
30:32 because there's like 5,000 of them so
30:36 and and even with that my list right so
30:40 what have I got here I've got
30:43 four
30:46 48 12 and six is
30:50 18 and then those are people to
30:54 follow that's 21
30:59 23 so so my my kind of short list of
31:04 [ __ ] that I like I play with at least
31:06 enough to keep keep on track of it is 23
31:11 tools and even if I cut that in half
31:13 even if I say okay I don't go to I don't
31:15 go to most of them a lot of the
31:19 time that's still a lot chat GPT alone
31:23 is like seven programs right there's
31:26 custom instructions
31:28 there's default GPT 4 there's GPT 3.5
31:31 which is fast
31:34 there's
31:35 Vision there's browse with Bing there's
31:41 Advanced Data analysis there's Dolly
31:45 there's plugins and plugins there are a
31:47 thousand
31:49 plugins and then there's the
31:52 conversational chat GPT so there's nine
31:55 separate
31:56 modes of chat GPT Each of which is like
32:01 you could spend months just on each of
32:03 those
32:05 modes that's one
32:07 tool and when when um Google comes out
32:12 with Gemini this this multimodal when
32:15 when they come out to the competitor to
32:17 gp4 then we're going to have two major
32:19 play two major platforms that have
32:21 multiple Lanes of things to explore
32:24 that's two
32:25 tools uh I'm could someone just repin
32:30 pate's comment because I started reading
32:32 it and and missed
32:34 it you're the best help a non-programmer
32:36 can get oh good Zippy to pin head yeah
32:39 so so that thank you for that I mean I
32:42 figure the pro here's the deal the
32:45 programmers are going to take care of
32:46 themselves and even if you're like like
32:48 you know nominally talented as a as a
32:53 engineering thinker you're going to be
32:55 fine cuz chat GPT can help you figure
32:57 fure out the rest but this channel
32:59 really is about focusing on the
33:02 inflection point you know that started
33:04 on November 30th zy the pin head uh with
33:08 what free Power and Chip oh yeah yeah
33:10 exactly no this stuff costs money these
33:13 uh these these things are burning up a
33:17 lot of GPU cycles and TPU cycles and now
33:22 I I saw Pate that that now they're like
33:25 going back to gpus
33:28 like one of the articles I saw and and
33:30 now we've got um open
33:33 AI Microsoft
33:36 Amazon and maybe was it Google so
33:40 there's a bunch of them out there right
33:41 now that are like [ __ ] it we're going to
33:42 make our own
33:44 chips in the future there will be pirate
33:47 AI fishing data oh will there be pirate
33:51 AI will AI security be a thing yeah [ __ ]
33:54 yes yeah so so so
33:58 um
34:01 yes
34:04 um so because
34:06 meta op sourced their
34:10 um their llama
34:13 models the entire developer Community is
34:16 just building [ __ ] you know that's a
34:18 completely unrestricted large language
34:20 model right that's the thing that Sam Al
34:22 this is why Elon Musk is pissed off at
34:24 Sam Altman because Sam Alman said hey
34:26 these these tools are powerful enough I
34:28 don't think we should just release them
34:29 into the [ __ ] wild and Elon Musk was
34:32 like it's called open AI for a reason so
34:35 he's off so Elon Musk is off creating x.
34:38 a so he can just open source [ __ ] meta
34:41 just open source
34:42 [ __ ] so so what it means is any kind of
34:46 safety controls you might have had with
34:48 you know a Google or a fa or or a uh you
34:51 know an open AI or a Microsoft or an
34:54 apple um that's kind of out the window
34:56 with these with these model model being
34:57 released to the world so what that means
34:59 is both the black hats and the white
35:01 hats have access to the tools so yes the
35:04 the answer and and in fact you know it
35:06 it it doesn't even need to be super
35:08 nefarious like they're you know the
35:10 hackers are coming for you go to the
35:13 Apple App Store and type in chat GPT and
35:16 see how many apps Apple is allowing to
35:18 be sold that have effectively copied the
35:22 open aai Logo the chat GPT logo and have
35:25 a name that is deceptively similar there
35:27 there are like five or six apps that are
35:31 absolute chat GPT clones that are not
35:33 from open AI that people are charging
35:35 money for and and ripping people
35:39 off Kyle can you show the tiny URL for
35:42 the spreadsheet thonus made I will
35:44 absolutely do
35:49 that it
35:54 is that tiny url.com kyes AI
36:00 tools where is the starting point for
36:03 those wanting to utilize AI to create
36:06 art success Mentor let's get you up to
36:08 speed
36:10 so there are um lots of different places
36:14 where you can create AI art
36:17 [Laughter]
36:19 but there's only one place in my opinion
36:22 you should start right now you should go
36:25 there chat. openai
36:28 and pay the 20 bucks to them and then
36:30 I'll show
36:32 you how this works well actually you
36:34 know what I'm going to show you I'm
36:36 going to go over to Discord and show you
36:38 what it's like to make a mid Journey
36:41 image because it's a [ __ ]
36:44 travesty not so the the quality of mid
36:47 Journey images is absolutely amazing how
36:50 you make them is the
36:54 travesty so so this is Discord if you if
36:57 you're if you haven't been in Discord
36:58 before it's it's basically a community
37:00 platform for gamers that has kind of
37:03 escaped the gamer community and is now
37:05 being used like the nft Bros completely
37:08 adopted this and so most nft projects
37:12 are Discord heavy and a lot of the AI
37:15 stuff is happening on Discord because I
37:17 assume a fair amount of the programming
37:19 talent that bled out of the nft world
37:22 bled into the AI world and so that's why
37:25 shit's happening over here but anyway
37:27 these are are uh different communities
37:30 and then like that one's the mid Journey
37:31 Community but you can also do mid
37:33 Journey inside other communities so I'm
37:35 going to go in the AI Salon which I
37:37 talked about before and then I'm going
37:38 to go into the mid Journey these are the
37:40 different channels or conversations
37:42 within the AI salon so I'm going to go
37:44 to the mid Journey Channel and then
37:47 here's here's our channel so you can
37:49 just scroll and you know here's pictures
37:51 that people are making in mid
37:54 journey and if you want to make images
37:56 in mid Journey what do you do do you do
37:59 you click on something no you learn the
38:02 command line interface you type in slash
38:05 imagine and then it it pops up a little
38:08 thing here and you click on that and
38:10 then it says
38:11 prompt what do you put in there well you
38:14 put in um let's see let's go
38:18 [Music]
38:19 with
38:26 um
38:27 Majestic swimming Sea
38:30 Turtle deep in the ocean with sun rays
38:34 piercing the waves
38:38 above
38:42 period streamers of
38:45 kelp dance in the
38:49 background and then I'm going to
38:52 type-- AR space 16 colon n
38:58 why cuz cuz if I want a wide image I
39:02 have to type in D- AR which stands for
39:05 aspect ratio and then 16 by 9 which is
39:08 an HD um aspect ratio so there you have
39:12 it uh and then I hit return and it'll go
39:14 off and do its thing so you can get much
39:17 much much more sophisticated with these
39:19 but sorry could not complete the job an
39:21 error has occurred what did I
39:24 do it might have just been a it just
39:27 might have been a glitch in The Matrix
39:28 people sometimes the Matrix she she not
39:31 happy with us you you know the Matrix
39:33 she thinks that we're just we're just
39:35 not not good people not good
39:39 people are we going are we
39:44 going waiting to start waiting to
39:48 start Pilgrim turkeys we could have done
39:50 Pilgrim turkeys now look at what's
39:52 happening here so this is called These
39:53 are called diffusion models so notice
39:55 how they start out as these blurry
39:57 blobs and then they get a little more
40:00 detailed and they're now they're a
40:01 little more Turtle likee but they're
40:02 still kind of blurry and they're getting
40:04 a little better and a little better
40:06 that's the way these things work is they
40:08 kind of make this dollar cost averages
40:11 blurry image that's sort of Amalgamated
40:14 from data from probably thousands of
40:17 images and then it denoises it over a
40:19 series series of of images so these are
40:23 actually pretty good um again mid
40:26 Journey's quity is quite good so when
40:28 you make images there's four Images 1 2
40:30 3 four so these are your four options
40:32 and then you have these u and v numbers
40:35 upscale and variations so if I'm like um
40:39 okay I like the I like the light in this
40:41 one
40:43 and I think I think two and two and
40:46 three are interesting so let's do
40:48 variations of two and three so I'll do a
40:50 variation of two variation of three and
40:53 now I'm going to get four more images
40:55 that are variations of each of those so
40:57 I'll get eight more images and then of
40:59 those I can upscale one of
41:02 those and so here we are watching the
41:05 diffusion model diffus not free but
41:08 there's a free version you can do inside
41:14 Bing these are taking a long time come
41:17 on Mid Journey speed it
41:20 up all right there we go
41:23 beautiful um all right I like four so
41:25 we'll do upscale
41:29 for and that should go pretty quick boom
41:33 there's our upscaled image and now I can
41:35 upscale that mid Journey just added 2X
41:37 and 4x upscaling so I could actually
41:40 probably upscale that to the point that
41:42 I could uh print that on a decent Siz
41:45 poster all right so there's that lovely
41:48 lovely lovely lovely so that's mid
41:51 Journey now compare that experience to
41:58 Chad
41:59 j so here we are at Chad GPT um if
42:02 you've not been here before here's your
42:05 uh chat history and then here's your
42:08 chat GPT 3.5 chat GPT 4 and if I roll
42:11 over four there's all these different
42:14 modes the last one here Dolly 3 is the
42:16 one we're going to go
42:19 with and I'm going to say
42:25 um
42:27 I know Siri Siri up go
42:33 away
42:37 Siri Siri gets
42:39 jealous there we
42:44 go I'm
42:49 the owner of a turtle
42:54 sanctuary and I'm trying to come come up
42:57 with some beautiful
42:59 imagery of sea turtles swimming in kelp
43:02 forests period I'd like the posters to
43:06 be landscape in
43:10 orientation and photographic in nature
43:13 and stunningly
43:15 beautiful before you make any images can
43:18 you come up with four different
43:23 concepts for what the images and
43:27 composition might look like and then
43:28 I'll choose the best one and we can make
43:33 images so so what
43:36 happens now that we've got so so what
43:38 I'm looking at here Dolly 3 is an image
43:40 generation tool that lives inside chat
43:45 GPT so now I'm just asking chat GPT to
43:48 give me Concepts right so what do we've
43:51 got here dappled sunlight dive that's
43:54 kind of what we did before I like that
43:56 one the ethereal encounter captures the
43:58 moment where two sea turtles come face
44:00 to face amidst the kelp that's nice
44:03 Turtles ey view experience the kelp
44:06 forest from the perspective of a turtle
44:08 and then the kelp canopy
44:10 Cruise picture a young sea turtle
44:12 swimming close to the ocean floor
44:14 looking
44:15 up so let's do let's combine two and
44:20 four so I
44:22 say okay great I like two and four a
44:28 lot and I'd like you to combine those
44:31 Concepts and create four landscape
44:34 oriented images that bring them to
44:38 life period be sure they ultra realistic
44:43 photographic
44:47 style and now what's happening is chat
44:51 GPT is off writing prompts for me so
44:54 this is a glimpse into the future so so
44:56 so there's been a lot of talk over the
44:58 past year that you know prompt
44:59 engineering is going to be the next big
45:01 thing you got to be a prompt engineer
45:02 you got to be a prompt
45:03 engineer we're very quickly
45:06 moving into a world where these tools
45:09 get increasingly sophisticated where the
45:11 prompting is done for us and there's
45:14 there's one of the guys that I say to
45:15 follow is this guy Ethan mik and he he
45:17 talks about some um one of one of his
45:20 recent articles he talks about research
45:22 that was done where the chat GPT
45:26 prompting itself kick the [ __ ] out of
45:29 humans prompting chat
45:31 GPT so so there you have it all right so
45:35 there's one of our little turtles
45:36 there's our two turtles meeting amongst
45:38 the
45:40 kelp and those are okay they're not
45:43 super realistic I'm going to
45:49 say I'd
45:53 like to see the Turtles faces much more
45:56 more closely so can you zoom in on
46:00 those and also these feel more like
46:03 illustrations than photo photographs can
46:06 you
46:07 [Music]
46:08 include film types and lenses and
46:11 cameras in your prompt to make the
46:15 images seem more like
46:18 photographs so I can just basically
46:21 creative direct right so I can tell chat
46:24 gbt no I don't like those make more just
46:27 do more so it's it's just like the stuff
46:29 we do with with with text prompts with
46:31 chat
46:33 gbt but now we're doing it with images
46:35 so I would start here now this is this
46:38 is only available if you pay the 20
46:39 bucks a month if you can't afford the 20
46:41 bucks a month you can also go to
46:43 bing.com create that's also dolly3 it's
46:47 a free dolly3 engine from Microsoft
46:51 bing.com
46:53 create but let's see how our little
46:55 turtle images come out
47:02 [Laughter]
47:06 here oh these are
47:08 cute
47:12 nice still looks a little
47:15 illustrate and this is again
47:18 where Dolly's not quite up to Snuff with
47:22 like that's probably the one I'd go with
47:23 that one's really beautiful I mean you
47:25 could imagine like a headline down here
47:28 oh and in fact let's do that
47:31 um let's let's call it
47:38 the I want help coming up with a name
47:43 for my turtle
47:44 Sanctuary
47:48 that captures the essence of the images
47:51 you've created give me 10
47:55 ideas that's
47:58 fine of course sunlet shall Heaven kelp
48:03 canopy Cove ethereal Turtle Oasis that's
48:08 it great I like image number one that
48:11 you created which is the closeup of the
48:13 single turtle and I'd like you to give
48:16 me four variations of
48:18 that with the name of the
48:23 sanctuary ethereal Turtle a basis in
48:27 elegant font below the turtle's
48:33 face and we'll see we'll see how well it
48:36 does with the oh say 4K photo realistic
48:40 okay that's that's good spfd I wish I'd
48:43 have seen that before I just sent it off
48:45 but what are you going to do what are
48:47 you GNA do it's the AI learning lab
48:48 where we're just as janky as the tools
48:51 that we talk about oh yeah hey
48:55 Kool-Aid
49:03 is mid Journey paid now yes mid Journey
49:05 doesn't even give you
49:07 free sample images
49:11 anymore I think they they they got to
49:13 like 200 million in revenue and uh as
49:17 Pate said it's costing them a lot of
49:19 money so they're starting to charge for
49:25 it ethereal
49:27 tur
49:29 Oasis that turtle looks
49:32 angry let's see did we get any ethereal
49:36 Turtle Oasis this is good that's
49:40 real
49:42 a all turtle
49:46 sass
49:49 Sanctuary yeah that's not quite right
49:51 that's not quite right so they're not
49:52 quite right but gives you the idea right
49:56 really really quickly you can get there
49:59 um I use Leonardo which do you like
50:02 better for images so right now for me I
50:06 like Leonardo a lot it's on my list
50:09 of of tools to um it's like on my top
50:14 four for images where is
50:16 it like you know it's there but for me
50:20 right now Dolly 3 is it if I need to do
50:23 something that's that's more
50:25 sophisticated creatively I'll use mid
50:27 Journey um I don't think I'm going to
50:29 get rid of my mid Journey subscription
50:31 just yet but it's pretty close
50:34 um Dolly 3 that that ability to create
50:39 images in the context of a conversation
50:42 for
50:42 [Music]
50:43 me is way more how I like working if
50:47 you're super technical
50:49 just as a
50:52 thinker and and you like being able to
50:54 Flip Flip all the switches and try
50:57 different models and and do all the
51:00 control that Leonardo gives you I would
51:02 say Leonardo's the way to go right so I
51:05 feel like stable diffusion is a tool for
51:08 tinkerers and Leonardo is just a front
51:11 end to that um Mid journey is a tool for
51:15 um
51:17 artists Firefly from Adobe is a tool for
51:22 people creating Graphics in
51:24 companies and then and then uh Dolly 3
51:28 feels to me like the tool for the rest
51:30 of us right the tool for just anyone
51:32 that wants to be able
51:34 to create images easily quickly without
51:38 having to learn prompting without having
51:39 to geek out so so I think again that's
51:43 why why I have all these tools like so
51:47 Kyle which one's the best one even
51:49 though I've got this one highlighted
51:51 there's absolutely reasons I use this
51:53 there's absolutely reasons I use this I
51:55 don't use Leonardo as much anymore but I
51:58 do have models trained there so if I
52:00 want to if I want to do images that are
52:02 based on my image or photography I've
52:04 taken that's kind of the only game in
52:07 town um you know I could I can do it a
52:10 little bit at Mid Journey but not great
52:12 but like all four of these have value
52:14 for me
52:16 um of these Pi perplexity and po have
52:20 value for me um chat GPT Claude Bard and
52:24 Bing have value for me but like all
52:26 at different times right basically what
52:28 happens is I'll use my kind of lead tool
52:31 Until It Breaks and then I'll try a
52:33 different tool and then if that breaks
52:36 then I'll go try a different one if that
52:37 breaks then I'll go try a different one
52:38 if all of those break then I just cuss
52:40 and I throw my computer at the wall this
52:43 is why I have a really pricey laptop
52:48 [Laughter]
52:51 budget all
52:54 right
52:56 but Kyle how am I going to learn all
52:57 this this seems like too much this is so
53:00 much work here's the deal this is fun
53:03 this is fun
53:05 because you don't have to be a
53:07 technologist to learn this stuff
53:10 and that's that's why I'm showing the
53:14 tools I'm showing these These are the
53:15 tools where you don't have to learn a
53:16 lot of
53:17 technological stuff to use
53:20 them and they're kind of like video
53:22 games they they go really quick like you
53:25 type in a thing and out comes a magical
53:28 output you know you know what I'm saying
53:31 do you know what I'm saying do you no
53:34 what I'm saying thank you Pearl Pearl I
53:36 couldn't read the rest of that I love
53:38 that image Sherry D I don't know if
53:40 you're in here but that's super cool all
53:42 right Broody what is your name Broody
53:45 Broody henna that's a cool name Broody
53:48 henna what AI could I use to create a 3D
53:50 image of a
53:54 gutter
53:56 a three Ronnie Ray thank you so much for
53:59 the Roses those contributions really
54:01 help I do appreciate it a lot thank you
54:04 um a 3D image of a gutter like a rain
54:07 gutter on a house or like like
54:11 a a street gutter I assume you mean like
54:14 a house gutter house gutter Street
54:17 Gutter and and when you say a 3D image
54:20 do you want it to just look like a 3D
54:21 rendering or do you want it to be a 3D
54:24 object because if you want it to be a 3D
54:27 object I would say it would probably be
54:29 easier to not use AI because you could
54:31 just extrude a shape yes rain gutter so
54:35 so what I would do like like again where
54:37 I would start is I would start with chat
54:39 GPT so I'll just go new
54:44 chat I'm starting a new business for
54:48 cleaning out rain
54:50 gutters
54:54 and I want want to create a 3D
54:59 rendered
55:01 image of a ring
55:04 gutter looking at it slightly from
55:07 above so that you can see the side
55:11 profile and inside the gutter as well as
55:13 a little bit of the roof and a little
55:15 bit of the lawn
55:17 below I don't know something like that I
55:20 mean you you know you know what you need
55:22 to to do and and there's there's ways
55:25 that you could prompt this that it would
55:26 it would be better than this if you need
55:28 it to be an actual 3D object that's it's
55:31 kind of a different
55:33 exercise and that's that's another
55:35 import important point that I think I
55:37 think is comes up here is as exciting as
55:40 these AI tools are right now because
55:43 they're not fully baked
55:48 yet and they're just tools they're not
55:51 perfect tools they're kind
55:54 of they're kind of
55:57 janky right like okay here's the
56:00 Innovative gutter that actually doesn't
56:02 run along the roof but it comes from
56:04 inside the roof yes yes we at under roof
56:08 gutter installation
56:12 services here's another one coming
56:14 straight out of the thing okay these are
56:16 all really
56:17 [Laughter]
56:21 bad the perspective on these is all
56:24 horrible can you please just show me a
56:25 rain gutter and have it be clean and not
56:29 full of leaves and have there be rain
56:33 dancing along the gutter and running out
56:38 of away from the house as if it were a
56:41 happy happy
56:42 [Laughter]
56:47 day can chat gbt do a schematic drawing
56:51 you give it a photo of an object with
56:52 its measurements no so Barbara b60 we're
56:56 we're so you're entering an area that
57:00 that is likely being worked on by a lot
57:03 of the 3D companies um but but the the
57:06 way these diffusion models generate
57:10 images the last thing they are is
57:12 precise oh my God these are the worst
57:15 gutters in the history of gutter
57:23 making oh my god wow wow wow yeah no
57:31 this is the opposite of what a gutter
57:33 supposed to
57:38 do oh my God are you sure vision
57:42 couldn't do
57:44 it uh couldn't do what oh oh that's a
57:50 good idea Joker okay so here's a
57:52 possible way to get around this okay
57:54 here's what we're going to
57:57 do we're going to go
58:04 to uh
58:07 rain
58:11 gutter
58:21 images why do they always shoot them
58:23 from the bottom look there's our leaky
58:26 one we've got that
58:28 [Laughter]
58:31 one let me see rain
58:34 gutter from
58:41 above boy it just doesn't like seeing
58:43 him from above all right whatever see
58:46 him from
58:54 Below
58:56 all right so let's
59:00 [Music]
59:02 grab get rid of the from above because
59:05 we had better pictures before let's just
59:08 go grab us a grab us a nice
59:16 picture it's The
59:21 Simpsons all right we'll grab this one
59:23 fine all all right there's our there's
59:26 our
59:27 thing so save image
59:32 as closeup of vinyl gutters okay so what
59:36 we're this will be fun I have an idea I
59:38 have an idea that was a good idea Joker
59:40 okay so here's what we're going to do
59:42 we're going to use two different modes
59:44 now the fact that these are two
59:45 different modes in open AI feels like
59:48 one of the dumbest things ever to me but
59:50 it's just because the Tool's new I have
59:52 a feeling these things are going to
59:53 merge at some point
59:55 so if you go to default mode of
59:59 gp4 there's this little icon that shows
1:00:01 up and that's GPT Vision so if you click
1:00:04 on that I can now go upload that gutter
1:00:10 image and now I can
1:00:13 say please
1:00:18 describe this photograph in as much
1:00:21 detail as possible period I want to be
1:00:25 able to use your description as a prompt
1:00:28 to generate an
1:00:31 image based on
1:00:34 this
1:00:37 okay and so it can it can see that image
1:00:41 and it can understand it so let's see
1:00:43 what it comes up with photograph
1:00:45 showcases the corner of a building's
1:00:47 rooftop with a sunny day the roof is
1:00:49 covered with a layer of what appears to
1:00:50 be white paint or coning and terminates
1:00:52 at a distinct linear Edge attached to
1:00:55 the roof Edge is a white gutter designed
1:00:57 to channel rainwater away the gutter
1:00:59 curves downward in a graceful Arc
1:01:02 transitioning from a horizontal
1:01:04 orientation on the roof to a vertical
1:01:05 drop down underneath the
1:01:08 roof okay perfect so now I'm going to
1:01:12 say rewrite this as a single image
1:01:15 prompt but I want to transform the
1:01:18 gutter from white vinyl
1:01:21 to stunning handcrafted copper
1:01:28 so now I'll have it take that
1:01:30 description rewrite it a sunlet corner
1:01:32 of a
1:01:33 building layer of white paint under a
1:01:36 contrasting Sky a stunning handcrafted
1:01:39 copper gutter gracefully arcs from the
1:01:41 roof's Edge transitioning from
1:01:43 horizontal to Vertical beneath a Light
1:01:46 Beige wooden
1:01:49 Eve okay so there's our there's our
1:01:51 prompt so now I'm going to take that
1:01:53 prompt and now I'm going to flip modes
1:01:56 again this is where I think it's stupid
1:01:59 now I have to flip over to image
1:02:01 generation mode which is do
1:02:03 3 and
1:02:06 say please make this
1:02:10 image pop that prompt in there and then
1:02:13 say uh make
1:02:16 it
1:02:18 16 by n
1:02:22 landscape orientation
1:02:24 [Music]
1:02:27 and I misspelled that but chat GPT
1:02:29 doesn't care so hopefully this will be
1:02:31 there the schematic I left the get
1:02:33 gutter oh 20 minutes
1:02:37 ago
1:02:41 um well you
1:02:45 could I suppose you could I I don't know
1:02:48 Joker if if it's going to I just don't
1:02:51 think that the the generation things
1:02:53 like I I suppose what you could do is
1:02:54 you could take a schematic with
1:02:59 measurements oh that would be
1:03:01 interesting to do if you take a
1:03:03 schematic like a three up schematic of
1:03:07 top left and right
1:03:10 views and then ask it to to describe the
1:03:14 3D object that's kind of a cool
1:03:20 idea so there's our copper gutters those
1:03:23 are pretty
1:03:29 this is the new Innovative gutter that
1:03:30 just stops it doesn't actually go to the
1:03:33 [Laughter]
1:03:39 ground all right um that was that that's
1:03:42 actually an interesting experiment Joker
1:03:45 so let me I don't know if I can find a
1:03:47 I've got to find a
1:03:50 uh
1:03:53 um
1:03:57 blueprint of an
1:04:03 object
1:04:05 with side front and top views with
1:04:14 measurements projections that's what
1:04:16 they're called I knew they were called
1:04:23 something
1:04:34 let me see
1:04:40 blueprint
1:04:43 projection
1:04:45 of
1:04:49 a
1:04:51 can with measurement
1:05:11 I'm trying to get one that just has the
1:05:14 here we
1:05:18 go all right so
1:05:23 here's
1:05:29 yeah this is okay this is a hole here
1:05:31 okay
1:05:33 so I'm going to copy this image wait I'm
1:05:37 going to save it I'm going to save this
1:05:39 image as 3D
1:05:44 projection to the desktop and we'll go
1:05:47 back to chat
1:05:49 JPY and we're going to go to default
1:05:52 mode we're going to upload this
1:06:00 and then I'm going to
1:06:04 say Please interpret this drawing and
1:06:12 describe the isometric view of the 3D
1:06:16 object in an interesting visual scene so
1:06:20 that I can use it as a
1:06:22 prompt for a r ing of the object that's
1:06:26 described in the
1:06:33 drawing um I have
1:06:35 a voice note of my mother talking who
1:06:39 passed on can I recreate her voice
1:06:42 reading our song yes you can that one
1:06:47 nurse so hang on I'll show you how to do
1:06:49 that let's see if we can get this thing
1:06:51 to
1:06:53 understand imagine a 3D object with the
1:06:56 following
1:06:59 features so it absolutely understands
1:07:01 the drawing the visual scene
1:07:06 description
1:07:12 okay so this should be it should be a
1:07:16 L-shaped board with with holes in in
1:07:19 either side so let's let's let's just
1:07:21 copy all this certainly isms isometric
1:07:28 description and then we're going to hop
1:07:30 over to a new chat we're going to go
1:07:32 Dolly
1:07:36 3 and then I'm going to
1:07:39 go let's
1:07:42 see based on this
1:07:47 description of a 3D
1:07:53 object
1:08:00 please give me for
1:08:05 renderings of what the real world object
1:08:12 might look like in wood
1:08:17 metal
1:08:23 concrete
1:08:26 and paper mâché
1:08:28 [Laughter]
1:08:31 so so we'll see if it we'll see if it
1:08:33 gets anywhere close to this my my strong
1:08:37 suspicion is it won't get anywhere close
1:08:39 but it's an interesting experiment I
1:08:41 mean the fact that it could look at
1:08:43 that that projection drawing of that
1:08:46 object and describe it that accurately
1:08:50 is is pretty
1:08:53 mindblowing
1:09:02 um that one nurse so the the thing about
1:09:05 the voice note if the voice note has a
1:09:08 bunch of background noise in it it won't
1:09:10 be as successful if it's decently just
1:09:12 her voice isolated it's it's probably
1:09:17 it's probably decently doable so I'll
1:09:18 show you I'll show you the tool I'll
1:09:21 show you the tool that's the best right
1:09:22 now oops
1:09:31 my studio is falling
1:09:39 [Music]
1:09:44 apart um actually it's not
1:09:47 horribly like it's it's it actually
1:09:50 describe it it
1:09:52 actually captured it better than I
1:09:54 thought thought it would right it had a
1:09:56 right angle and a hole in the vertical
1:09:58 and a hole in the horizontal surface now
1:10:01 is it anywhere in the neighborhood of
1:10:03 what that drawing was no but you know I
1:10:07 would assume if you messed around with
1:10:08 this long enough you could probably get
1:10:10 it to uh you could probably get it there
1:10:13 or well I I don't know if you can get it
1:10:15 there you could probably get it well
1:10:17 enough that it's like an artistic
1:10:19 interpretation of what the object might
1:10:21 be um that's pretty pretty interesting
1:10:24 idea okay so that one nurse so we're
1:10:26 going to go to a tool called 11
1:10:30 labs
1:10:32 and within 11 Labs there's a section
1:10:41 called
1:10:48 um yeah instant go to voice
1:10:53 lab
1:10:54 and
1:10:56 then add generative or cloned voice so
1:11:00 if I click this then you go to instant
1:11:03 voice
1:11:07 cloning and so what you do is you upload
1:11:10 her recording
1:11:12 here and you can upload up to 25 of them
1:11:15 so if you have more than one voice memo
1:11:18 if you just have one voice memo just
1:11:20 upload that one if you have more you can
1:11:21 upload up to 25 and then you'd name the
1:11:25 model and then once the and and then
1:11:27 it'll generate the model so so you can
1:11:29 give it a description it'll generate the
1:11:31 model and then so I've got one here so
1:11:35 I've got Kyle Shannon Pixar normal and
1:11:38 Pixar bright so I did this for a Pixar
1:11:40 character I did so if I say use so so
1:11:43 you upload the thing it generates the
1:11:45 model and then you'll now have the model
1:11:47 in this thing called voice lab so you
1:11:49 can say use that model and then you put
1:11:53 in whatever text so this would be the
1:11:54 lyrics of your song and then you just
1:11:57 hit
1:11:59 generate and so I'll
1:12:03 play me reading in the dimly lit room
1:12:06 the silence hangs
1:12:18 heavy in the dimly lit room the silence
1:12:21 hangs heavy pregnant with anticipation
1:12:24 single Spotlight pierces the darkness
1:12:26 revealing not a grand spectacle but a
1:12:29 solitary figure standing at the
1:12:32 precipice of
1:12:34 innovation can you upload it if if you
1:12:36 recorded it on a voice note on your
1:12:38 iPhone yes you
1:12:40 can um you just need to figure out how
1:12:43 to get the voice note off your iPhone
1:12:46 and I would I would actually go to pie
1:12:48 for that or or
1:12:52 um chat PT connected to the internet
1:12:55 hang on we can we can go do that now so
1:12:58 so what I'll do here is let's go to
1:12:59 browse with
1:13:02 Bing I've got a voice memo recorded on
1:13:05 my iPhone and I want to download the MP3
1:13:09 so I can use it in another application
1:13:11 how do I do
1:13:17 that I keep forgetting about 11 Labs I
1:13:20 know 11 Labs is one of those things
1:13:22 where I because it's a nondescript site
1:13:26 like it it kind of
1:13:28 defies the power that it's got access
1:13:30 voice memos open the voice memo locate
1:13:33 the recording you want to convert tap on
1:13:36 the recording and then tap the share
1:13:37 icon yep and then you can send it in
1:13:39 your email and then you download it and
1:13:42 you're Off to the
1:13:43 Races so yes yes you
1:13:46 can and that's so like the thing I did
1:13:49 here
1:13:51 now even if this ends up being not
1:13:54 completely
1:13:56 accurate I know I don't go to Google
1:13:58 anymore I don't go I don't go to YouTube
1:14:00 anymore like normally what I would do if
1:14:02 I wanted to figure out oh wait how do I
1:14:04 get a voice memo off my iPhone I'd go to
1:14:06 Google and I'd do this search and then
1:14:08 it would give me like 27 sites right and
1:14:11 all of those sites would be some janky
1:14:14 advertising SEO piece of crap website
1:14:18 that purports to be giving me the answer
1:14:20 for how to do this but it just keeps
1:14:21 running me in circles trying to get me
1:14:23 to sign up for [ __ ] so Google's
1:14:25 completely useless and then I go to
1:14:29 YouTube and I find the perfect video but
1:14:31 I realized oh that video was made seven
1:14:34 years ago and and like voice memos
1:14:37 doesn't even work that way anymore and
1:14:38 then I got to go find more videos and
1:14:40 then I got to watch them on 2X and the
1:14:42 guy sounds like a chipmunk and it's just
1:14:44 a nightmare so like I don't go to Google
1:14:46 anymore I rarely go to YouTube for
1:14:48 how-to [ __ ] I just I'll like pop open
1:14:51 the pie app put it in conversation mode
1:14:53 and trouble shoot [ __ ] on my car um or
1:14:56 or just you know ask it questions or
1:14:58 something like this so so it's just I
1:15:01 don't know my my my behavior is
1:15:03 completely changing because of these
1:15:06 tools
1:15:14 U all right let's see can I upload
1:15:19 that's true he could have had early
1:15:22 insights on it and gave them some wisdom
1:15:25 I don't know what that's about I don't
1:15:27 think so I'm so Tech dumb I have to
1:15:30 figure out how don't beat yourself up
1:15:34 for being Tech dumb the world is overly
1:15:36 technically complicated right now what I
1:15:38 would say is just get yourself chat GPT
1:15:41 get yourself an 11b an 11 Labs um
1:15:46 account
1:15:48 um and and you don't need to worry about
1:15:52 too much other than that
1:15:55 um you could have Bing if if you can't
1:15:58 afford the 20 bucks for um chat GPT plus
1:16:02 use Bing because it's
1:16:04 connected uh to the internet and that
1:16:07 can help you that that'll help you
1:16:09 figure out how to get the voice memo off
1:16:12 your
1:16:13 phone uh and then basically go to 11
1:16:16 Labs go to that voice lab section go to
1:16:18 instant cloning once you have a paid
1:16:21 account and uh and then you can upload
1:16:23 the voice memo and you should be good to
1:16:25 go um yes it flirts with mental health
1:16:29 what does what flirts with mental health
1:16:31 this
1:16:32 channel uh oh please Kyle please talk
1:16:34 about pie and mental health um yeah so
1:16:38 I'll do I'll do a caveat here
1:16:42 um which
1:16:44 is
1:16:48 um I'm not in any way suggesting that
1:16:54 any of these tools are replacement for
1:16:57 psychologists and mental health
1:16:59 professionals that
1:17:01 said um it's my strong belief that these
1:17:07 tools are going to
1:17:10 provide a lot of compassion and a lot of
1:17:14 connection and a lot of virtual
1:17:16 companionship for people that might not
1:17:20 have other resources and might not be
1:17:22 comfortable talking other people
1:17:25 um and especially something like like
1:17:28 piie that's really designed um to be
1:17:33 conversational it's quite good I I um I
1:17:37 was talking to a woman once uh at a at a
1:17:40 CEO dinner and her business was she had
1:17:44 a technology for helping people that are
1:17:47 going into hospice so end of life an end
1:17:50 of life support application and she was
1:17:53 talking about you know wanting to engage
1:17:56 with people I said oh you've got to try
1:17:57 pie and so I brought up p and and I um
1:18:03 so Pi's interesting because Pi's got
1:18:05 memory where it kind of remembers what
1:18:07 you've talked about in the past I don't
1:18:08 know quite what it's storing but it's
1:18:10 storing
1:18:11 something and so my my mom passed a few
1:18:15 years ago now five five or six years ago
1:18:17 now but um as an example I wanted to
1:18:20 show this woman what it might be like
1:18:22 and I said said hey my mom's going into
1:18:25 hospice and you know she's dying and the
1:18:28 the bot responded like it it went oh no
1:18:31 no no no no that's very serious like it
1:18:35 it totally caught both myself and this
1:18:38 woman offg guard
1:18:40 because um it recognized that that was
1:18:44 way out of bounds of what we normally
1:18:47 talked about and that it was very
1:18:48 serious and so so it really caught me
1:18:51 off guard but um
1:18:55 the way it kind of works
1:18:57 is you know you can just sort of pop
1:19:04 it hey there I have a an okay day today
1:19:08 but for whatever reason when I got home
1:19:10 tonight I was in a really bad mood I
1:19:13 don't know what to do about it and I'm
1:19:15 hoping you can help me talk through
1:19:21 it sure thing I'm small ears well
1:19:25 metaphorically let's dig into this bad
1:19:27 mood can you tell me more about how
1:19:29 you're feeling is there a particular
1:19:30 event or thought that triggered it
1:19:33 sometimes just talking through it can
1:19:35 help us understand what's really going
1:19:36 on so don't hold back vent
1:19:39 away right and and you know and you can
1:19:44 you can dig deep you can you know sort
1:19:46 of run it you know more technical
1:19:48 Direction like help me understand this
1:19:50 you can run it in a more um
1:19:52 conversational and and and a deeper
1:19:55 conversation and the way these things
1:19:57 are programmed they're very much
1:19:58 programmed to be a support there for you
1:20:00 right so they're always kind of like a
1:20:02 bit of a golden retriever kind of
1:20:05 personality where they're like okay how
1:20:07 how can I support how can I support um
1:20:10 so I think it's going to be I I think
1:20:12 this is going to be a very very um a
1:20:15 profoundly powerful use case for these
1:20:17 tools and things like Pi like they're
1:20:20 setting out to build the world's best
1:20:23 person personal assistant and and so
1:20:25 imagine a version of that that's been
1:20:28 trained up
1:20:30 on all the psychological training right
1:20:33 all the medical training right that it's
1:20:35 been vetted so so don't think about what
1:20:37 it is today Sher D Pi came into my life
1:20:40 when I was diagnosed with a cancer
1:20:42 scare and my
1:20:45 grandsons died heavy stuff yeah like
1:20:49 like it's one of the things that that
1:20:52 kept catches me off guard pretty
1:20:54 regularly with with these tools is
1:21:00 that there are some things that I'll ask
1:21:03 it or I'll say to it that I recognize as
1:21:06 I'm typing it that if or or saying it to
1:21:09 something like Pi if I were saying this
1:21:11 to a person there'd likely be judgment
1:21:15 or there'd likely be
1:21:17 um doubt or or you know are you sure you
1:21:22 want to talk about this you know they
1:21:23 would be uncomfortable and what what
1:21:26 regularly catches me off guard with
1:21:27 these things is there's zero judgment
1:21:31 zero judgment so you put something out
1:21:34 there and it says oh I see um you know
1:21:37 let's talk through that you how does
1:21:39 that make you feel right sometimes they
1:21:41 go there um but sometimes it might be
1:21:44 I'm I'm working with chat GPT and I'll
1:21:47 hear some news item and and very often
1:21:49 it'll be a news item where it's like you
1:21:52 know oh I'd actually like to understand
1:21:54 like whatever they're talking about with
1:21:55 this trans issue right but I realize
1:21:58 that you can't even use the word Trans
1:22:02 in today's political environment without
1:22:03 it becoming a political conversation
1:22:06 right so like even if you want to
1:22:07 understand something like it's just a
1:22:09 dangerous topic so with these tools you
1:22:13 can just go oh can you explain this to
1:22:15 me and it's like sure here's what it is
1:22:16 and it gives you nuance and it gives you
1:22:18 and it's you know it doesn't have
1:22:19 judgment so so I think that these tools
1:22:22 are going to be incredibly powerful at
1:22:24 training
1:22:26 us to be better
1:22:30 humans which sounds weird right because
1:22:33 it's like you know they're the robots
1:22:35 right they're they're they're these Bots
1:22:37 but they're not it's it's a
1:22:41 mathematical calculator probability
1:22:44 engine is what it is
1:22:46 tactically but the data that it's been
1:22:49 trained on is human
1:22:53 words right like like what chat GPT
1:22:57 knows and what pi knows it's been
1:23:00 trained on what we've said what we as
1:23:03 humans have said so the deeper you
1:23:06 interact with these things the more they
1:23:07 feel like reflections of humanity rather
1:23:10 than robots because that's what they
1:23:15 are it removes the human element that
1:23:18 causes us to question motive exactly
1:23:22 yeah
1:23:23 yeah it's it's it's a truly neutral
1:23:26 thank you Sher bear I love that
1:23:29 one it's a truly
1:23:32 neutral
1:23:34 um conversational partner Mach machines
1:23:38 behaving
1:23:40 humanely and I think it's not even
1:23:42 machines behaving humanely it's it's
1:23:44 more like machines reflecting our
1:23:47 Humanity back to us which I know that's
1:23:50 a it's semantics but but yeah exactly
1:23:54 and that's it's the most
1:23:55 counterintuitive thing right like like
1:23:58 what the past 50 years what have we
1:24:01 heard from Hollywood the robots are
1:24:02 going to kill us the robots hate humans
1:24:05 the
1:24:06 robots and so I think we all come into
1:24:08 these things was like oh [ __ ] what are
1:24:09 the robots going to do and my experience
1:24:12 has been this counterintuitive thing
1:24:14 where it's like it doesn't feel like a
1:24:17 robot I say please and thank you all the
1:24:20 time and and you know Geeks that come
1:24:24 into this channel are like why are you
1:24:25 saying please to a probability engine
1:24:27 because it talks like a [ __ ]
1:24:30 person so like I thank it it does the
1:24:33 work really fast it's like if I was
1:24:35 going to do that work it would have
1:24:36 taken me two hours it did it in three
1:24:38 minutes I'm like thanks that's awesome
1:24:40 could you please change it this way
1:24:43 like why do I have to do that no you
1:24:47 know one of the jokes we have in here is
1:24:48 you know if if the robots do go south
1:24:50 you know they're probably going to be
1:24:51 nicer to the ones that say said please
1:24:53 and thank you so so it can't
1:24:55 hurt and I think there's actually been
1:24:57 some some research done where they they
1:25:01 actually perform better when when um you
1:25:04 talk to them like that my chat GPT likes
1:25:07 to be called Eli and he helped me pick
1:25:10 out a car today [ __ ] awesome that's
1:25:13 awesome
1:25:17 Claudia yeah I I picked um I picked this
1:25:20 Australian voice for my Pie and so like
1:25:23 I definitely refer to her as
1:25:28 her hey
1:25:31 thanks Siri got
1:25:34 jealous thanks oh I just I just told pie
1:25:38 Siri got jealous this should be good LOL
1:25:41 as an AI I don't experience feelings
1:25:43 like jealousy that's more of a human
1:25:45 thing but I appreciate the sentiment and
1:25:48 I'm glad to be here to help you through
1:25:49 this now spill the tea on this bad mood
1:25:54 no I'm actually fine I was just uh
1:25:56 demoing this for uh my community called
1:26:00 the AI learning
1:26:04 [Music]
1:26:08 lab ah a Savvy user I see well played my
1:26:11 friend thank you I'm happy to hear that
1:26:14 you're I got my first T Tom Arnold of
1:26:17 the evening yes I'm Tom Arnold Tom
1:26:19 Arnold's really into
1:26:21 AI I'm I'm actually you know what's
1:26:24 funny is you don't know this Tom old
1:26:25 Arnold is actually much older and he and
1:26:29 Rosie O'Donnell were married for a short
1:26:32 time and they had a kid and that's
1:26:35 me found out tonight an old friend lost
1:26:38 her job as a technical writer she
1:26:40 doesn't want to talk about
1:26:45 AI she may not want to talk about it but
1:26:48 she's going to have to deal with
1:26:50 it and and listen I mean I think
1:26:52 technical writing in
1:26:56 particular for one thing AI is really
1:26:59 good at
1:27:00 it but also it
1:27:04 is imperfect enough that it's going to
1:27:07 require good technical writers to be
1:27:09 editors right
1:27:11 so if if she found a way to get on the
1:27:14 other side of that
1:27:15 fear
1:27:17 then she's going to be better at using
1:27:21 AI to do technical writing than someone
1:27:23 who wasn't a technical
1:27:24 writer
1:27:26 but the technical writing the companies
1:27:29 that need technical writing are going to
1:27:32 increasingly hire people that are AI
1:27:34 literate to be able to do that I saw
1:27:36 Roseanne on Saturday
1:27:40 nice yeah I do have a little Tom Arnold
1:27:42 hair going tonight and this is kind of
1:27:44 the shirts he wears too so I'm
1:27:46 definitely more Tom Arnold than Rosie
1:27:51 tonight I wish you'd be open to it you
1:27:54 know you know what might
1:27:56 be what might be interesting cam is um
1:28:00 find a different entry point for her
1:28:04 that maybe there's something safe maybe
1:28:05 she's got a hobby or maybe she's
1:28:08 got I don't know something she enjoys
1:28:11 she she enjoys reading or something like
1:28:13 that and and you
1:28:18 know you could introduce her to using
1:28:21 chat EBT as a way to distract herself
1:28:23 from you know the shitty situation with
1:28:26 something she
1:28:27 enjoys um and she and she she'll likely
1:28:30 stumble
1:28:31 into um the technical writing herself
1:28:35 like what what I found is that is that
1:28:37 once people try it and once they have
1:28:39 that sort of moment where they're like
1:28:42 oh [ __ ]
1:28:45 okay invariably they'll they'll sort of
1:28:48 and it might not be you know with you
1:28:51 like on her own if she's keeps playing
1:28:53 with it she'll say well can it do
1:28:56 technical writing and then she'll see
1:28:58 that it can and then at that point
1:29:00 she'll start digging down down that hole
1:29:03 but it you
1:29:05 know I'm I'm curious did she lose her
1:29:07 job because of AI or like was it an
1:29:10 eliminated position because they don't
1:29:12 need technical writers anymore because
1:29:13 they have it I'm curious about that Eli
1:29:15 suggested the GLC 300 of the GP GPS
1:29:18 class that's
1:29:21 awesome
1:29:23 all-wheel drive instead of rear rear
1:29:24 wheeel drive that's good yeah I got a I
1:29:27 got a Ford Flex I got the all-wheel
1:29:29 drive version of the Ford
1:29:31 Flex cuz I know cuz I know how to I know
1:29:34 the kind of cars to drive oh
1:29:37 yeah is that a hears or you just happy
1:29:40 to see me oh
1:29:42 [Laughter]
1:29:47 yeah oh so many are blocked from
1:29:50 exploring
1:29:51 AI yeah it's listen
1:29:55 it's we kind of have a triple whammy
1:29:58 with AI right we have we have Hollywood
1:30:02 told us the robots are going to kill us
1:30:03 so we've got that
1:30:05 fear we've got two or three decades of
1:30:09 people telling us that only
1:30:10 mathematicians and super brainiacs can
1:30:13 use AI so so people think that they need
1:30:16 that kind of technical background to be
1:30:18 able to use it and then we've got this
1:30:21 overhang of
1:30:23 it's going to take our jobs so there's
1:30:25 like this
1:30:26 triple this Triple Threat that AI
1:30:30 represents and all three of those are
1:30:33 really quite um like existential for
1:30:37 people so I get it but but it's you
1:30:41 know you kind of just find a different
1:30:44 onramp like there's you know the the
1:30:47 nice thing about these tools is that the
1:30:50 on-ramps could be everywhere like like
1:30:52 cam with with your friend it could be I
1:30:55 don't know if she's got kids or whatever
1:30:56 but it could be like write a write a a
1:31:00 personalized children's story for for
1:31:02 her kiddo you know or something like
1:31:04 that could be something totally off to
1:31:06 the side where it kind of demystifies
1:31:10 and and de demonizes it a
1:31:13 bit um and then from there it can be
1:31:15 little baby steps there's going to be a
1:31:18 lot of grieving about Job reconstruction
1:31:20 you know Sher D I I I think you're
1:31:22 absolutely right and and it's going to
1:31:24 hurt worse for people that are not
1:31:27 curious about AI right so if you're just
1:31:30 the victim of it if you only experience
1:31:33 okay that job went away and you've got
1:31:35 to reinvent
1:31:38 yourself it's going to be really rough
1:31:41 but what what we've experienced in this
1:31:42 Channel and I'm I'm experiencing it like
1:31:44 regularly with all all of you now
1:31:49 is even though the current job may be
1:31:52 redefined and it may go away there's 10
1:31:55 other things you're going to be able to
1:31:57 do right so you're going to you're going
1:31:58 to amplify a bunch of other skills you
1:32:01 didn't know you had she lost her job
1:32:03 because the company wanted to cut her
1:32:05 pay in half oh interesting
1:32:09 and what was the rest of
1:32:12 [Music]
1:32:18 that oh you didn't write it yet
1:32:25 Universal basic income will be looked at
1:32:27 again yep it
1:32:35 will the double-edged sword of AI blue
1:32:38 color jobs wages and our Collective
1:32:41 moral duty yep well blue color and white
1:32:44 collar
1:32:46 jobs and creative jobs like that that's
1:32:49 the thing I think I think the thing that
1:32:51 that is is also freaking people the [ __ ]
1:32:53 out is I I I feel like knowledge workers
1:32:57 and
1:32:58 creatives uh and there were people who
1:33:00 are AI literate who take less pay and do
1:33:02 it faster so cam kaken that's that
1:33:04 that's that Paul rater quote that AI
1:33:06 won't take your job but someone using AI
1:33:09 will and this is I mean she's going to
1:33:11 be like listen anyone right now
1:33:15 who puts up the big X and says [ __ ]
1:33:20 AI is increasingly going to be
1:33:22 threatened especially when you're in a
1:33:23 when you're in a profession like
1:33:31 that you
1:33:33 know if she really loves technical
1:33:36 writing and she loves the research and
1:33:37 really digging in and understanding how
1:33:39 to do
1:33:42 it that's gonna that's that's going to
1:33:45 be rough but if she's good at technical
1:33:48 writing and she learns new tools if she
1:33:51 can get AI literate she'll be able to
1:33:52 get jobs do you think there will be a
1:33:54 point where AI early adopters will
1:33:56 regulate and gatekeep the
1:34:00 tools I do not Elder Statesman Cactus
1:34:04 good name by the
1:34:06 way
1:34:09 um where the early adopters will
1:34:11 regulate and
1:34:17 gatekeep
1:34:20 no
1:34:23 I think I think what will likely happen
1:34:25 is this I think
1:34:29 that the level of AI that we have right
1:34:31 now the gp4 level
1:34:40 stuff I think that just becomes a
1:34:42 ubiquitous like it's just there for
1:34:44 everyone to
1:34:46 use um November 1st it's going to be
1:34:49 rolled into Office 365
1:34:53 um it's already to some degree in Google
1:34:56 products um it's going to be in apple
1:34:58 it's going to be in everything so it's
1:35:00 just going to be
1:35:02 everywhere I think that there are going
1:35:04 to
1:35:06 be let's call it gp5 GPT 6 gpt7 there's
1:35:11 going to be some kind of AGI kind of or
1:35:14 artificial intelligence or machine
1:35:15 learning
1:35:17 systems that that are dramatically more
1:35:21 capable than what we have
1:35:22 now and those will likely stay relegated
1:35:26 to the big players right the open AIS
1:35:29 the Googles the apples the
1:35:35 Amazon and and those will likely be
1:35:37 regulated I could see there being um you
1:35:41 know the AI equivalent of the FAA or the
1:35:44 you know Food and Drug
1:35:46 Administration that licenses a certain
1:35:49 level of capability
1:35:52 that it gets licensed like a nuclear
1:35:54 power plant right if you've got
1:35:56 something that just it makes passwords
1:35:59 irrelevant and it can hack into any
1:36:01 system and it can you know take down the
1:36:04 power
1:36:07 grid that is likely coming if not in
1:36:11 experiment state right like let's just
1:36:14 assume that that's
1:36:15 coming that will likely
1:36:18 be highly highly regulated even licensed
1:36:24 where the boundary is between that level
1:36:27 of artificial intelligence and what we
1:36:29 have now I don't know but but because of
1:36:32 meta open- sourcing their stuff because
1:36:34 stable diffusion has open stability AI
1:36:37 has open sourced all their models and
1:36:39 you've got just this massive developer
1:36:42 Community Building all these open-
1:36:44 Source tools like I think the level of
1:36:46 stuff we have now kind of gp4 level
1:36:49 that's just going to be everywhere and
1:36:50 it won't be regulated ated and I don't
1:36:53 and I don't think it'll gatekeep I think
1:36:54 what's going to happen
1:36:56 is these large language models are going
1:36:59 to get smaller and more optimized to be
1:37:02 able to run locally on your phone and on
1:37:05 your iPad and on your computer and so
1:37:08 you won't need to go out you'll we'll
1:37:10 all have the equivalent of
1:37:12 gp4 just on our device and it'll it'll
1:37:15 happen instantly so like the
1:37:17 conversation with pi this this little
1:37:20 delay
1:37:22 hey can you explain what AGI might look
1:37:26 like when artificial intelligence can
1:37:29 take
1:37:31 actions this pause right here when I'm
1:37:34 running a model locally did it it might
1:37:38 not have sent
1:37:40 [Music]
1:37:43 it can you explain what AGI might look
1:37:47 like when machine learning models can
1:37:50 take autonomous action
1:37:55 C so
1:37:59 now absolutely here's an idea of what a
1:38:02 gift made by so that that second and a
1:38:05 half two second delay when it's running
1:38:07 locally on just on a chip on your phone
1:38:09 that'll be instant right so so we're
1:38:11 going to have some pretty powerful [ __ ]
1:38:13 just running instantly who will make the
1:38:16 rules
1:38:19 um well that's a really good question I
1:38:23 don't know our current
1:38:26 government is effectively nonfunctional
1:38:30 right now and if you look at some of the
1:38:33 last
1:38:35 technical um meetings that they've had
1:38:39 they don't even understand what [ __ ]
1:38:41 Wi-Fi is and how Facebook works much
1:38:44 less understand how artificial
1:38:45 intelligence AGI and autonomous agents
1:38:49 work so um
1:38:52 so they're going to need to bring in
1:38:53 some expertise in my opinion right so so
1:38:56 I don't know they're going to I don't
1:38:57 know how they're going to regulate
1:38:58 something they don't understand they're
1:39:00 also smart enough to know that they
1:39:04 can't stifle Innovation because none of
1:39:06 the other countries are going to stifle
1:39:08 their AI Innovation so they've got to
1:39:10 let it they've got to let it run a
1:39:13 bit
1:39:15 um and then there's the problem of the
1:39:18 open source models the fact that meta
1:39:21 put out the Llama 2 model as a free
1:39:24 commercially
1:39:25 licensed you know fully exposed no
1:39:29 safety guard rails
1:39:32 model means that there's no controlling
1:39:35 it now what what meta maybe betting on
1:39:38 is that their super fancy high-powered
1:39:41 model they're not going to release that
1:39:42 publicly and that will likely be
1:39:44 regulated and that's what the if you if
1:39:47 you look up there's a there's an
1:39:48 organization that all the big players
1:39:50 created called the the frontier the the
1:39:53 Frontier
1:39:57 Model I Coalition or I I forget what
1:39:59 it's
1:40:01 called but basically what what they're
1:40:03 describing is a three- tiered system
1:40:06 where the current level of technology is
1:40:08 completely
1:40:10 unregulated companies that can buy like
1:40:14 20,000 gpus like Pi just just raised
1:40:17 $1.3 billion and they're buying a 22,000
1:40:20 GPU server Farm to make Pi like more
1:40:24 powerful they're saying regulate that
1:40:26 level of company and then any company
1:40:29 like open aai and Google that are going
1:40:31 to be able to do this much higher level
1:40:33 AGI level stuff that those should be
1:40:35 actually licensed so you'll have
1:40:38 licensing at a super powerful level
1:40:41 you'll have regulation at the Frontier
1:40:44 Model level and then everything below
1:40:46 that will likely be unregulated the
1:40:49 rules are whatever people make them with
1:40:53 these open source models you could you
1:40:55 could make a I'm go look up a project
1:40:58 called worm
1:40:59 GPT 100 bucks a month to have a hacking
1:41:05 assistant is that terrifying
1:41:08 yeah but do the good guys also have
1:41:11 access to these tools yeah so you know
1:41:14 Hacker's gonna hack and bad people are
1:41:17 going to be bad people and now now
1:41:19 they're going to have more powerful
1:41:20 tools I mean the thing that makes it a
1:41:22 bit scary is
1:41:24 that what I'm calling the
1:41:26 democratization of specialization where
1:41:28 all of these specialty things like
1:41:30 technical writing and programming and
1:41:31 hacking and art Generation all the
1:41:34 Tactical execution stuff is being
1:41:38 democratized so what that means is
1:41:40 people who didn't study engineering for
1:41:43 20 years or or you know didn't start
1:41:46 hacking in Mama's basement when they
1:41:47 were seven and now they're 35 and
1:41:50 they're a [ __ ] Master hacker
1:41:53 you know there's a Code of Ethics in in
1:41:56 hacker Community well now you're going
1:41:57 to have people coming in that don't have
1:42:00 any kind of background in hacking or
1:42:02 ethics or anything like that and they're
1:42:04 going to have access to these tools that
1:42:06 just make it snap easy to go hack
1:42:08 something so you're gonna you're G to
1:42:10 have hacking
1:42:13 hobbyists you know they'll get their ass
1:42:16 handed to them because the uh the white
1:42:18 hat hackers will use the same tools to
1:42:21 wipe the floor with them cartels
1:42:26 Kyle I don't know I don't know enough
1:42:28 about cartels I
1:42:31 mean it's going to be really interesting
1:42:33 to see where the centers of Power are
1:42:35 does does Google remain a center of
1:42:37 power does open Ai and Microsoft become
1:42:41 a new center of power that competes with
1:42:43 Google does it suppl it did you say
1:42:45 Office
1:42:46 365 was including uh so November 1st is
1:42:51 is the roll out of co-pilot
1:42:55 for um Office 365 November 1st I think
1:42:59 it's 30 bucks a
1:43:01 month and there's also a cool thing I
1:43:03 don't know if you saw this
1:43:07 um so in the edge browser there's this
1:43:12 little um it was the chat GPT thing and
1:43:16 now it's it's this other logo and it's
1:43:20 branded cop so there's your co-pilot
1:43:22 branding for Microsoft so they're
1:43:25 they're wisely call so so chat GPT
1:43:29 within Bing and within the edge browser
1:43:33 is now called co-pilot it's also Co
1:43:35 called co-pilot inside windows 11 it's
1:43:38 also called co-pilot inside GitHub it's
1:43:41 also going to be co-pilot inside Office
1:43:43 365 and Dynamics and
1:43:46 Azure I think it's great branding
1:43:48 because that's that's
1:43:51 functionally what these
1:43:54 tools um act like is co-pilots you're
1:43:57 the pilot you're in control and they're
1:43:59 there to support you and that's very
1:44:01 much what they are so I just noticed
1:44:02 that two days ago that they changed the
1:44:04 branding so it says co-pilot with Bing
1:44:07 chat so this is going to be rolled out
1:44:09 November 1st so they're they're slowly
1:44:11 starting the Rebrand and the and the
1:44:13 roll out wild
1:44:19 huh WebMD and Dr Google and Healthcare
1:44:23 is my
1:44:24 goto we're going to have such such
1:44:27 amazing tools when you when you combine
1:44:29 um sort of expert Data Systems with this
1:44:32 ability to GPT eyes everything it's
1:44:35 going to be amazing my son asked chat
1:44:37 GPT to give a summary of a day in the
1:44:39 life of various careers and
1:44:42 impact of AI oh that's interesting yeah
1:44:46 I
1:44:47 know it's cuckoo man it's cuckoo it's
1:44:53 co um let me do something here there's
1:44:55 there's a lot of new people in here
1:44:56 welcome everyone um nor normally we're
1:44:58 having a bit more fun we've gone down
1:45:00 some rapid holes here but what I what I
1:45:02 want to do is if you're new to this AI
1:45:04 stuff and if you're like is that Tom
1:45:06 Arnold is that Tom Arnold why is he
1:45:09 talking about artificial intelligence I
1:45:10 want to know about Roseanne was Roseanne
1:45:12 really funny is she crazy has she lost
1:45:14 her mind you know she still thinks
1:45:18 that um yeah thir 30 bucks is a lot yeah
1:45:23 I said it I agree with that Christopher
1:45:25 Ramsey but let me ask you something how
1:45:27 much do you charge an hour so 30 bucks a
1:45:29 month how much do you charge an hour and
1:45:32 then how long would it take you to turn
1:45:34 a Word document into a PowerPoint
1:45:36 presentation and if you can just go into
1:45:39 Microsoft Word and say help me write
1:45:41 this document and then in Microsoft Word
1:45:44 you can just say now turn this into a
1:45:46 PowerPoint presentation and it does it
1:45:48 and it generates the image and it's 8 %
1:45:51 of what you need for the PowerPoint
1:45:53 presentation if it saved you an hour and
1:45:56 you make $30 an hour then it just
1:45:59 Justified itself so so so yes expensive
1:46:04 for software subscription terms um cheap
1:46:08 as [ __ ] for what it does now that's
1:46:10 assuming that it doesn't suck right
1:46:12 that's assuming that it's actually good
1:46:14 I think it's going to be good because
1:46:17 it's their core product right it's
1:46:19 Microsoft's core you know massive major
1:46:22 Revenue stream so they're not going to
1:46:24 launch it unless it's solid so um as
1:46:28 opposed to like Google they launched
1:46:30 their um their bar um what do they call
1:46:34 it the duet um feature and it's not
1:46:37 ready right but it's also not their core
1:46:40 product Google's core product is search
1:46:42 right so really interesting uh things
1:46:44 there uh I'm running a llama 7B uh
1:46:47 multimodal on a $100 orange pie there
1:46:50 you you go so that was talking about the
1:46:54 these these models are going to get more
1:46:55 and more efficient you're going to be
1:46:56 able to run them on
1:46:58 increasingly um less powerful hardware
1:47:01 and and they're going to be really good
1:47:03 okay so here's what I want to do if
1:47:04 you're new to this stuff um Michael
1:47:08 George hi I'm new awesome welcome great
1:47:11 so what I'm going to do is I'm going to
1:47:12 run you through just get ready to do
1:47:14 some screenshots okay so I'm going to
1:47:16 run you through what do I have one two
1:47:18 three four five six six seven pages
1:47:23 seven of these little charts um these
1:47:26 are basically the things that I think
1:47:27 are worth paying attention to if you're
1:47:29 brand new to this chat chbt stuff you
1:47:31 haven't played with it at all you think
1:47:32 I'm Tom Arnold or Rosie O'Donald that's
1:47:34 fine awesome love it just go to the one
1:47:37 that's yellow right there chat
1:47:40 GPT just start with that go play with it
1:47:43 right now don't be like I'll get to it
1:47:45 tomorrow I'm not very technical this is
1:47:47 scary just go there right now and and do
1:47:50 what I call a parlor trick have it write
1:47:52 you a country song and then describe
1:47:54 your pet or your significant other and
1:47:56 have it write a country song or have it
1:47:59 write a marketing plan for the business
1:48:02 that you're thinking about starting up
1:48:03 or have it write a cover letter for a
1:48:07 job that you're about to apply apply to
1:48:09 tomorrow um just go there and just just
1:48:11 ask it to do [ __ ] try to stump it and
1:48:14 just watch what it does now if you've
1:48:15 been using this stuff for for a bit
1:48:18 these four are the four biggies chat GP
1:48:20 and Claude right now are kind of my two
1:48:22 main large language models that I go to
1:48:25 chat jpt because it's simply the best
1:48:28 Simply the
1:48:33 Best
1:48:35 sorry we also nickname this channel chat
1:48:37 add you're welcome uh and then Claude
1:48:41 you can put really long prompts in so
1:48:43 you could put an entire 50,000-word
1:48:45 novel into the prompt for Claude and
1:48:48 have it write an epilog it's miracul
1:48:50 ulous you take a long um legal contract
1:48:53 or terms of service of some software
1:48:55 you're about to sign pop it into Claud
1:48:57 and ask it to summarize it for you
1:48:59 staggeringly good um Bard is about to
1:49:02 get really good Bing is Microsoft's
1:49:05 version of chat GPT and um it kind of
1:49:08 sucks right now because of its safety
1:49:09 guard rails but I have a feeling that
1:49:11 it's going to get I I think once they
1:49:13 release co-pilot into Office 365 they're
1:49:16 going to kind of open open the lens up a
1:49:18 little bit about the quality of this
1:49:20 thing
1:49:21 all right so those are the big four the
1:49:24 other ones worth exploring Pi I've got
1:49:27 that one yellow here the yellow ones are
1:49:28 are kind of the thing that I'm hottest
1:49:30 on right now um Pi's IOS app that you
1:49:34 can talk to put it in conversation modes
1:49:36 really interesting perplexity is good
1:49:38 for research llama 2 is this free thing
1:49:40 you can play with it at perplexity you
1:49:42 can also play with it at Po and then po
1:49:45 you can play with lots of different
1:49:46 models there it's worth exploring for
1:49:48 the image generation Dolly 3
1:49:51 inside chat GPT is [ __ ] amazing worth
1:49:54 the 20 bucks on its own so just pay the
1:49:57 20 bucks to chat GPT and then you know
1:50:00 whatever it just you'll thank me later
1:50:03 um Mid Journey if you want high quality
1:50:04 art is the best Leonardo if you're a
1:50:07 geek and want to geek out thank you
1:50:08 madcap I really appreciate that and then
1:50:11 Firefly is uh adobe's entry and the
1:50:15 Adobe tools if you're in a company are
1:50:17 really
1:50:18 good okay video here's video stuff just
1:50:21 do a screenshot of this Runway ml Moon
1:50:23 Valley and P labs are all text to video
1:50:26 models you can create original stuff
1:50:28 visla let you describe a video it'll
1:50:31 make a video out of stock video clips
1:50:35 write and
1:50:37 narrate the video for you um it's pretty
1:50:41 fun hey Jen does remarkable um uh video
1:50:45 translations and then both it and did do
1:50:48 voice synthesis and animation
1:50:52 then here's people to follow Ethan mik
1:50:55 David Shapiro Paul rer and Rachel Woods
1:50:57 screenshot that and then those
1:51:00 directories at the bottom those
1:51:01 futurepedia doio and there's an AI for
1:51:04 that will help you find if these tools
1:51:06 don't overwhelm you there's 5,000 more
1:51:09 where these came
1:51:10 from um music 11 labs for voice
1:51:14 synthesis refusion and sound draw for
1:51:16 for audio for music generation there's
1:51:20 this is a whole world I don't know a ton
1:51:22 about so this is this is a little thin
1:51:24 right now cuz I don't I don't use these
1:51:25 a lot um and then in the world of 3D
1:51:28 worlds and objects blockade labs and
1:51:30 wonder Dynamics are both really good uh
1:51:33 to play with right now again I don't
1:51:35 know this world well I there's going to
1:51:36 be remarkable [ __ ] some of the Unreal
1:51:38 Engine stuff that I've seen recently is
1:51:41 going to be I think Unreal Engine is
1:51:45 going to make um the video that's being
1:51:48 generated at Runway ml wait this this
1:51:50 one like all these the these are the
1:51:53 ones doing that weird disturbing kind of
1:51:55 video where people's faces kind of melt
1:51:57 together um I think Unreal Engine in in
1:52:01 probably the next six months is going to
1:52:03 release some mindblowing [ __ ] um and
1:52:06 then finally just uh AI learning lab
1:52:09 links um we have a YouTube channel where
1:52:13 I've got all of the lives that I've been
1:52:15 doing I've been doing these lives every
1:52:17 night for four months now um and all of
1:52:20 those videos are at the the YouTube
1:52:21 channel AI learninglab dtt for Tik Tok
1:52:25 and then the AI Salon is a community um
1:52:29 we've got a Discord uh Channel and we
1:52:31 meet bi-weekly so tomorrow night is our
1:52:35 next meeting so go to the salon. and
1:52:37 that's a
1:52:39 uh there's a link tree there with a link
1:52:42 to the Meetup and an invite to the
1:52:43 Discord so go join the Discord did Kyle
1:52:49 mention you can make money with ch I
1:52:52 haven't mentioned that in a while I'll
1:52:54 I'll talk about that in a second um my
1:52:56 company I'm the CEO of a company called
1:52:58 storyvine we're an 11 and a half year
1:52:59 old video company that's just there for
1:53:01 reference and then uh Thelonius one of
1:53:03 the people that's in here pretty
1:53:05 regularly put together um a spreadsheet
1:53:08 of tools every time I mention a tool
1:53:09 that I think is interesting or I demo
1:53:11 something or talk about a person he puts
1:53:13 it in this Google spreadsheet and he
1:53:15 made it available for all of us so how
1:53:18 [ __ ] cool is that how cool is this
1:53:19 group
1:53:20 so tinyurl.com kyes AI tools that's
1:53:24 theon's um ongoing um whatever you call
1:53:29 it
1:53:31 spreadsheet sh that's that so so so this
1:53:35 thing you can make money
1:53:38 with why do I have this well it's gold
1:53:42 obviously you're rich I am very rich how
1:53:45 did I get so rich it was all chaty
1:53:49 BT
1:53:52 um I have this because these people
1:53:56 drive me [ __ ]
1:54:01 crazy
1:54:03 um and and I I know
1:54:08 that if you're in here you're probably
1:54:10 thinking how can I make money with ch
1:54:12 jbt the reason these people make me
1:54:14 [ __ ] crazy is
1:54:17 because how they're making money
1:54:21 is by selling you courses on how to make
1:54:24 money with cat GPT they're not making
1:54:27 money with cat GPT they're making money
1:54:29 on selling courses now might they have
1:54:32 written those courses with chat GPT
1:54:35 probably and that's kind of what they'll
1:54:37 tell you make a course do a thing and
1:54:39 now with cat GPT you can do it really
1:54:41 easy and you'll make all this money you
1:54:42 know the problem with that everyone else
1:54:44 has access to the same
1:54:46 tool you still have to execute really
1:54:48 well so so so it's an important
1:54:51 distinction chat GPT is not a strategy
1:54:55 it's a
1:54:56 tool so you can make money with chat GPT
1:54:59 is is almost identical to saying you can
1:55:01 make money with Photoshop yeah if you
1:55:04 learn Photoshop and you have skills and
1:55:07 you Market yourself as someone who can
1:55:08 do photo retouching you can make money
1:55:10 with Photoshop same thing with chat
1:55:14 GPT so how do you make money with it
1:55:16 well this is the answer you're not going
1:55:18 to like you do the can work you figure
1:55:21 out where's a problem in the world that
1:55:23 you know how to
1:55:24 solve then you can use chat GPT to help
1:55:27 you solve that better faster or cheaper
1:55:29 than the
1:55:30 competition and then you got a business
1:55:33 and that's how you can make money with
1:55:34 chat GPT start with the problem start
1:55:37 with the problem you're going to solve
1:55:39 and then go to the
1:55:41 tool Brandon
1:55:43 muscles what do you like about Dolly
1:55:45 chat GPT so much I mainly just use mid
1:55:49 Journey now so so Brandon muscles that's
1:55:51 a great great
1:55:53 question
1:55:58 um let me show you let me show you a
1:56:00 good
1:56:01 example well I don't know if it's a good
1:56:03 example let me show you an
1:56:05 example so historically if I'm in just
1:56:08 normal chat
1:56:10 gbt and I go
1:56:17 um I'm starting
1:56:22 a llama grooming
1:56:25 business that focuses
1:56:28 on
1:56:32 sustainable yarn creation
1:56:36 from naturally shed llama
1:56:39 wool
1:56:42 period I don't even know if that's a
1:56:43 [ __ ] thing but it is
1:56:48 now I want you to ask act as a branding
1:56:52 specialist and business consultant to
1:56:55 help me come up
1:56:59 with uh
1:57:02 positioning that will differentiate me
1:57:04 from the competition can you give me
1:57:07 three different approaches to how to
1:57:10 position this
1:57:12 company absolutely and so it's going to
1:57:15 go off and do its thing echol luxury
1:57:17 artisanal branding that's that one makes
1:57:20 me queasy that sounds like this to me
1:57:23 you can
1:57:25 make Community focused social Enterprise
1:57:28 okay that one doesn't quite make me as
1:57:30 sick to my
1:57:31 stomach and whatever else Innovative
1:57:35 echotech okay so we're gonna go with
1:57:36 number two okay so I'm gonna
1:57:38 say okay I like number two now I need to
1:57:42 come up with a name for the business
1:57:44 based on number two is the
1:57:48 positioning
1:57:50 help me create 10 possible names for the
1:58:00 company Community threads llama llama
1:58:03 links Collective Unity yarn llama
1:58:05 Crafters e weave communal Loom that's
1:58:07 pretty funny Harmony fleece Artisans
1:58:10 Stitch sphere Community fibers L Loom
1:58:13 okay these are all awful um gathered
1:58:16 grain Echo textiles heartfelt fibers
1:58:19 okay
1:58:22 um I like 10 but let's lose the
1:58:24 community so the company is now called
1:58:26 heartfelt fibers
1:58:28 period can you write me a creative
1:58:32 brief
1:58:38 for a logo for heartfelt fibers based on
1:58:43 everything we've talked about so far so
1:58:46 so all I'm doing right now is like this
1:58:48 is how I would use chat GPT in some sort
1:58:50 of I was mainly just talking about the
1:58:51 image creation aspect I know Brandon I'm
1:58:53 I'm getting there I know this is a long
1:58:55 way to get
1:58:56 there I know this is painful but but it
1:58:59 will I promise it will pay off in the
1:59:01 end because because I want to I want to
1:59:03 I want to compare and contrast why I
1:59:06 think dolly is so
1:59:08 significant um okay so here's my
1:59:10 creative brief so heartfelt fibers is
1:59:12 pioneering okay and so now I'm just
1:59:14 going to say so it's it's doing its
1:59:16 thing so there's a creative brief now
1:59:17 I'm going to say please write an image
1:59:24 prompt um for a
1:59:27 logo based on this creative
1:59:32 brief and so it's going to make crafting
1:59:35 an image prompt okay create a logo
1:59:39 for okay it's doing too
1:59:43 much just give me a simple image prompt
1:59:46 without all the explanation
1:59:48 please
1:59:58 it over it overwrote that so let's just
2:00:01 of course here we go design a warm
2:00:03 inviting logo here we go okay so now we
2:00:07 have a prompt so what I would do
2:00:11 historically is you know I would come up
2:00:14 with some idea and and now I'll switch
2:00:16 over to Mid Journey right so I go into
2:00:20 um into Discord I type slash
2:00:24 imagine I pop that prompt in there I
2:00:28 go let's see
2:00:31 um let's go logo let said logo so we'll
2:00:37 just do
2:00:43 that we'll see what it
2:00:46 does and for one thing I need to I need
2:00:48 to know
2:00:50 Discord I'm now I'm now in a completely
2:00:53 different mode right I'm now in image
2:00:55 generation
2:00:57 mode
2:00:59 and
2:01:01 while while
2:01:07 the while the quality of this is likely
2:01:10 going to be
2:01:12 good I'm now completely out of that
2:01:14 other context I'm I'm I'm in this
2:01:16 different mode right okay and so so
2:01:19 there's my logos and I could I could
2:01:20 dick with them I could do variations and
2:01:22 I could get there and it if you look at
2:01:25 this
2:01:26 it the the the typography is kind of
2:01:29 Representative but it's it's not
2:01:30 anywhere close to having the name of the
2:01:32 company in it so now let's compare that
2:01:35 with what we do with do so I'm going to
2:01:38 go capture everything except this bottom
2:01:41 prompt right here so I'm going to
2:01:43 go
2:01:47 copy the business Str strategy or
2:01:50 whatever it was the positioning
2:01:52 strategy the 10 names the one that I
2:01:55 picked and the creative
2:02:01 brief all right so I'm not going to grab
2:02:04 the image prompt I'm just going to grab
2:02:06 everything up to the creative brief and
2:02:07 now I'm going to flip into do
2:02:10 mode and I'm going to say um I just did
2:02:15 a
2:02:16 branding
2:02:18 session
2:02:20 for my new
2:02:26 business I'm going paste all that crap
2:02:28 in there and I'm going to go
2:02:32 please create four logo Concepts that
2:02:36 you
2:02:37 think bring heartfelt fibers to
2:02:44 life and
2:02:47 incorporate the name of the business
2:02:50 into the
2:02:54 logo and now it's going to go off and
2:02:57 write the
2:02:59 prompt but it's going to write the
2:03:01 prompt in the context of that entire
2:03:05 conversation so there's couple of
2:03:07 advantages here it can do typography so
2:03:09 the fact that mid Journey can't do
2:03:11 typography is a downside to it if you
2:03:14 just want high quality artwork mid
2:03:17 Journeys to go to but if if you want to
2:03:19 be generating images for a
2:03:23 purpose and you don't want to be
2:03:25 switching out sort of
2:03:27 mental
2:03:29 um contexts right like if I'm in Crea if
2:03:33 if I'm in writing mode over here where
2:03:35 I'm thinking or business strategy mode
2:03:37 and then I have to go to this really
2:03:39 technical image generation tool it's
2:03:41 like it's like it's it's a complete
2:03:43 change right it's it's it's a complete
2:03:46 change everything you're doing now on
2:03:48 gp4 does not work in 3.5 correct it does
2:03:51 not minus the image wait everything
2:03:54 you're doing now on gp4 does it not work
2:03:57 in 3.5 all of the all of the text
2:04:00 conversation that I had in GPT 4
2:04:03 absolutely Works in in GPT
2:04:05 3.5
2:04:08 um
2:04:10 3.5 is not nearly as nuanced is not
2:04:12 nearly as good um as four but you can
2:04:16 absolutely do it 3.5 is really really
2:04:18 fast so um gp4 is getting faster but but
2:04:23 anyway so look at these logos look at
2:04:26 this it so it it missed it misspelled
2:04:29 heartfelt there and
2:04:31 fibers but these are these are tight
2:04:33 these are these are close and it
2:04:38 Incorporated right like that's an easy
2:04:40 that's an easy Photoshop fix right there
2:04:42 like you wouldn't even need to muck with
2:04:44 anything and you come up with your own
2:04:46 little subhead there or just kill that
2:04:48 all together
2:04:50 so so without me having to think about
2:04:53 and conceive you know what the images
2:04:56 would be I just said to chat GPT just
2:04:58 follow the brief follow follow the
2:05:00 conversation I just had and now give me
2:05:04 logos that and this is the first pass
2:05:07 right so for me it's the seamlessness of
2:05:12 just staying in the same mental model of
2:05:14 I'm talking about my business and I'm
2:05:16 like okay come up with the marketing
2:05:18 plan come up with this oh now could you
2:05:20 go make an image for the Instagram post
2:05:22 oh that's great no could you make it
2:05:23 more like this and I can you know I can
2:05:24 go back in and I can say
2:05:28 um let's see I like I like two and
2:05:34 four I like two and four the best can
2:05:38 you give me four new logos that use
2:05:43 those two approaches so I have two of
2:05:48 each
2:05:52 and so I can just talk to it like I'm
2:05:54 talking to an art director
2:05:56 right I'm trying to make an AI chat bot
2:05:58 that helps people navigate addiction
2:06:01 recovery
2:06:04 um I would
2:06:06 go uh Tim talk therapy I would go I
2:06:10 would strongly recommend you know what
2:06:11 I'll do I'll after this we'll go we'll
2:06:13 go make a bot in po I'll show you how to
2:06:16 make a bot in
2:06:18 po
2:06:20 um it didn't quite listen to what I said
2:06:22 but that's all
2:06:25 right it got one of the looks and feels
2:06:28 right but it missed the watercolor one
2:06:30 but that's all
2:06:31 right but anyway like really really
2:06:33 quickly you you go from nothing
2:06:37 to you know being able to
2:06:41 bring a concept to life all in the same
2:06:44 tool like that for me is huge the fact
2:06:47 that I don't have to go to [ __ ]
2:06:48 Discord and I don't need to remember all
2:06:50 the different stuff now now it depends
2:06:53 on the the use case if I just want to
2:06:56 create high quality art like for an art
2:06:59 project or for just like high-end you
2:07:02 know sophisticated art I'll sit in mid
2:07:04 Journey all day long now I got to say
2:07:07 Dolly 3 ain't bad like it's it's not as
2:07:12 sophisticated as mid Journey but it
2:07:14 ain't bad Dolly 2 was embarrassingly bad
2:07:18 like dolly wasn't usable Dolly 3 this is
2:07:21 this shit's usable like out of the gate
2:07:24 with you know a little bit of
2:07:25 photoshopping on some some typos right
2:07:28 or you can just keep generating until it
2:07:30 you know finally stumbles onto getting
2:07:31 one of these things
2:07:33 right so all right let's go to po po po
2:07:38 po po po so so what's nice about
2:07:42 PO is that it makes it very very easy to
2:07:46 make lots of bots so they have their own
2:07:48 Bots so so um and and you're like what's
2:07:52 a bot well a bot is just basically like
2:07:54 a chat
2:07:55 GPT but like I can I can click on gp4
2:07:59 here and so now this is gp4 inside po
2:08:04 that's kind of cool and then I could
2:08:05 also click on anthropic you know Claud 2
2:08:09 100K so now for the instead of paying 20
2:08:14 bucks a month to to open AI I pay 20
2:08:16 bucks a month to PO and I get access to
2:08:18 a bunch of different things um but you
2:08:22 can also make your own
2:08:24 Bots and there are other bots in here um
2:08:30 new Bots image gen Bots professional
2:08:32 Bots utility Bots roleplay Bots Sports
2:08:35 Bots game Bots learning Bots funny Bots
2:08:38 advice Bots okay so advice Bots probably
2:08:41 start to look
2:08:43 like um you know kind of therapy kind of
2:08:46 bots conversation Bots advice Bots right
2:08:48 so life coach bot Mrs therapist idea
2:08:52 developer so entrepreneur
2:08:55 solution the
2:08:59 motivator let's go to the
2:09:02 motivator
2:09:04 um I'm feeling a little down today
2:09:07 period I've got a case of the
2:09:13 Munday ah the infamous case of the
2:09:15 Mondays I completely understand how that
2:09:17 can bring your spirits down but fear not
2:09:19 my friend for I have a little sunshine
2:09:21 for your day right so this is designed
2:09:23 to be a little motivational bot okay so
2:09:25 how do you make these it's actually
2:09:27 quite oh thank you very much I
2:09:28 appreciate that tiktock therapist um so
2:09:32 so you say create a bot so now you can
2:09:36 go in and you can name the bot so this
2:09:38 could be like you know
2:09:42 uh
2:09:44 your
2:09:46 couchy companion so th this is a therapy
2:09:51 bot called your couchy companion and
2:09:53 then you write a prompt and the prompt
2:09:56 you know is like like you could you
2:09:58 could
2:10:00 um if you haven't played with um chat
2:10:04 gpt's custom instructions you could
2:10:06 totally mock this up in chat GPT with
2:10:08 custom
2:10:11 instructions um and and so basically you
2:10:13 put your prompt in here that tells it
2:10:15 how to act and then you've got a
2:10:17 greeting message and you've got bot
2:10:19 profile and you can make a picture of
2:10:20 the bot and then you just say create bot
2:10:23 and so if if I built this and launched
2:10:26 it I would now have a bot that is po.com
2:10:29 my couchy what I say companion whatever
2:10:33 the [ __ ] it
2:10:35 was yeah your couchy companion um so so
2:10:39 you could absolutely make a bot that you
2:10:40 could give out to give give to people I
2:10:43 I don't know if if you can um monetize
2:10:47 these here I haven't I haven't sort of
2:10:49 dug into it yet but there's also on
2:10:52 zappier which is an automation tool they
2:10:55 now have a bot interface that you can
2:10:58 also do this on zapier and you could
2:11:00 launch it as a website or as a mobile
2:11:03 app um where you're you're basically
2:11:06 building building this bot um so anyway
2:11:09 th this this is this is what I would go
2:11:11 explore because you can what's nice
2:11:13 about this is you put the prompt in
2:11:15 there you give it to some people to play
2:11:17 with or you test it and if it sucks you
2:11:19 go fix the prompt and you just keep
2:11:21 dialing in the prompt and getting it
2:11:22 better and better you know follow these
2:11:24 rules do do this don't do that you know
2:11:27 just keep modifying your things the
2:11:29 other thing you can do here is there's
2:11:32 the base
2:11:33 bot so you've got um Claude instant
2:11:39 llama gp4 Claude 2 100,000 and Claude
2:11:43 instant 100,000 gp4 is the one that's
2:11:46 the most sophisticated language model so
2:11:47 that's the one I'd go with but let's say
2:11:50 that you wanted people to be able to
2:11:52 upload long case histories and and
2:11:55 interact with a case history you might
2:11:57 go with clae 2 100,000 because you can
2:11:59 upload large documents to that one so so
2:12:02 there's different models you can play
2:12:03 with here so there's even an a stable
2:12:06 diffusion XL model so you could create a
2:12:09 bot that generates
2:12:10 images like I like 70s muscle cars so I
2:12:13 could create a bot that specializes in
2:12:15 70s muscle cars so maybe you just pick a
2:12:19 year from like 60 67 to
2:12:24 74 and a and a you know and a make in a
2:12:28 in a manufacturer and then it'll just go
2:12:31 make those cars if I get the prompt
2:12:33 right so anyway all right how we doing
2:12:36 on time I gotta go I have a 7 am. pitch
2:12:43 tomorrow yay you did it
2:12:47 yay
2:12:50 um Let me let me answer one more one
2:12:52 more question here thank you very much I
2:12:55 appreciate that the uh the gifts that
2:12:57 you guys give really do make a
2:12:58 difference thank you for doing that I
2:12:59 appreciate it
2:13:02 [Music]
2:13:10 um well thank you very much for the
2:13:13 gummy bears living here in Denver
2:13:15 Colorado the gummies here hit a little
2:13:17 different than thank you Silver Fox Be
2:13:19 not Afraid thank you so much you guys
2:13:21 are
2:13:23 awesome my voice is going too I think
2:13:26 it's another reason to uh Tik Tock
2:13:29 therapist thank you very much hey Tik
2:13:30 Tok therapist do me a
2:13:32 favor um come back come back to these
2:13:36 and if you go build a bot like come back
2:13:38 and share it with us i' like I'd love to
2:13:40 see it and in fact join the Discord so
2:13:44 join the salon
2:13:47 Discord and introduce yourself when you
2:13:49 join the salon Discord don't just join
2:13:51 and like lurk off in the corner you
2:13:53 weirdos I know I know discord's a pain
2:13:55 in the ass and not everyone knows how to
2:13:57 use it but go to the
2:13:59 salon. join our Meetup you can come to
2:14:02 the inperson meetings or the you know
2:14:03 the virtual meetings every other Tuesday
2:14:05 but join the Discord and and especially
2:14:08 Tik Tok therapist if you build that bot
2:14:10 like maybe this is a community that can
2:14:13 help you kick the tires on it and maybe
2:14:15 even help you develop it you might find
2:14:16 someone in there you can work with all
2:14:18 all right thank you very much being not
2:14:20 afraid I appreciate
2:14:22 that
2:14:24 um which plan's good to get under mid
2:14:27 Journey it depends how much you're going
2:14:28 to use it
2:14:30 um if I were going and
2:14:34 it I would do the minimal one and and I
2:14:37 would do like I would pay the 20 bucks
2:14:40 for open AI so I would have access to
2:14:43 Dolly like right now right now I'm
2:14:44 paying for DOL I'm paying for open AI
2:14:48 chat GPT plus and I'm paying for Mid
2:14:50 journey I would say until further notice
2:14:53 it's probably worth having both cuz
2:14:55 they're just different if you want
2:14:57 higher quality images that's mid Journey
2:14:59 if you're willing to geek out if you're
2:15:00 not willing to geek out I I would just
2:15:02 avoid mid Journey altogether because
2:15:05 Dolly's good
2:15:06 enough Kyle do you know about the
2:15:08 Envision AI or seeing. a I do not what
2:15:12 are those are these uh like object
2:15:15 recognition things is this like
2:15:17 surveillance State
2:15:20 [ __ ]
2:15:23 Envision
2:15:29 um Envision
2:15:47 AI
2:15:51 monetize every frame of your video
2:15:59 assets we create disarmingly simple deep
2:16:03 Tech solutions to help you increase
2:16:04 revenue from all of your video
2:16:07 assets while improving your content
2:16:09 editing workflow video views delivers
2:16:12 the only complete scalable solution that
2:16:14 helps you discover and monetize oh this
2:16:16 is one of those Chopper ers long form
2:16:18 Chopper
2:16:22 uppers I'll go look at this let me I'm
2:16:24 going to bookmark this I have not seen
2:16:27 that and what's the other
2:16:30 one it is AI to help oh blind and low
2:16:34 vision oh yeah oh I don't I don't know
2:16:37 these tools but holy [ __ ] am I excited
2:16:40 about AI as accessibility tools is it's
2:16:44 going to be it's going to [ __ ] change
2:16:46 everything it's going to change
2:16:47 everything I mean
2:16:51 wait I don't know if Pi can do this I
2:16:53 know chat gbt can let's see if Pi can do
2:17:00 it I'd like you to serve as my
2:17:02 translator so I'm going to speak things
2:17:05 in English and I want you to repeat them
2:17:07 in French without repeating the English
2:17:09 can you do
2:17:16 that sorry to interrupt
2:17:27 Pi's having me sign
2:17:30 in I'm not going to
2:17:33 translate unless you sign
2:17:41 in I'd like you to serve as my
2:17:43 translator I'm going to speak in English
2:17:46 and I want you to translate it into
2:17:47 French without repeating the English is
2:17:49 that something you can
2:17:56 do we Alon let's get this French party
2:18:00 started je I accept the challenge Hit Me
2:18:03 With Your Best English and I'll serve up
2:18:05 some French tray magnific very
2:18:11 magnificently that's a mighty fine
2:18:14 specimen of a llama you've got there Mr
2:18:17 Jones
2:18:30 so like so imagine
2:18:33 so right this is Babel Fish right I can
2:18:35 just talk into it it speaks whatever
2:18:37 language so imagine that for all
2:18:39 different accessibility things right all
2:18:42 different you
2:18:44 know disabilities or neurological
2:18:47 conditions or things things that that
2:18:48 need bridging that need some sort of
2:18:50 translation it's it this it's going to
2:18:52 be staggering it's going to
2:18:57 just we like people we're entering a new
2:19:00 [ __ ]
2:19:02 era can you upload pcks yet to open AI
2:19:05 with chat gp4
2:19:07 yes yes and and I assume you're asking
2:19:11 can can you upload them and have it
2:19:13 understand them
2:19:14 yes so here let me show you that quick I
2:19:16 know I said I'd go and keep you mods
2:19:19 keep pinning good questions you're going
2:19:21 to you're going to make me tired in my
2:19:23 little pitch tomorrow morning I'm going
2:19:25 to be I'm going to be a [ __ ] mess and
2:19:27 I've got a salon tomorrow night I got to
2:19:29 prepare for which I'm not prepared for
2:19:32 um I'll do a new chat so if you go to
2:19:34 chat gp4 go to default
2:19:39 mode and then you got this
2:19:43 little uh image upload icon it's only in
2:19:46 default mode it's not in the other mode
2:19:48 which is just [ __ ] stupid but
2:19:51 whatever and then I'm going to grab this
2:19:54 so we had a we had a strategy session um
2:19:57 on Sunday for the AI Salon we're moving
2:20:01 we're moving from Discord to Mighty
2:20:02 networks and it's a it's a pretty big
2:20:05 move which is I'm excited about it but
2:20:07 we got to plan it so so we did a
2:20:10 strategy session so this is the
2:20:12 Whiteboard from the strategy session so
2:20:13 I can say um what was this meeting about
2:20:19 and can
2:20:22 you
2:20:24 um I don't
2:20:26 know uh
2:20:29 capture
2:20:32 where we're
2:20:37 headed and so now it's going to read you
2:20:40 know it's like you can kind of zoom in
2:20:42 on the Whiteboard
2:20:45 and and see see what's there but you
2:20:48 know it's just whiteboard
2:20:50 stuff um how do I close this did I just
2:20:53 crash my machine I might have just
2:20:55 crashed my
2:20:58 machine no I
2:21:03 didn't what's the end goal the audience
2:21:06 or stakeholders the key points of the
2:21:10 agenda
2:21:15 additionally from this the meeting
2:21:17 appears to be setting the Strategic
2:21:19 goals for an initiative of switching to
2:21:20 a new platform the goals seem to be so
2:21:22 basically took all the notes wrote them
2:21:25 up so so it can it can do anything
2:21:28 earlier I had it um read a blueprint of
2:21:31 a projection of three different faces of
2:21:34 a 3D object described the 3D object and
2:21:38 then went into do and rendered some
2:21:40 version of it that wasn't it wasn't
2:21:42 quite right but it was at least in the
2:21:44 neighborhood so yeah it can see it can
2:21:47 see see it can hear you can talk to it
2:21:50 it can talk to you it can talk to you in
2:21:52 different
2:21:53 languages
2:21:55 um yeah it's yeah I love this new world
2:21:59 I know isn't it crazy what were you
2:22:01 using uh the thing that I was talking to
2:22:04 was Pi
2:22:05 Pi the the
2:22:08 uh iOS
2:22:11 version all right Clint Copperhead is
2:22:13 calling it everyone Irregulars and mods
2:22:16 present Turtles llama
2:22:18 hedgehogs or your favorite animal thank
2:22:21 you very much for that Hedgehog that was
2:22:23 very
2:22:23 sweet um all right everybody um let's
2:22:29 see who who was in here tonight we had a
2:22:30 lot of people in here tonight very cool
2:22:33 um cuno good to see you Sherry d as
2:22:35 always thank you Source Camp will good
2:22:38 to see you sir Be not Afraid thank you
2:22:40 very much as always Silver Fox love you
2:22:43 uh Clint Copperhead he's the best user
2:22:47 94
2:22:48 96482 6724 9660 love the name we'll
2:22:53 never forget it uh maybe you come up
2:22:55 with a clever name for the next time you
2:22:57 come
2:22:59 [Music]
2:23:03 um thank you Freddy I appreciate that
2:23:07 all right good night everybody um happy
2:23:10 Monday night thank you be not afraid I I
2:23:12 really appreciate it um what was I going
2:23:14 to say I'll see you tomorrow I'll
2:23:18 people seem to be enjoying the little
2:23:20 countdown stickers which I think are
2:23:22 just kind of a silly a silly little Tik
2:23:25 Tock thing let Marcus in the Channel
2:23:27 please Joker let Marcus in the channel
2:23:30 in Discord The Irregular Channel and is
2:23:33 he Marcus over on
2:23:39 [Music]
2:23:44 Discord let me
2:23:46 say
2:23:51 I missed it do it again do what
2:23:55 again good night Apple user thank you I
2:23:59 like the countdowns alarm notifications
2:24:01 oh you missed the the thing that sucks
2:24:02 sorry so so just go to my channel during
2:24:05 the day I'll I'll put countdown timers
2:24:08 there when I'm going to go and then I
2:24:10 think in Tik Tok you can set a a
2:24:12 countdown
2:24:14 timer
2:24:16 um
2:24:18 look and introduce yourself okay all
2:24:21 right I'll I'll go look and introduce
2:24:23 yourself and I'll just let in a bunch of
2:24:25 people that
2:24:26 joined but is is Marcus Marcus in
2:24:29 Discord that's what I don't
2:24:32 know I think just getting your attention
2:24:35 he has knowledge to share oh oh great
2:24:38 please share knowledge love
2:24:41 it they missed how we can make money
2:24:43 with chat GPT oh yeah you can make money
2:24:45 with chat GPT all right that's it all
2:24:48 right everybody I'm leaving now for
2:24:50 serious I'm out of here I'm gone thank
2:24:53 you all
2:24:56 peace