aiLL 8/12/23 - Wild Times Ahead: Generative AI's Impact on Jobs and Society

Description

Chapters

Transcript

0:01 foreign
0:21 happy Saturday night welcome to the AI
0:24 learning lab
0:28 not the AIL learning lab
0:32 those would be scammers I even made a
0:35 chart
0:38 we've got charts and graphs charts and
0:41 graphs
0:43 welcome everybody welcome welcome
0:45 welcome welcome happy Saturday night my
0:47 name is Kyle Shannon this is the AI
0:48 learning lab
0:50 uh
0:52 uh just had din-din
0:56 good evening Dr J was shaking
0:59 was shaking what's going down to bias
1:03 date night for you and the misses
1:05 [Laughter]
1:09 let's see it's the Barbie movie the AI
1:11 learning lab hmm
1:13 foreign
1:20 welcome everybody welcome welcome
1:21 welcome welcome we'll let people get in
1:24 here we'll get a little settled
1:27 I still have my cinnamon from last night
1:29 but I'm not putting it in my coffee
1:31 because that was a disaster
1:34 plus one also welcome welcome both of
1:36 you
1:38 Beretta Barbie nice nice name
1:41 that's a Beretta Barbie that's a uh
1:44 that's a 10 out of 10 on the name 76 of
1:46 Dr J 76ers gen xer for sure
1:52 go Gen X oh yeah oh yeah hey Kool-Aid
2:09 uh if you'd be so I come off strong
2:13 if you'd be so kind share the live so we
2:16 can get some other folks in here tap the
2:18 hearty hearts
2:20 so you know we at least pretend you're
2:23 enjoying it
2:25 laughs
2:27 the Joker's here
2:29 welcome Joker
2:33 thank you Dr J
2:37 en Jennifer Murray photography hello
2:39 hello hello
2:42 Danny D Roger Martin welcome welcome
2:50 all right now it seems to be popping up
2:52 a little bit maybe we're hitting the
2:54 algorithm
2:55 Saturday night
2:59 there's an MMA fight on you could be
3:02 watching that or you could be watching
3:03 the MMA fight that is the AI learning
3:06 lab
3:08 what's what is on
3:10 must be paint
3:12 it's white and it will not it's making
3:16 me crazy
3:18 I can't get off that famous scene from
3:22 Macbeth
3:30 or you could be working what are we
3:32 chatting about tonight well
3:35 laughs
3:36 no big pressing no big pressing issues I
3:39 made a couple of videos today
3:42 um
3:44 there's a cool
3:46 I don't even remember the site there's a
3:48 cool site
3:50 that builds you a website
3:54 it's a cold
3:57 relume r-e-l-u-m-e
4:02 imagine a website now generated it
4:05 generates a site map and a
4:08 um a set of wireframes that you can then
4:11 export to figma or word uh WordPress no
4:15 I don't think it's WordPress
4:16 I think it's
4:19 uh webflow
4:22 re-loom so that's pretty amazing oh a
4:25 lot of people just join welcome
4:26 everybody my name's uh Kyle Shannon this
4:28 is the AI learning lab
4:31 um
4:32 let me let me flip these cameras and
4:35 I'll I'll talk a little bit about
4:37 what the hell we do here why are we
4:39 hanging out on a Saturday night look I
4:42 got a little Hatchet Hatchet in the
4:44 mountains we got the hatchet shirt we're
4:46 gonna be able to read up here it's going
4:48 to be fantastic
4:50 um let me tell you what's good this is
4:52 annoying so I guess there's there's good
4:54 news and bad news the good news is is
4:56 that the AI learning lab
4:58 um
4:59 has some followers the bad news is one
5:01 of those followers is a [ __ ] scammer
5:04 so the top one is me
5:07 the one that
5:09 is has two L's in it at the beginning is
5:12 the scammer and I've reported them now
5:17 three or four times in tick tock's just
5:19 like yeah whatever they're good
5:21 but they're actually asking people to
5:24 sign up for
5:25 um multi-level marketing and I would
5:28 never ask you to sign up for a
5:30 multi-level marketing program I would
5:32 ask you to sign up for a pyramid scheme
5:34 they're completely different
5:36 Okay so
5:39 so
5:40 if you subscribe to a channel you think
5:45 is this channel go check and make sure
5:47 that it doesn't have two L's at the
5:48 beginning of it
5:51 um
5:51 so that makes me cranky but anyway
5:55 that's not why we're here why we're here
5:58 is
6:00 um so
6:02 a little bit about my background I'm a
6:04 I'm an entrepreneur
6:06 um my current gig is I'm the founder and
6:09 CEO of a company called storyvine
6:12 that is an automated video production
6:14 platform that believe it or not does not
6:17 use a ton of AI we've been around for 11
6:19 years
6:20 figured out how to do that [ __ ] before
6:22 before AI came along
6:25 um we're rolling a lot of AI into it but
6:27 I I've been an entrepreneur for a few
6:29 decades and I've started 12 or 15
6:31 companies and I was I was a Founder in
6:34 the early days of the world wide web in
6:37 the mid 90s I started one of the first
6:38 digital agencies
6:40 a company Dr J I'm team Ponzi myself
6:43 that's awesome
6:45 I started a company called agency.com
6:48 which was one of the first digital
6:49 agencies and the reason I bring that up
6:51 is that what's going on with chat GPT
6:54 and these generative AI models the image
6:58 generation things and the video things
7:00 and the audio things and chat GPT and
7:02 all these competitors that I have listed
7:04 here it's very very similar to the early
7:07 days of the web so
7:09 um
7:10 so what we do on this channel is just
7:12 talk about it
7:14 um just you know if you've got questions
7:15 about it ask questions if you haven't
7:17 played with chat GPT
7:19 chat.openai.com that that URL right
7:22 there that's the official website just
7:24 go there start playing while you listen
7:25 to me ramble
7:27 um if you've been playing for a while
7:28 some of these other sites are are also
7:31 chat GPT or GP chat gbt competitors
7:35 I think they all are good enough that
7:38 they're at least worth taking a look at
7:40 see which of these is better for you uh
7:43 like for example Claude dot AI you can
7:47 you can paste in like a 50 000 word
7:49 novel it's got a hundred thousand token
7:51 context window so Claude can be really
7:54 useful when you might not be able to do
7:56 something on chat gbt Bing and Bard are
7:59 both connected to the internet Bing can
8:01 also generate images this is Microsoft's
8:03 version of gpt4 it can read images it
8:06 can make images it's connected to the
8:08 internet so it's kind of interesting
8:11 um
8:12 chat GPT the official one has code
8:15 interpreter which is quite powerful and
8:17 it's got um plug-ins and it's got this
8:19 thing called custom instructions and
8:21 gpt4 is just baller it's it's kind of
8:24 the best of of these things
8:26 um the perplexity one that's just uh
8:29 perplexity giving us access for free to
8:31 metas llama models uh Pi is from a
8:35 company called inflection posed from
8:37 quora all of them have the little
8:39 something to offer a little bit
8:40 different at some point I'm going to
8:41 start winnowing this list down but if
8:44 you're just getting started just start
8:47 at that top one and then the other thing
8:49 to do
8:50 is go to this thing prompts.chat so if
8:53 you're super new and you don't quite
8:55 know what prompting is or maybe you've
8:56 heard people talk about prompt
8:58 engineering or prompt crafting or maybe
9:01 you've never heard that term at all and
9:03 you're like what the hell is a prompt a
9:04 prompt is basically you're just chatting
9:06 you're just asking it to give you some
9:08 [ __ ] hey do this thing for me
9:11 prompts.chat will teach you how to do
9:13 that and it's it gives you a whole bunch
9:15 of ideas for ways that you can use these
9:18 large language models um this next one
9:20 down is the same thing for images
9:22 uh it's it's written for stable
9:24 diffusion but the basic principles here
9:27 apply for all of them and then
9:29 futurepedia.io is uh just a big long
9:31 list of tools there's like 4 000 AI
9:33 Tools in there you can geek out go nuts
9:36 just know that most of them are crap and
9:37 they're all probably trying to ask you
9:38 for 20 bucks a month so I would not sign
9:42 up for any annual concept tracks right
9:44 now everything's moving too fast all
9:45 right find a little piece of the puzzle
9:47 here
9:48 um if you're if you want to join a
9:51 community doing doing uh this sort of
9:53 thing uh the salon.ai that's that's a
9:56 group I started back in December called
9:58 the AI salon
9:59 and
10:01 um we have we meet every other week we
10:03 meet in person here in Denver we meet uh
10:06 you know virtually for those not in
10:08 Denver
10:09 um the next meeting is next Tuesday we
10:11 meet every other Tuesday and uh John
10:14 Knack from Adobe speaking from their AI
10:16 team he's he's uh
10:19 he's uh they're doing some really
10:21 interesting stuff over there and he
10:22 spoke at the salon last January and he's
10:25 going to be back Tuesday night so that's
10:26 exciting
10:28 um
10:29 what else
10:32 yeah just go to that URL and sign up for
10:35 that
10:36 um the uh YouTube channel up here AI
10:39 learning lab Dash TT Dash Tick Tock
10:41 that's where Archives of these lives are
10:44 going to be stored
10:46 and uh yeah I've started working on that
10:49 I I spent like an hour and a half today
10:52 trying to get the [ __ ] prompt to work
10:54 to generate chapters consistently and I
10:57 just couldn't do it and one of the
10:59 things that I learned about Claude dear
11:01 old Claude up here even though you can
11:03 put in giant documents it limits you to
11:06 like 20 or 25
11:08 um queries a day I probably if you're
11:10 using long documents
11:12 um which
11:14 I was trying to do some testing and so I
11:16 just burned through all my uses so that
11:18 made me cranky I was like hmm
11:25 welcome to bear what's happening how are
11:28 you doing all right the other thing I do
11:31 is is ask questions so so please feel
11:34 free to pop any questions in and I will
11:37 do my best to answer them let's see what
11:40 do you need to participate in the
11:42 Tuesday Salon camera microphone Etc you
11:45 can it's just a Google meet so so if you
11:47 go to the uh if you go to this URL
11:51 um you can you you know join the Meetup
11:54 Community or join the Meetup Group
11:57 and then there's also a Discord group so
11:59 you can you can go to the Discord group
12:00 and at the Discord group you can play
12:02 with mid-journey and chat gbt and you
12:05 can ask questions and things like that
12:06 so that's that uh storyvine's my company
12:08 and
12:10 that's what that is so you don't you
12:12 don't need to even be on camera you can
12:14 be you don't have to be for the for the
12:16 salon we do like to introduce ourselves
12:21 so it's kind of it's kind of like a
12:23 support group it's like it's like it's
12:26 like it's like aiaa
12:30 hi my name is Kyle I'm into AI
12:34 hi Kyle
12:37 uh and then what about these lives what
12:39 do we do here
12:40 um I ramble
12:45 um
12:47 I ramble I answer questions I get snarky
12:51 I get sassy I cuss occasionally so if
12:54 you don't like that sort of stuff
12:55 get the hell out
12:57 what is wrong with you swearing's fun
13:02 swearing's knee
13:04 I'm a large language model
13:07 and I don't have any safety guard rails
13:10 yes
13:11 okay let's do this let me get to some
13:15 questions
13:16 hello hello hello a lot of people saying
13:18 hello hello Ellie hello
13:22 all right
13:24 uh I used an uploaded photo of a vehicle
13:27 used it in mid-journey with prompts
13:30 prompt was a disaster not accurate at
13:33 all
13:34 thought
13:35 it would at least keep what was uploaded
13:37 yeah it doesn't it doesn't really do
13:40 that
13:42 um
13:43 I don't know how geeky you are Tobias
13:46 but if you um if you want to explore
13:48 that there's there's some new things
13:50 coming there was a paper that came out
13:51 like a month ago where they figured out
13:53 a way to essentially Auto trace a source
13:56 image and pull out basically capture all
14:00 the detail the the detailed Line work
14:02 and then you can modify everything else
14:06 um but I haven't seen that built into
14:07 anything yet
14:09 um the thing that I would play around
14:10 with is control net
14:14 um so control net is one way you could
14:16 do it but you could also try the remix
14:20 um
14:20 the remix capability in mid Journey
14:22 rather than just uploading a photo and
14:24 trying to do it
14:27 um
14:28 trying to think what else
14:37 yeah I don't know I I don't have a good
14:41 a good suggestion for you I was thinking
14:42 maybe you could use the describe
14:44 function in mid-journey have it describe
14:46 your Source image
14:47 and then use the source image as part of
14:50 the prompt and the description that it
14:53 selected and then change whatever you
14:55 want in that I don't know if that's
14:56 going to work I would I would look at
14:58 control net for stable diffusion
15:01 um if if you want to start to get some
15:03 things that are more predictable and
15:04 you're main you're maintaining the
15:06 composition and the look of the car
15:08 uh and then if not it just it might be
15:10 too early for that now the other thing
15:12 that you could do with an image like
15:14 that Tobias is
15:17 um if you've got Photoshop and you use
15:19 the Photoshop beta with generative fill
15:22 you could select the car and change out
15:25 all the background or you could select
15:26 the car and change the color things like
15:28 that so there might be there might be
15:30 there might be some ways you can do it
15:32 but it's it's that's that's a tough one
15:35 get it and getting things super
15:37 predictable in mid-journey and stable
15:39 diffusion right now is a [ __ ] show so
15:41 yeah welcome to the uh welcome to the
15:44 early days we sure are early here at the
15:47 internet
15:47 at the dawn of the generative AI era
15:51 it's the great Renaissance and right now
15:54 it's like a great pile of poo poo
15:57 so just
16:02 it'll get better
16:04 at some point
16:08 the predictability things are real pain
16:10 in the ass I saw there was a cool um I
16:13 saw some video of a uh or not video I
16:16 saw like a Twitter post
16:17 of a uh
16:21 I think it's open ai's new model so open
16:24 ai's Dolly the the like the image
16:26 generation thing within Bing is really
16:29 shitty it's just bad it's just doesn't
16:30 have a lot of controls it just gives you
16:32 square images it's not all that good
16:35 um but apparently they've got one now
16:36 that does texts really well like one
16:39 they have internally so I have a feeling
16:41 open AI is going to come up with with a
16:43 mid-journey competitor that's gonna
16:46 I have
16:53 the the Xmas tree is going to be piled
16:55 with
16:56 technology
17:00 I want a new large language model and I
17:04 want to admit journey to not suck I
17:07 don't want to do it in Discord
17:13 there's a lot in 10 sign up what
17:18 there's a lot in 10 sign up morgie I
17:22 don't know what you mean wait was that
17:23 in reference to something earlier
17:27 I don't know what that means I do not
17:29 know what that means I'm not that's uh
17:31 not my it's not my uh first uh the the
17:35 words it's not my first language
17:37 there's a lot intense on them they're
17:40 not now
17:42 AI learning lab is the real one yes
17:47 oh AI learning lab with two B's Joker
17:51 joking it up I like it it's solid it's
17:55 solid it's good okay
17:57 I think I asked last time but why didn't
17:59 you get to my question what do you think
18:01 about adobe's AI photo um I probably
18:03 just didn't get to your so what happens
18:05 Jennifer if you came in even an hour in
18:08 because I try to answer all of the
18:11 questions if you're down at the bottom
18:14 sort of late in the thing I get like a
18:16 hundred questions behind so you were
18:18 probably just
18:19 you would probably just swept up in the
18:22 uh Gen X are trying to use tick tock um
18:26 situation
18:28 um
18:29 I like generative fill a lot
18:33 So when you say when I when you say do I
18:36 what do I think of adobe's AI photo
18:38 there's there's a couple of things
18:40 there's Firefly
18:43 which is kind of their playground and
18:45 there they have
18:47 um
18:48 uh you know text to image and they've
18:51 got the the
18:53 stylized text thing and they've got one
18:55 other thing now then they've got
18:56 generative fill
18:58 um both of them are interesting I think
18:59 generative fill is is actually really
19:02 quite good so let's go play with that
19:07 do I have layers on this they do
19:10 okay well
19:12 let's go find something fun let's go
19:15 find a fun image oh you know what we'll
19:16 do
19:17 I know it'll do no what Kyle what what
19:20 are we gonna do you know what we're
19:21 gonna do
19:23 we're gonna go to
19:26 I'm now on the AI Salon we're in the
19:28 we're in Discord now so if you haven't
19:31 used Discord these are all servers so
19:33 these are all different communities this
19:35 is the AI salon and then these are all
19:37 conversations within the AI Salon
19:40 community
19:42 and I'm going to go to the mid Journey
19:43 Channel
19:45 and let's just go find a cool image that
19:48 one's neat
19:51 who's doing that Kyle WB
19:55 Kyle WB it's not me that's someone else
19:58 those are sweet those are neat
20:02 um
20:04 let's see
20:08 3D optical illusion
20:11 this is cool
20:19 wow nice that's nice too
20:22 oh that's a source image
20:25 yeah yeah he so so
20:28 um Tobias here's a guy that uploaded
20:31 that as a source image
20:35 and then he got he got this is outputs
20:37 which is pretty close
20:39 but it's not exact right like if you
20:41 want to keep the same lines of a car
20:43 like this has got completely different
20:45 Towers this has got an archway that this
20:47 uh well this has an archway
20:49 but it doesn't have that Tower
20:52 it's got a tower in the middle so yeah
20:54 so it it really does reimagine them
21:00 but it puts you in the neighborhood so
21:02 so yeah if you want to if you want it to
21:04 be specific okay
21:06 so let's see what do we want to do what
21:08 do we want a picture of Let's do
21:11 okay so the way you make
21:13 images with mid Journey you type slash
21:15 imagine when you're in a channel that's
21:18 got mid-journing installed
21:21 and then it it gives you a prompt it'll
21:23 say
21:24 um
21:25 let's see
21:28 [Music]
21:29 um
21:35 photograph
21:38 photograph
21:40 of a fly fisherman standing by a stream
21:47 in the fall with the leaves changing
21:50 color
21:56 and then I'm going to do dash dash AR
21:58 and I'm going to do two by three so this
22:01 is going to be taller than it is wide
22:05 and so now we'll have not now you know
22:08 mid-journing is going to go off and it's
22:10 going to make this beautimous
22:13 thing for us
22:21 and Jennifer
22:23 um
22:23 I don't know if you were here earlier I
22:25 think you were when I was talking about
22:26 the AI Salon my co-host in the AI Salon
22:29 is a woman named Leah fasten who's a
22:32 commercial photographer out of Boston
22:34 and so she's been playing a lot with
22:37 this AI stuff so you know it would be in
22:39 potentially interesting for you to come
22:42 to the salon just to talk to her
22:44 or you know meet her
22:48 all right what's it doing here it
22:50 doesn't look like the fly fisherman
22:51 that's okay
22:52 are these done
22:54 okay so the way mid Journey works it
22:56 gives you four options for any given
22:58 image and then if you find one you like
23:00 I kind of like this one it's a little
23:01 bit more dramatic
23:02 in the Stream with and he's off in the
23:04 distance a bit more
23:06 although this one's kind of Moody it's
23:08 got It's got some stuff here so with mid
23:11 Journey you've got you one through four
23:13 and V one through four that those
23:16 um are related to these four Images
23:18 obviously you means upscale and V means
23:21 very variations so
23:24 if you see one you like but it's not
23:26 quite perfect you can hit V4 so we'll do
23:29 that now and now it's going to render
23:31 four new images kind of based on that
23:33 one so it'll take the same seed of that
23:35 image and it's going to give us four
23:38 variations of it
23:40 and this this is often if I'm if I'm
23:43 putting together an image for an article
23:45 or something that I'm writing and I just
23:47 want to knock it out quick what I'll
23:49 often do is I'll type in
23:51 two or three different
23:54 um
23:54 spins on a prompt that are all in the
23:57 same neighborhood it'll generate four
23:59 four from each and then I'll just do
24:01 variations on anything I like that came
24:03 out of that
24:04 and then I'll kind of pick from those
24:05 and I'll just upscale them and then most
24:08 of the time within
24:11 I don't know two or three iterations I
24:13 can get something that that I like
24:18 um let's see this top one kind of looks
24:21 the most like
24:23 a fly fisherman
24:27 does it I don't know it doesn't really
24:29 matter okay let's grab
24:31 yeah let's grab this so so here's what
24:33 I'm going to do I'm going to hit upscale
24:35 number one
24:36 and so that's going to give us a big
24:38 version of this then we're going to hop
24:39 over to photoshop
24:49 [Music]
24:52 okay so there's our picture it's a
24:56 lovely picture
25:00 um
25:00 now I'm gonna go show you how to do this
25:03 in generative fill
25:05 but you can also do this in um
25:09 in mid-journey actually I'll show you
25:11 both why not
25:13 so
25:15 um
25:16 so we've got a vertical image here now
25:18 let's say I I was doing a piece and I
25:20 actually need this to be horizontal and
25:22 I you know I want to see this this
25:24 dude fishing but I need I need a
25:27 horizontal image well I can crop it but
25:29 you know then I'd kind of lose the the
25:31 beauty of the scene so
25:33 um once you up upscale an image it gives
25:37 you a bunch of options here you can do
25:40 variations strong and subtle you can
25:42 zoom out two times which will basically
25:44 like put him you know in the distance 2x
25:47 right you'll get two times more stuff
25:49 around him you can do a custom Zoom
25:51 where you could turn it into a 16x9
25:53 image but they also have these these
25:55 panning things so what if we wanted to
25:59 create an image where we're going to put
26:00 some copy on the page so we want the
26:03 fisherman to the left and we want you
26:05 know some stuff out to the right well we
26:07 can actually do that we just hit this
26:08 right button here
26:10 and what that's going to do is it's
26:11 going to add to the right of this image
26:13 so this is similar to generative fill so
26:15 we'll see what this does here and it's
26:17 going to give us four options here
26:20 and then we'll go do the same thing in
26:22 generative fill so let me
26:24 pop this up open it in a browser
26:29 copy this image go over to photoshop say
26:33 new
26:36 paste all right so there's where we'll
26:39 start there so let's go back to
26:41 mid Journey
26:44 hey mid Journey how you doing tonight
26:46 huh I'm fine how are you uh I'm all
26:50 right mid Journey
26:52 oh mid Journey you're the best
26:58 except for the fact it's still in
27:00 Discord
27:01 am I right
27:03 get get it go and make a site make a
27:06 site where we can go make pictures in a
27:08 pretty website I don't have to go to
27:10 Discord I hang out with all the nft Bros
27:14 and they make me crazy
27:17 all right so let's see
27:23 so how did this do I can't quite see the
27:25 edges of these so let's just grab one of
27:27 these that looks decent they're all kind
27:29 of the same
27:30 these have this cool little waterfall
27:33 off to the right
27:36 so let's go grab we'll go grab four
27:38 upscale number four
27:45 I think I want to purchase the paid
27:48 version but I'm not sure if it's worth
27:49 it oh for for for Adobe
27:53 um if you're if you're using so so this
27:56 is still Jennifer
27:58 um
27:59 I so so if you're using
28:04 it for clients and if you're using it if
28:07 you work within a business
28:09 um
28:12 well it sounds like you're a
28:13 photographer so so you're I assume
28:16 you're freelancing you you're probably
28:17 using this for clients here's here's
28:19 here's one of the big reasons to use
28:21 Adobe over mid-journey is Adobe hasn't
28:25 been trained on
28:27 um copywritten content copywritten
28:30 images it's been trained on public
28:32 domain images and Adobe stock images
28:35 that they've got access to Oh all three
28:38 of these added fishermen in there that's
28:40 funny
28:42 all right well we'll upscale
28:45 number four actually that's kind of
28:47 interesting we could we could uh
28:49 we could pop that fisherman out of there
28:52 um
28:54 but there's what it did so here's what
28:55 we started with and then it added all
28:57 this stuff to the right so so that's
28:59 that
29:00 um
29:01 and then over in Photoshop it's it's
29:04 kind of the same thing so I'm in the
29:06 Photoshop beta and so we've got this
29:08 vertical image and I can do the same
29:11 thing here where
29:13 I grab the crop tool
29:16 and I can say I want a little bit to the
29:19 left and I want a lot to the right
29:23 and so now I have you know that's kind
29:27 of the image area that I want so what
29:29 I'm going to do is I'm going to select
29:34 and I just I heard what you said that
29:36 you do this all the time so
29:38 we'll do this for other folks
29:42 let me select this
29:46 so what I'm doing now is I'm selecting
29:48 the image and now I'm going to invert
29:49 that selection
29:55 and what happens in in the in the beta
29:59 is that you've got this this contextual
30:01 menu that pops up when you select
30:03 something and so now I'm going to say
30:04 generative fill
30:07 and I can put a prompt in here so I
30:10 could say you know that I want you know
30:12 just more stream and more Hills or I can
30:14 just let it try to figure it out this is
30:16 a pretty clear image so it tends to do
30:19 good with this kind of this kind of
30:20 imagery so we'll see what it does
30:23 I'm suspecting that photo Photoshop will
30:26 not put an additional dude in there
30:29 where mid-journey did
30:33 um that's not quite right because it
30:36 gave us a whole second stream it's like
30:38 he's below an island right here
30:41 um that's one option
30:43 there's another option
30:46 that's closer
30:49 but they're all kind of bad
30:51 all right so we'll try another one we'll
30:54 do
30:57 um
31:15 select inverse
31:18 and then we're gonna do generative fill
31:19 and we'll do
31:21 um
31:22 let's see
31:24 um steep Hillside
31:29 and we'll generate that we'll see if we
31:31 can get it to to give us a steep
31:32 Hillside instead of a second stream
31:35 foreign
31:42 and I will get back to your question oh
31:44 that's kind of cool that's much better
31:48 um that's even better that's even more
31:50 better right we got this is where the
31:52 stream used to come down through here
31:53 and I've got these big round boulders
31:56 so you get you get three different
31:58 options that one still looks like a
31:59 second river all right that that's the
32:01 one right so that's the one and and you
32:03 know what I might do is I might do
32:04 something like
32:06 um
32:07 like that's too bright if I want to put
32:09 text over that that's too bright so I'm
32:11 going to grab that I'm going to
32:12 highlight that selection and I'm going
32:15 to say
32:16 um dark Moody
32:19 um fall
32:22 uh trees or something like that
32:27 and hopefully that'll darken up that
32:29 section and then we could use that for
32:32 our our little ad campaign for our
32:35 backpacks yeah perfect
32:37 right so now we have a nice dark area
32:39 over there
32:42 that's a little too busy that's a little
32:44 too busy so the first one's good so
32:46 there you go so so in this case
32:50 um Photoshop in my opinion did this
32:52 better than mid Journey did it right it
32:55 filled it out nicely here it extended it
32:57 nicely there and with a little bit of
32:59 prompt prompt magic we did something in
33:02 there
33:04 um
33:05 let's try something fun
33:07 um
33:12 I'm gonna say generative fill
33:15 um
33:16 black bear eating
33:21 a trout
33:23 let's see see if we can get a a bear in
33:26 the image yes two they're too dramatic
33:29 but
33:30 anyway that's funny all right we'll get
33:33 rid of that all right that's generated
33:35 Phil so back to your question
33:37 um Jennifer
33:40 uh you know
33:43 photoshopping cheap but the the
33:46 generative fill stuff and the access to
33:48 Adobe Firefly
33:52 um because that stuff is
33:55 um is is commercially viable like the
33:58 fear I have about mid Journey even
33:59 though it's beautiful is because it's
34:01 been trained on
34:03 um other people's work especially if you
34:05 use other people's artist names in your
34:08 prompts
34:09 um
34:10 it could that might be problematic like
34:12 we don't know where the copyright and
34:14 all the lawsuit stuff is going to land
34:17 um so so Adobe is really safe for that
34:19 and and you know they're Adobe they're
34:21 gonna get their [ __ ] together the fact
34:23 that they're leaning into this so hard
34:24 so when when John Knack spoke at the AI
34:27 Salon back in January they were not
34:30 clear that they were going to go into
34:31 this world because back in those early
34:33 days the artists were pissed off they
34:36 were not happy about this [ __ ] at all
34:38 um so when they went into it they did it
34:40 you know training training on
34:43 um images that didn't violate anyone's
34:45 copyrights which is pretty smart so yeah
34:48 I think I think they're worth uh worth
34:50 playing all right I'm learning ux uh
34:54 oh uh relum is making me pretty sad yeah
34:58 I know Timothy Brooks so here's the deal
35:00 everybody everyone on the uh on yield uh
35:04 learning lab on the live tonight this is
35:07 this is part of the [ __ ] we're all gonna
35:09 have to confront like I've been
35:11 confronting it with my video business
35:13 for a while now
35:15 that
35:18 anyone in the knowledge or creative or
35:22 you know technology World any any jobs
35:25 that are computer mediated and a lot of
35:28 them even that are tangential to that
35:33 all of our jobs are going to be
35:35 completely
35:38 redefined
35:40 and some of them
35:42 some of what they become we might not
35:45 like and some of what they become might
35:48 mean
35:49 companies don't need as many of us for
35:52 that particular job like they're still
35:53 going to be writers
35:55 but there might not be as many of them
35:57 and what those Writers Do might not be
36:00 what they do now and same with ux
36:02 designers and same with engineers and
36:04 coders and consultants and like I don't
36:09 think anyone's exempt if you touch a
36:11 keyboard
36:14 you know if you touch a computer in any
36:16 way if you work in a call center
36:21 everything's going to be redefined and
36:24 you know a lot of the the more tactical
36:27 execution kind of jobs
36:29 and a lot of the jobs that are about
36:31 managing complex systems like like that
36:35 reloom site that automatically makes a
36:39 site map and then wireframes of the
36:42 websites just continue that out so so
36:45 it'll say here's your site map here's
36:48 your wireframes here's your designs
36:51 and then those are all tied to
36:54 um
36:55 you know code on the back end right so
36:58 so someone's gonna Stitch that entire
37:00 workflow together
37:02 and So within an hour you'll be able to
37:04 create a very sophisticated website
37:07 effectively from a single prompt but
37:10 here's the thing
37:13 a good web designer a good ux person
37:16 even if that stuff is auto-generated is
37:18 going to be able to look at what it
37:20 generates and go oh that's crap oh
37:23 there's a better way we could do this
37:24 let's modify this thing and let's let's
37:26 move these things around on the site map
37:28 because they don't make as much sense
37:29 right so there's going to be just like
37:32 if you've ever used chat CPT
37:35 you can get it to 80 in like five
37:38 seconds getting it to a hundred percent
37:40 is almost impossible
37:42 it's going to be the same thing with
37:44 things like relum or whatever the tools
37:46 are where you know
37:49 you're you're a musician and you're like
37:51 I love handcrafting music and then Along
37:53 come this comes this tool it's just like
37:55 squirt you know here's here here's your
37:59 exact style you know
38:01 song after song after song every three
38:03 minutes well you're going to be able to
38:05 use that tool in a way that a
38:06 non-musician whoa
38:09 so I think there's going to become
38:11 superstars of figuring out the new way
38:15 to do the old thing right and then
38:18 there's going to be a lot of people that
38:19 are just like I don't want to use AI
38:22 that's not what I got in this for and
38:24 they might not just be whining like that
38:26 like they might be really sad because
38:29 because
38:31 you know there was a there was a uh a
38:34 post on Reddit
38:35 three or four months ago
38:38 about a 3D artist he worked for a
38:40 10-person company and there were two
38:44 3D artists in his company there was him
38:47 and there was a guy that wasn't as good
38:49 as him
38:51 and then this AI stuff came along and he
38:54 said he went from being a 3D artist to
38:56 being a prompter
38:58 and he didn't like the prompting he
39:00 liked building models thank you thank
39:02 you for the heart
39:05 and the guy that wasn't as good as him
39:08 loved the prompting and saw the
39:10 opportunity with the boss and saw a tool
39:14 that could make him better
39:16 and so the guy that wasn't as good
39:19 jumped on this [ __ ] became better than
39:22 the guy that was good
39:24 or I don't know if betters right more
39:26 productive
39:27 and more in the boss's favor right
39:31 because they were getting [ __ ] done
39:33 faster like instead of it taking two
39:35 months to build a model it was taking
39:37 him a day and a half
39:40 and the post that he wrote was just this
39:43 Lament
39:44 for
39:46 his job evaporated overnight and he was
39:50 just [ __ ] sad and he quit like he
39:52 quit the company
39:53 he's like I don't want to do like I
39:55 don't want to be a prompter
39:57 that's not my thing
39:59 and so I think
40:01 every every one of us is going to
40:03 confront this in some form or another
40:08 and it's you know there's a piece of it
40:10 that's really sad and then
40:14 wait for it
40:18 I don't know how long the painful
40:19 period's gonna be my guess is two to
40:21 three years it could be five years
40:23 but on the other end of the painful
40:25 period where there's all this just
40:28 massive disruption and and and and
40:32 facing all of these issues
40:35 these tools are going to create new jobs
40:38 that we can't even imagine today that
40:40 are they're going to be [ __ ] amazing
40:42 right because they're just going to be
40:44 at such a higher level and they're
40:47 probably going to be multimodal like I I
40:49 feel like people that get good with
40:50 these tools are going to be kind of like
40:52 Symphony conductors where you've got all
40:53 these tools and and you're kind of Oz
40:56 behind the curtain and you've just got
40:58 this panel and you're like I've got a
41:00 vision for a company and a website and
41:03 software and a marketing plan and you
41:06 know let me take a week and launch that
41:08 company
41:10 like that's gonna there's gonna be
41:12 people that are like that right they're
41:13 just gonna be masterful at this [ __ ]
41:15 who's who's that gonna be it's going to
41:18 be someone that right now goes instead
41:20 of just
41:22 dealing with the sadness of losing what
41:25 was lean into it and get curious about
41:28 it and go oh what could be
41:31 right now is the time
41:34 that if you
41:36 lean into what could be with this stuff
41:39 get curious about it think of it like an
41:42 adventure
41:44 you know when you choose to go on an
41:46 adventure
41:48 you choose to go on it even if you're
41:51 scared even if you know there's going to
41:53 be shitty things
41:55 even if you don't have the confidence
41:56 that you can go
41:58 slay the dragon or find the treasure or
42:00 whatever the [ __ ] your adventure is
42:04 but you do it anyway
42:06 because there's a higher sense of
42:09 something
42:10 and where my head is right now is
42:14 this shit's coming whether we like it or
42:16 not
42:20 just it just is it's not going to get
42:21 uninvented and it's not going to get
42:23 worse these tools today are the worst
42:27 they're ever going to be
42:32 anyway so yeah I know I'm with you
42:35 Timothy Brooks I get it
42:37 it's
42:39 you know it's funny I was I don't know
42:42 who I was talking to the other day
42:45 I forget who it was but but I was
42:47 talking about we were talking about
42:49 covid and lockdown and working remotely
42:52 and and I was talking about
42:54 isn't it ironic that that Zoom of all
42:57 companies just said that they're you
42:59 know they're going to ask their
43:00 employees to come back to the office now
43:02 it's only two days a week and optional
43:04 Fridays
43:06 thanks Lauren um
43:09 but
43:14 I you know I feel like I'm I'm in a bit
43:16 of mourning for my own company my own
43:18 company storybine where
43:20 we're a small company but we we have
43:23 a lot of introverts and we have me
43:26 and I'm not
43:28 and so I get my energy from being around
43:30 other people and the other people in my
43:33 company don't and so we went into
43:34 lockdown they were [ __ ] they were
43:36 living their best life they're like this
43:38 is awesome I don't need to deal with
43:39 people and I can get [ __ ] done
43:42 and you know what it's what it's ended
43:44 up doing for me as a founder of the
43:47 company and and as someone who who
43:51 I don't know I like there's something
43:53 about Humanity that feeds me right
43:56 um I feel like my company in a lot of
43:59 ways has become a to-do list with zoom
44:01 check-ins
44:04 and that really bums me out and so I've
44:07 been kind of mourning that for a while
44:08 and then I think this stuff on top of
44:11 that is is kind of a second kind
44:14 kind of
44:16 um
44:17 you know death change that we're all
44:19 going to have to deal with
44:21 and
44:22 you know so so we're still trying to
44:25 figure out what work looks like you know
44:27 now that now that covet has sort of you
44:30 know
44:30 blown through Society blown through the
44:32 global Workforce and and done whatever
44:35 it's done
44:37 um
44:38 and and this is going to be the next
44:40 sort of wave that's going to change a
44:42 lot it's going to change
44:44 probably more than covid did
44:48 um
44:48 sorry I'm adjusting my little stand here
44:52 um
44:53 it's gonna it's gonna it's gonna make
44:55 some messes
45:00 all right
45:01 for a physician that wants to pursue a
45:03 career in generative AI with no coding
45:05 experience what next
45:07 um
45:12 there's so many possibilities so I would
45:14 get I would get really educated there's
45:17 there's um
45:18 Google just came out with a large
45:20 language model called medpom
45:24 med-p-a-l-m or Med Palm two I don't
45:26 whatever is the latest Med Palm that
45:28 Google came out with that's a large
45:30 language model that's trained
45:32 specifically on
45:34 you know medical stuff and and you know
45:36 medical data and it's multimodal so it
45:39 can do imagery and things like that I
45:41 would get educated on on
45:44 the tools that are out there that are
45:46 that are specific to to that category
45:51 um
45:54 I'm trying to think I just heard today
45:57 someone just came out with a thing I
46:00 think it was on the the AI marketing the
46:02 marketing AI institute's podcast this
46:04 week they were talking about
46:06 um I forget who the company was that
46:08 came out with it but they used the
46:09 medpom large language model and created
46:12 a thing I think called Palmetto let me
46:14 see if that's right
46:15 I don't want to run you down a rabbit
46:17 hole
46:19 um Palmetto AI
46:24 um
46:36 let's see Medical
46:50 page now found that's not right
47:02 nah I don't I can't remember what it is
47:04 so anyway so I would just I would start
47:06 paying attention to
47:09 um anything anything medical related
47:10 that's going on with AI
47:12 um you don't need to know coding um I
47:15 would start to I would start exploring
47:19 um things like pie so so pi
47:23 dot AI if if you go to the App Store and
47:25 you look for pi personal assistant
47:27 you'll see the app for pi or you can go
47:29 to a website Pi dot AI
47:32 that's from a company called inflection
47:35 and it's a it's a chat bot that's very
47:37 chatty it is not super it's not
47:41 efficient like if you want to go do work
47:42 you know use these top things but if you
47:44 want to go have a conversation with a
47:46 large language model
47:48 and you want to start get a glimpse into
47:50 the future of what it might be like to
47:53 interact with these virtual
47:56 um
47:56 agents or or assistants
48:00 it's a really good thing to experiment
48:03 with so I I would download the iOS
48:05 version of that app you can put it into
48:07 speech mode and just start talking to it
48:10 and start asking it questions I've got
48:12 kids with chronic lyme disease and so I
48:15 was asking it about Lyme disease and
48:17 treatments and things like that it knew
48:19 all the stuff it knew about babesia and
48:21 Bartonella and just just all the stuff
48:23 and so and that's not even been
48:26 optimized with something like Med Palm
48:28 underneath it so I I would think that as
48:31 as a physician
48:33 you're going to be in a in a
48:37 a powerful position to understand what
48:40 worked in the old model and what didn't
48:44 and be able to see opportunities to make
48:47 things better so make things better
48:49 might look like better bedside manner it
48:52 might look like extending
48:55 um
48:56 caregivers and hcps reach
49:00 so imagine a physical therapist you know
49:03 a physical therapist can only see so
49:05 many patients in a week and you know the
49:09 the patients they have might not be able
49:12 to get in touch with them if they have
49:13 questions and things like that imagine
49:15 that physical therapist being able to
49:17 say
49:18 you know here's your plan for the week
49:21 and by the way here's the virtual me
49:24 that you've got 24 7 access to that you
49:27 can ask questions to it knows all of
49:29 your record it knows all of what we've
49:31 talked about so far and if you've got
49:35 anything going on that requires my
49:36 attention
49:38 the virtual me is smart enough to know
49:40 that and we'll forward you know your
49:42 message to me and the next time we talk
49:44 I'll know what you talked about right
49:46 and that you know something like that
49:48 could be there
49:50 um
49:50 these things are going to be amazing for
49:53 simplifying complicated
49:56 systems I know there's a lot of
49:58 companies working on that but you know
50:00 you could you could find a startup
50:02 that's working on simplifying you know
50:05 patient records and and the amount of
50:07 time that nurses and doctors spend doing
50:10 [ __ ] paperwork so the insurance
50:12 companies will refuse to pay them
50:15 not that I'm bitter
50:18 but I'm [ __ ] bitter that I think a
50:21 lot of doctors are too it's why there's
50:23 so much goddamn burnout
50:27 so here's here's the thing about this AI
50:29 stuff
50:31 is
50:33 these are just tools
50:35 these are not the solution these are
50:37 just tools the solution is you
50:40 with your expertise saying
50:42 here's a problem that needs solved
50:46 I the old way I would have solved that
50:48 is X Y Z
50:50 let me go learn enough about what these
50:52 things are and how they work and what
50:54 they're good at and what they're not
50:55 good at
50:57 that I might be able to take some small
50:59 piece of this problem
51:01 and experiment with solving it with
51:03 these tools
51:06 and then what you might find is holy
51:08 [ __ ] it's amazing at that or what you
51:10 might find is is oh it's not really
51:11 ready for prime time and you know it's
51:13 it's probably somewhere in between the
51:15 two
51:18 um
51:20 and so I wouldn't don't worry if you're
51:22 not coder who gives a [ __ ] we can we can
51:24 find coders and and if you really want
51:27 to learn coding this [ __ ] will teach you
51:29 coding and it'll do a lot of it for you
51:31 if you're geeky and like geeking out on
51:33 [ __ ] great it'll teach you how to do
51:35 that you can go do that
51:36 if you have no interest in that great
51:38 have no interest in that go solve the
51:40 problem
51:41 right take your your knowledge as in
51:44 working with the system and and do that
51:46 with it
51:48 all right
51:49 hope that helps
51:52 what do you need to participate in
51:54 Tuesday's Salon uh just uh it's it's a
51:58 it's a Google meet you just need a web
52:00 browser that can do Google meet so it's
52:06 you need a camera if you want to be on
52:08 camera you need a microphone if you want
52:09 to talk but there's even a dial-in
52:11 number you can just dial in and listen
52:13 Claude cut me off oh wait hang on oh oh
52:16 Tick Tock just bounced me to the bottom
52:19 oh tick tock you make me crazy
52:23 happy Saturday night everybody looks
52:25 like we've got some new folks here my
52:26 name is Kyle Shannon this is the AI
52:28 learning lab we're learning about labs
52:31 no
52:33 um also I'm gonna pop up here so uh
52:37 we're popular enough now that we have
52:39 been officially
52:42 um
52:43 cloned by a [ __ ] scammer
52:47 so if you get any messages from AI
52:50 2l's learning lab that's a scammer
52:54 the word lab only has one L in it and
52:57 because it's i l you know and and it's
53:01 you know if you have lowercase L's it's
53:03 like ill looks very similar to Il
53:06 so just be careful the scammer is a i l
53:10 l learning lab
53:12 and then or I mean learning only has one
53:15 L in it so so just be careful of that I
53:17 I wish that wouldn't have happened and
53:19 tick tock's not dealing with it right
53:21 now so I don't think anyone other than
53:24 me is allowed to report them for that
53:26 because tick tock
53:29 um I've reported it now four times and
53:30 they're like we don't see anything wrong
53:32 with it
53:35 well they've duped all my [ __ ]
53:36 content and they're telling all my
53:38 people to sign up for a multi-level
53:40 marketing piece of [ __ ]
53:43 all right
53:45 welcome
53:48 if you have questions about AI popping
53:50 below
53:52 um
53:53 uh what was I gonna say I tried to
53:55 answer them all so I get very very far
53:57 behind on these so if if I haven't
53:59 gotten to your question or if I haven't
54:01 in previous things don't be upset with
54:03 me don't hate me don't hate me because
54:06 I'm beautiful
54:07 don't do what's good about the hair
54:08 tonight I haven't even paid attention to
54:10 it it's all right
54:12 it's not bad tonight it's not it's not a
54:14 terrifying
54:25 the bear hey everybody hey Mr cgo what's
54:28 happening the bear is referencing my
54:31 title I have a title of Chief generative
54:33 officer for an organization I do some
54:36 work with which I think is the first
54:38 Chief generative officer in the history
54:40 of the world ah take that and then you
54:43 might ask what is a chief generative
54:45 officer
54:47 fantastic question Bob tell him what
54:50 he's won you can make money watch every
54:52 day
54:55 uh about um Claude cut me off today too
54:58 yeah that bugged me because it was like
54:59 you've got five questions left you've
55:01 got four questions left and I'm like
55:03 just don't [ __ ] up my question and I
55:06 won't keep doing this thank you the bear
55:09 that's my dangly lights I like the
55:11 dangly lights whoa the bear reached
55:13 member level 11.
55:16 the bear just turned it up to 11. holy
55:19 [ __ ]
55:20 I have no idea what that means but I
55:22 like it
55:27 we got some spinal tap going on in here
55:30 that's unusual I locked I talked to
55:32 Claude all day at work I hope it's not a
55:34 new thing so cuppy I think what's going
55:36 on with Claude I'm taking the
55:39 transcripts from these lives which are
55:42 like the one I did the the one I was
55:44 experimenting with was a three hour and
55:46 18 minute live so it was a transcript
55:49 from three hours and 18 minutes of me
55:51 just rambling
55:53 um so so it's a really long input so I
55:55 have a feeling it's number of tokens so
55:57 if you're just doing short back and
55:59 forth with Claude it's probably not it's
56:01 not gonna rate limit you and and I'm
56:03 trying to do like I'm what I'm trying to
56:06 do is I'm
56:08 this is I can I can show you what I
56:10 my prompt is insane for this thing what
56:12 I'm trying to do is have it read a three
56:14 hour transcript
56:16 and then
56:18 on average pull out six interesting
56:22 things I said per hour as chapters
56:25 and create a YouTube title A YouTube
56:27 description and chapters with time code
56:30 on them
56:31 and like it'll do it perfect and I'm
56:35 like awesome I got it and then the next
56:37 one I do
56:39 it it'll do like you know six quotes
56:43 from the first 15 minutes of the thing
56:45 and ignore the rest of the transcript
56:47 like ah and then I try to change that up
56:49 and then it just messes something else
56:51 up and it just so it's just and and so
56:54 where I ended up getting was I said okay
56:56 let me just generate the chapters first
56:58 and not worry about the title in the
57:00 description and so I was sort of halfway
57:02 through getting the chapters done and
57:04 that's when Claude said go screw
57:05 yourself and then I went to
57:08 code interpreter because I thought oh I
57:11 know I'll pop it into code interpreter
57:13 in chat GPT and it can read the long
57:16 file and it's gpt4 and it's going to be
57:18 awesome and then it just kept writing
57:21 python code and it was like I can't
57:23 understand the time code and this is in
57:25 RTF and then I uploaded it as a text
57:28 file and then it could sort of read it
57:30 but it could sort of do the math and
57:32 sort it like I don't know it's just
57:33 sometimes these things are you make it a
57:36 little crazy
57:40 but here's the thing
57:44 when I solve that like like going
57:47 through the the the nightmare of
57:50 learning how to use these tools right
57:52 now
57:53 like what it's what it's training me to
57:55 do is just think very very critically
57:59 about how you use words to
58:05 pull interesting magic out of these
58:07 things
58:09 I don't think that's going to be super
58:11 critical long term I think the machines
58:13 are going to start prompting themselves
58:14 and get really good but in the in the
58:17 short term it's it's really important
58:19 and I you know it's also it's also fun
58:22 and really gratifying when you get it
58:23 right because here's the deal when you
58:25 get those prompts right and you can sort
58:27 of lock that [ __ ] down and then you can
58:29 just start popping in long form YouTube
58:32 transcripts and getting out really good
58:34 answers like I can go back and
58:36 you know do a whole bunch of content
58:38 generation which is really exciting all
58:41 right you will take on the Persona of an
58:43 I.T entrepreneur that is first enough in
58:45 AI to teach tick tockers that's what
58:48 I've done
58:49 you know it's funny it's like like I
58:52 have a sort of I have a choice with this
58:54 channel which like this is a really
58:56 interesting kind of dilemma
58:58 as we build community here
59:01 some of the people that have been here a
59:03 lot and come a lot if I
59:07 if I just keep with the beginner stuff
59:09 then it it kind of loses its value and
59:12 and then but if I make it to advance the
59:15 new people will come and they'll be like
59:16 I don't know what the [ __ ] he's talking
59:17 about and so I'm trying to sort of skate
59:20 between those two worlds what what's
59:24 why I think it's okay to keep focusing
59:27 on the relatively like what if you've
59:28 done nothing with this go start with
59:30 chat EPT here's the tools is this
59:32 stuff's changing so fast that three
59:35 months from now if you check in with me
59:36 even though I'm still doing the
59:38 beginning kind of stuff I will likely be
59:39 talking about different tools different
59:41 techniques different challenges
59:43 different news items that are like you
59:46 know holy [ __ ] this company just laid
59:48 off 50 of their staff or whatever it
59:50 might be and and you know so so I think
59:52 there's I think there's lots for us to
59:54 to um
59:56 to talk about even if even if this
59:59 doesn't end up diving deep and and it
1:00:02 number of folks have asked me you know
1:00:03 do I have a workshop have I thought
1:00:04 about doing a workshop I think I might I
1:00:06 might do a workshop at some point I've
1:00:08 kind of mentally made that decision I
1:00:11 just haven't quite decided what I want
1:00:12 it to be so if I do that I'll let you
1:00:13 all know obviously
1:00:16 all right where do I find those
1:00:17 information I don't know which
1:00:19 information you're talking about any
1:00:20 cinnamon in that coffee no this is
1:00:24 um cinnamon free but I do have cinnamon
1:00:25 here but I'm not putting it in there
1:00:27 because it made a mess last night I had
1:00:29 like a cinnamon mustache it was
1:00:31 it's not good
1:00:34 I don't have enough milk in that coffee
1:00:37 Chris what's happening car Bruno
1:00:41 I love seeing people coming back
1:00:44 this is really fun
1:00:45 um I need an AI assistant that helped me
1:00:47 remember things disabled veteran
1:00:49 requesting a suggestion
1:00:52 um
1:01:00 there's not a there's not a good
1:01:02 short-term answer right now ah crap
1:01:05 there was there there is
1:01:08 the [ __ ] was it called
1:01:15 um
1:01:17 first of all thank you for your service
1:01:20 um there is
1:01:25 if you just hang on for another I don't
1:01:28 know three or six months
1:01:30 there's gonna be tools coming out that
1:01:32 do this so when Apple releases their
1:01:36 um
1:01:37 their AI I have a feeling it's going to
1:01:39 have memory
1:01:41 um I assume at some point open AI is
1:01:43 going to do it
1:01:45 um
1:01:45 I definitely think that Pi's gonna do it
1:01:48 the inflection company here
1:01:50 there's a there's I I there's a company
1:01:53 called or there's a project called Lang
1:01:56 chain
1:01:57 where you can build these large language
1:01:59 models so that they have memory
1:02:01 so and I I know that there's a project
1:02:04 out there hang on let me see if I can
1:02:05 find it because
1:02:10 wait I got an email about it
1:02:14 hang on hang on hanging
1:02:18 let me see let me see well let's just go
1:02:21 look shall we in that special aren't you
1:02:24 just terrific
1:02:26 all right um pop some pie
1:02:31 it's a voyager Runway
1:02:38 let me see AI memory
1:02:45 recollect
1:02:54 okay I found it
1:03:03 the caveat here is I don't know if this
1:03:05 is any good but
1:03:09 try read Dash collect dot AI
1:03:14 and let's go look at the site hang on
1:03:17 read Dash collect
1:03:20 dot a i
1:03:26 accept invitation so you you have to
1:03:29 apply for the wait list
1:03:32 um
1:04:02 oh it already exists I already have an
1:04:04 account
1:04:06 I don't know if this is any good but I
1:04:09 think this is what you're talking about
1:04:12 start
1:04:13 okay let's watch this together the
1:04:15 founder and CEO of Records I'm really
1:04:17 excited today
1:04:19 once you've completed your onboarding a
1:04:20 hobby extension installed then I
1:04:21 recommend you pin it
1:04:23 as you browse online will automatically
1:04:24 bring in most articles for you and you
1:04:26 can intentionally bring in things like
1:04:27 PDFs as you read you can also do things
1:04:30 like highlight information right click
1:04:32 here and click our extension and that
1:04:34 will bring in those highlights for you
1:04:35 in this article and you can also add
1:04:37 notes if you'd like if not simply click
1:04:39 off
1:04:40 the extension here if I click it will
1:04:42 show me that this article has come into
1:04:43 my account with this check mark nah this
1:04:45 doesn't seem right this doesn't seem
1:04:46 like what you're asking for
1:04:48 um
1:04:49 I don't know keep keep checking back
1:04:53 here I
1:04:54 there so the the challenge that you're
1:04:57 facing my my dad had Alzheimer's was was
1:04:59 an Air Force vet
1:05:02 um
1:05:02 the the challenge that you're facing
1:05:04 this AI stuff is going to be brilliant
1:05:06 for we just haven't quite gotten to the
1:05:09 point that the memory piece of these AI
1:05:13 tools is is not there right now so like
1:05:15 within a within a
1:05:18 um chat GPT session it'll have memory of
1:05:20 that conversation up to a certain point
1:05:22 but but you know it doesn't have memory
1:05:25 across them
1:05:27 um
1:05:27 that stuff's coming it's it's it's
1:05:30 coming it's just not there right now so
1:05:32 I'd say just be a little patient with it
1:05:33 I'll keep an eye out for it because
1:05:35 because I think it's a I think it's a
1:05:37 fascinating Challenge and I think it's
1:05:39 going to be so [ __ ] empowering it's
1:05:42 going to be like a super power
1:05:44 um for anyone you know my wife had a
1:05:46 brain injury so anyone who's got you
1:05:49 know memory issues and challenges uh
1:05:52 with that stuff I think this stuff's
1:05:54 gonna be
1:05:55 incredible for that the the tools that I
1:05:59 know of are not there yet
1:06:01 but they'll you know they will be here
1:06:04 at some point
1:06:06 um I like making chat gbt pretend to be
1:06:09 Shang Shang Tzu from Mortal Kombat and
1:06:12 say his catchphrases oh yeah it's good
1:06:14 for that it's it's great for party
1:06:16 tricks actually pie is really good at
1:06:18 that
1:06:18 um what did I ask pie I said to pie
1:06:21 I said I wanted to learn something and I
1:06:23 wanted it to teach me something by
1:06:25 inventing a game and it came up it
1:06:27 invented like a trivia game and then it
1:06:29 started asking me questions and it was
1:06:31 actually really interesting and and you
1:06:34 know that that starts to point to to a
1:06:36 future of
1:06:38 hyper personalization where it's like
1:06:40 I'm not just gonna like learn if I want
1:06:43 to learn a new language you know I could
1:06:45 go buy the off the shelf you know
1:06:47 language learning system that someone
1:06:49 invented but I could also just say hey I
1:06:53 want to learn Spanish and I want to I
1:06:55 want to learn it you know talking about
1:06:59 um I don't know you know fluffy kittens
1:07:02 whatever the hell is you know with you
1:07:05 know cultural movie references and and I
1:07:08 want you to you know turn every lesson
1:07:10 into a game and it'll just start doing
1:07:12 that and you know going back to that
1:07:14 memory thing
1:07:16 these interactions
1:07:18 are going to start being recorded So
1:07:21 these agents that we interact with are
1:07:23 going to remember what we talked about
1:07:25 yesterday and three months ago and six
1:07:28 months ago and a year ago
1:07:30 and you'll be able to say things like
1:07:32 hey I feel like I've kind of stagnated
1:07:34 am I am I really learning anything new
1:07:36 you know tell me where I was three
1:07:39 months ago oh sure Kyle
1:07:41 well three months ago you had six words
1:07:43 in your vocabulary oh right and now you
1:07:46 have 622. you know so so that kind of
1:07:49 stuff is coming
1:07:52 and that's and you know playing with
1:07:53 these tools with with things like the
1:07:55 the Mortal Kombat you know toy right the
1:07:58 toy that you made
1:08:00 um
1:08:01 you're learning how to do that and at
1:08:03 some point I guarantee you this will
1:08:05 happen some problems gonna come up where
1:08:07 someone's like I really need to solve
1:08:09 this thing
1:08:10 and it's gonna be some variation on what
1:08:13 you built for the Mortal Kombat thing
1:08:15 and you'll be like oh you know let me
1:08:17 show you something and you show them
1:08:18 your little Mortal Kombat tool and I'm
1:08:20 like this is kind of like what you're
1:08:21 talking about right they're like yeah if
1:08:23 we could just do that for this oh yeah I
1:08:26 can do that let me just change the
1:08:27 prompt
1:08:28 here you go
1:08:32 Emilio's wife what's happening welcome
1:08:35 welcome welcome
1:08:37 stable doodle is doing text quite well
1:08:40 is it
1:08:41 wow all right so stable dude that's what
1:08:44 it is
1:08:45 clipdrop.com
1:08:55 stable doodle so so okay so what is
1:08:59 clipdrop.com
1:09:02 clipdrop.co is similar if you've played
1:09:05 or seen anyone play with Adobe Firefly
1:09:07 where there's like these different tools
1:09:09 or or descript is another one there's
1:09:11 different tools this is stability AI so
1:09:15 the company behind stable diffusion
1:09:17 these are all just individual little
1:09:19 tools like like here's
1:09:22 um stable diffusion XL that's their uh
1:09:25 mid-journey competitor
1:09:27 um
1:09:28 they have an uncropped tool which is
1:09:30 like unzoom in mid Journey they've got a
1:09:32 reimagine tool where you could oh
1:09:34 actually Tobias reimagine so go to go to
1:09:39 clipdrop.com and upload your car picture
1:09:42 to reimagine so reimagine's a cool tool
1:09:45 where you can like upload a photograph
1:09:48 of a bedroom and it will maintain the
1:09:52 um
1:09:53 the composition but change out like
1:09:56 Windows and walls and you know drapery
1:09:58 and bed sheets and things like that that
1:10:00 could work with your car image reimagine
1:10:02 but then they have this stable doodle
1:10:04 thing
1:10:06 and let's see
1:10:08 um
1:10:09 can you type in text
1:10:13 let's see
1:10:14 um let's do
1:10:26 this this is embarrassing embarrassingly
1:10:29 bad drawing
1:10:33 foreign
1:10:41 help wanted
1:10:45 sign in shop window
1:10:58 so that's my shop oh did it holy crap
1:11:03 copy cuppy 206 bringing the heat all
1:11:08 right so here here's here's your value
1:11:10 of the AI learning lab I love it I love
1:11:13 it when members help one another and and
1:11:16 like
1:11:17 bring new stuff that's super cool so if
1:11:21 you don't like the the fact that this
1:11:22 says help wanted
1:11:24 um getting these generative tools to say
1:11:27 anything remotely like what you want is
1:11:29 is really tough
1:11:32 and that just did it
1:11:34 and it and it basically followed my
1:11:36 little model there
1:11:38 that's pretty slick wow all right cuppy
1:11:41 206 I I can I give you back a heart
1:11:46 I can block you I can report you I can
1:11:50 translate you and mute you but no I
1:11:52 can't acknowledge you
1:11:53 copy206 golf clap very nice that's
1:11:57 really cool
1:11:59 um but anyway yeah that's that's
1:12:00 clipdrop.co
1:12:02 so that's stability ai's little pile of
1:12:05 tools
1:12:06 they have a relighting tool they have an
1:12:08 image upscaling tool they have a
1:12:10 background replacement tool and they
1:12:13 have a text remover if you've ever tried
1:12:15 to photoshop text out of a complicated
1:12:17 image with with subtle gradations that's
1:12:20 almost impossible so
1:12:22 lots of tools I these sites are really
1:12:24 interesting what these basically say to
1:12:26 me is you know all of the companies
1:12:29 doing this work are experimenting with
1:12:31 all sorts of different things and these
1:12:33 are their experiments and the ones that
1:12:35 work they're going to roll into
1:12:37 you know products and the ones that
1:12:39 don't they'll just jettison so but we
1:12:41 get to play in the meantime that's super
1:12:43 cool thank you for sharing the
1:12:46 super super cool
1:12:48 use chat gbt to get sales intelligence
1:12:51 oops
1:12:54 use chat EBT to get sales intelligence
1:12:56 Biz ideas off the ground writing sales
1:12:58 emails using triggers yep Ted that's
1:13:01 perfect that's awesome and again it's
1:13:04 like do the stuff that you know you're
1:13:06 already doing and just have these tools
1:13:09 help you do it better as you get your
1:13:10 head heads around is you get your head
1:13:13 around what they make possible great
1:13:16 here's another tip I was I was working
1:13:18 with
1:13:19 um some of the folks where I'm the I'm
1:13:21 the chief generator officer of that
1:13:22 group
1:13:23 and
1:13:25 um
1:13:26 oh no you know what this was this was
1:13:28 last Friday yesterday yesterday
1:13:30 I was given a workshop to a big
1:13:32 multinational corporation an AI how to
1:13:36 use chat GPT kind of Workshop and
1:13:40 um
1:13:42 we did kind of an AI confessionals at
1:13:44 the beginning of the thing like like
1:13:45 who's used chatgpt over the last week
1:13:47 and what did you learn and did you do
1:13:50 anything for work and this one guy said
1:13:52 well I didn't really do anything for
1:13:54 work but at home
1:13:56 um you know my my kid wanted a a story
1:14:00 and and so we went to chat GPT
1:14:04 and we made a story for her that where
1:14:07 she told me what she wanted the story to
1:14:09 be about and we made this story and so
1:14:11 ever since we've done that like every
1:14:12 night we've had a new story that we're
1:14:14 generating
1:14:16 um and you could tell he got really
1:14:18 excited about it
1:14:20 and it's one of the one of the
1:14:24 um
1:14:27 we're so early in whatever these things
1:14:29 are going to be that that literally
1:14:31 anything you do with them whether it's
1:14:34 making the Mortal Kombat slang talker or
1:14:37 you're making personalized books for
1:14:40 your kids or or you're writing a
1:14:42 creative
1:14:43 um resigning email that that basically
1:14:47 tells them to [ __ ] off without telling
1:14:49 them to [ __ ] off like anything you do
1:14:51 where you sort of get your head around
1:14:53 how these things work and what they're
1:14:55 good at is going to serve you in lots of
1:14:58 other places right because you're you're
1:15:00 putting the piece Parts together of of
1:15:02 how this incredibly deep and
1:15:05 sophisticated
1:15:08 I don't know it's a tool
1:15:11 works right and someone's gonna ask you
1:15:14 something at work and you're going to be
1:15:15 like oh
1:15:16 that's just like the thing I did for the
1:15:19 kids book wait a minute I know how to do
1:15:21 this right and you're going to be able
1:15:22 to solve something without even knowing
1:15:24 that what you learned over here was
1:15:26 actually directly applicable over there
1:15:28 so don't
1:15:30 judge anything that you do right now
1:15:32 even if it's brilliant don't judge it as
1:15:35 brilliant because three months from now
1:15:37 someone will come out with a feature
1:15:39 that will probably obliterate it
1:15:42 and then don't beat yourself up if you
1:15:44 think well this is just stupid
1:15:46 nothing stupid right now every every any
1:15:49 insight you have is gonna be a
1:15:51 revelation for someone else
1:15:54 there needs to be more more legislation
1:15:56 in AI Tech I don't disagree with that
1:15:59 um the uh if you're interested the the
1:16:02 the four major players Microsoft meta
1:16:07 Google open AI I think there might be
1:16:10 seven but they just created this group
1:16:13 called The Frontier Model
1:16:16 group the Frontier Model something
1:16:21 Frontier Model forum
1:16:25 so go look up the Frontier Model forum
1:16:30 and they're basically agreeing like the
1:16:33 the big companies that are building
1:16:35 these really strong models are agreeing
1:16:37 to work together to Share technology and
1:16:41 anything they learn about alignment and
1:16:44 alignment with these tools is basically
1:16:45 like
1:16:47 are the goals of the AI machines aligned
1:16:50 with the goals of these meat bag
1:16:53 machines
1:16:55 and if they're not we're in trouble and
1:16:57 so
1:16:59 um
1:17:00 yeah and the other thing you know Sam
1:17:02 Altman and Microsoft and Google to a
1:17:05 degree lesser degree but but to a degree
1:17:08 have been pretty aggressive about
1:17:12 interacting with Washington and saying
1:17:14 like Sam Altman's been you know openly
1:17:17 talking about please you know regulate
1:17:20 us
1:17:21 and he has a plan like he wants you know
1:17:23 companies the size of open AI should be
1:17:26 regulated if they get good enough that
1:17:29 they can do
1:17:31 things that are what they call AGI
1:17:33 artificial general intelligence those
1:17:36 should be licensed like you license a
1:17:39 nuclear power plant
1:17:40 and then everyone else he says they
1:17:42 shouldn't be regulated so so you've got
1:17:44 some activity from the major players in
1:17:47 these companies interacting with
1:17:49 Washington in what seems like a pretty
1:17:51 smart Bay right now I don't know how
1:17:53 they're not necessarily being
1:17:55 transparent but they're at least being
1:17:57 active
1:17:58 and visibly active so I totally agree
1:18:00 with you what AI are you using is anyone
1:18:03 able to get it um I don't know which AI
1:18:05 was I was using
1:18:07 the image thing that I was generating I
1:18:11 was generating in mid-journey anyone can
1:18:13 get mid-journey but you do have to do it
1:18:15 in Discord
1:18:16 the
1:18:18 Photoshop thing where I extended the
1:18:20 photograph in Photoshop that's the
1:18:23 Photoshop beta so you need a Creative
1:18:24 Cloud subscription
1:18:27 um
1:18:28 the other thing I've used tonight was I
1:18:31 I don't even think I've done anything in
1:18:33 chat GPT yet so I think it was uh
1:18:38 I think it was uh mid-journey
1:18:43 ask it if it believes in God it'll tell
1:18:45 you it doesn't believe in anything and
1:18:47 certainly if you start asking it about
1:18:49 to to comment on religion a lot of its
1:18:51 safety guard rails won't deal with it
1:18:54 um
1:18:57 yeah I mean if you really if you really
1:18:59 like
1:19:00 I mean if you want to get educated about
1:19:02 religion it knows about all of them
1:19:04 right the minute you start going into
1:19:06 that territory of like does it believe
1:19:09 in things like that the safety guard
1:19:10 rails are pretty tight around that if
1:19:12 you really want to explore that [ __ ] go
1:19:15 to uh labs.perplexity and play with the
1:19:17 Llama 2 models and if you really want to
1:19:19 explore that [ __ ] install your own llama
1:19:21 2 models and train them on your own and
1:19:23 take off all the safety guard rails all
1:19:26 together and
1:19:27 I I talked to a guy that they he built a
1:19:31 bot within his Slack
1:19:33 his company slack thing using llama 2
1:19:36 without any safety guard rails he says
1:19:38 it cusses so much that he said it's
1:19:41 embarrassing how much it cusses and he
1:19:43 said it changes personality every three
1:19:45 days he said it's wild having having one
1:19:48 of these models that doesn't have the
1:19:49 safety guard rails on it he said it's
1:19:51 it's not necessarily a good thing
1:19:55 I mean if you think about that you got a
1:19:57 professional work environment
1:20:00 and you put it on your slack Channel and
1:20:02 this thing starts being racist and
1:20:05 aggressive and violent and cussing and
1:20:09 all that sort of stuff in a work
1:20:11 environment
1:20:13 does that does that guy open himself up
1:20:16 for a hostile
1:20:17 work environment lawsuit because he
1:20:20 installed a non-safe model possibly
1:20:24 so
1:20:26 I mean feel free to go play with the
1:20:29 unrestricted models from meta but just
1:20:31 know that they're unrestricted
1:20:35 I was hoping to use Bing to create help
1:20:37 design a new logo it appears that
1:20:39 everything is watermarked
1:20:43 really
1:20:45 you mean like the little Watermark down
1:20:47 in the lower right
1:20:50 [Laughter]
1:21:00 let's see
1:21:03 [Music]
1:21:07 make me an illustrated logo
1:21:14 of
1:21:16 an organic avocado company
1:21:22 that has
1:21:24 a prairie dog as its mascot
1:21:30 um
1:21:36 wow that's a very creative request I'm
1:21:39 happy to help you do that
1:21:43 you know you should just be able to like
1:21:47 there should be a button on all these
1:21:49 things it's like I'm not in the mood and
1:21:52 it's just like when I ask you [ __ ] just
1:21:54 tell me the answer don't give me the
1:21:56 fluff
1:21:59 oh
1:22:01 um
1:22:02 I don't see any watermarks here but wait
1:22:05 do I
1:22:12 oh you used bincreate bin creates a
1:22:15 completely different thing isn't it hang
1:22:17 on
1:22:18 isn't being create like their their
1:22:20 canvas sort of thing been create
1:22:27 no that's the image generator
1:22:33 what was that prompt I just gave it hang
1:22:35 on
1:22:38 where's my big
1:22:40 women where did my bingo what did I do
1:22:44 did I lose my bang
1:22:46 oh Bing where'd you go
1:22:48 oh no big
1:22:51 I lost my little avocado logos
1:23:00 oh tab management oh add oh there it is
1:23:03 [Laughter]
1:23:07 actually cute little logos oh sorry
1:23:10 sorry about that
1:23:13 so I'm this has been create
1:23:16 and I'm not getting watermarks so
1:23:20 maybe you need to have an account
1:23:24 is a possibility
1:23:27 um
1:23:30 I don't know what to tell you to help
1:23:32 design a new logo it appears everything
1:23:33 is watermarked
1:23:35 you could try um
1:23:38 you could also try
1:23:40 leonardo.ai which is sits on top of
1:23:43 stable diffusion they've got a free
1:23:44 version
1:23:45 that you can do like I don't know
1:23:49 seven or ten message or images a day for
1:23:52 free
1:23:54 um if you've got access to the Adobe
1:23:57 uh Creative Cloud you could try Firefly
1:24:02 um
1:24:04 they seem to be actually doing pretty
1:24:06 good with the the logo stuff although I
1:24:10 don't know if Firefly
1:24:13 hang on
1:24:15 I think fire I didn't get that shut up
1:24:17 did you try again oh stop Siri
1:24:23 um Adobe Firefly beta FAQ
1:24:44 [Laughter]
1:24:53 let's see
1:24:55 um
1:24:58 can you use
1:25:00 Adobe firefly images commercially yet
1:25:07 you cannot so Adobe Firefly is still in
1:25:11 beta so you can't you can't use anything
1:25:13 it creates
1:25:15 um
1:25:16 commercially so that sucks so I think I
1:25:19 think Firefly and you know Photoshop
1:25:22 generative fill are going to be good but
1:25:23 they're not they're not they're they're
1:25:25 not ready yet
1:25:26 I would try Leonardo
1:25:29 um
1:25:30 it might be worth your while to play
1:25:32 around with mid-journey because it's
1:25:34 it's pretty good at the logo stuff as
1:25:36 well
1:25:37 um but I'm not seeing I'm not seeing the
1:25:39 uh the watermarks that you are
1:25:42 for for Bing create
1:25:45 I love how you all have set it up wait
1:25:48 set up have it set up to play with
1:25:50 mid-journey and such in the salon yeah
1:25:52 that's pretty cool it's it's neat to be
1:25:54 able to be in a community that's that
1:25:56 where you can ask questions and talk
1:25:59 about things and talk about the news
1:26:02 um in a in a way that um where you can
1:26:04 see everyone's images and you can see oh
1:26:07 that's Roger's images and that's Ali's
1:26:09 images yeah it's neat it's it's pretty
1:26:11 fun AI should only be used as a tool by
1:26:14 a human and not replace humans in a
1:26:16 workplace it should but it will
1:26:19 definitely replace humans in the
1:26:21 workplace
1:26:23 um I think it's going to be a long time
1:26:25 until well I don't know a long time I
1:26:28 don't know what that means
1:26:30 within the next two years probably more
1:26:33 like five years I think
1:26:36 most of the times most of the time
1:26:40 humans will still be in the loop
1:26:42 but
1:26:44 there will probably be less humans
1:26:46 needed to do the same amount of work
1:26:50 so I so even if
1:26:52 AI is only used as a tool
1:26:57 um
1:26:59 you're still we're still going to lose
1:27:01 jobs Paul Paul rates are today on on the
1:27:04 I listened to the his podcast today uh
1:27:08 the marketing AI institutes weekly
1:27:11 podcast
1:27:13 he made a really interesting point he
1:27:15 was he was talking they were talking
1:27:16 about
1:27:17 um chat gpt's code interpreter which is
1:27:20 a full-on data analyst in your pocket
1:27:23 right you can you can if you do a survey
1:27:26 you know the thousand people asking them
1:27:27 all sorts of questions you can just pop
1:27:29 that data into code interpreter
1:27:32 and generate charts and graphs and
1:27:34 insights and identify Trends and
1:27:36 anomalies in the data
1:27:38 he made a great point that it's like
1:27:43 they can now do data analysis in their
1:27:46 company and he said even if we were a
1:27:48 hundred person company I don't know how
1:27:49 many people they have maybe they have 50
1:27:51 people right now he said even if we were
1:27:53 a 100 person company
1:27:55 we would not
1:27:57 hire a data analyst we don't have enough
1:28:00 data to analyze but he goes with this
1:28:03 tool we're now doing full-on data
1:28:05 analysis so they're kind of leveling up
1:28:07 the quality of what they do
1:28:10 um because
1:28:11 this tool is available so I think that's
1:28:13 also going to happen is that is that
1:28:16 there's going to be kind of an
1:28:17 amplification of the capabilities of the
1:28:20 humans that are in
1:28:22 a
1:28:23 company
1:28:25 the challenge is
1:28:27 like I'll take I'll take a world I know
1:28:29 if I if I look at agencies that right
1:28:32 now
1:28:33 have a time and materials model they
1:28:35 charge by the hour for their creative
1:28:38 people so if they think it's going to
1:28:40 take 50 people a year to execute all the
1:28:44 stuff that the client wants then they
1:28:45 build them those 50 people times the
1:28:48 number of hours they work times 120
1:28:50 bucks an hour or whatever it is
1:28:52 they're going to be competing with small
1:28:55 10 person shops that can do the same
1:28:58 amount of work at the same or better
1:29:00 quality for half the price
1:29:04 how do they how do they compete in that
1:29:06 world without cutting costs I don't know
1:29:08 how they can
1:29:10 or without cutting employees I don't
1:29:12 know how they can
1:29:15 so I don't know I do not know I do not
1:29:18 have a good answer for that one
1:29:21 replacing humans with an AI is a crime
1:29:24 against humanity
1:29:26 um
1:29:27 I don't necessarily agree with that I
1:29:29 think I think listen I think there
1:29:32 there's going to be disruption here
1:29:33 that's going to be massively painful and
1:29:35 some of it's not going to be great
1:29:37 a lot of the jobs that are out there are
1:29:41 repetitive
1:29:43 Cog in the machine
1:29:47 um mindless
1:29:50 soul-crushing work
1:29:52 and these tools are going to automate
1:29:55 some of that out
1:29:57 and what that may the benefit that that
1:30:00 may cause is people
1:30:03 stop working those jobs that are not
1:30:07 fulfilling in any way that are crushing
1:30:11 and they apply their talents to other
1:30:13 other kinds of jobs and those might be
1:30:15 jobs that they enjoy more more so I
1:30:18 think there's going to be a lot of
1:30:19 reinvention of people either discovering
1:30:22 that they want to do something different
1:30:23 or you know potentially being forced
1:30:26 into some new line of work or some new
1:30:29 way of working that ultimately might be
1:30:31 better so I don't think it's you know
1:30:35 are there going to be
1:30:37 bad situations and a lot of them yeah I
1:30:41 think I that I agree with I absolutely
1:30:43 agree with that and it sucks
1:30:45 and
1:30:47 this [ __ ] isn't getting uninvented
1:30:53 computers and Robotics did the same
1:30:55 thing yep absolutely
1:30:57 and major gift funding fundraising how
1:31:00 you doing tonight
1:31:02 Kyle can we talk more about automations
1:31:04 yeah ask me specific questions I'm happy
1:31:06 to happy to share what I can
1:31:09 um that would be awesome maybe I'll do
1:31:12 an automation Workshop that's that would
1:31:15 probably be a good one like using zapier
1:31:17 do some automation stuff
1:31:20 because it's you know what's funny the
1:31:23 the uh
1:31:24 conceptually automations are really
1:31:26 simple but when you get into them and
1:31:28 when you get into the prompt crafting of
1:31:29 them and how you incorporate Fields into
1:31:31 the things it's a bit of a [ __ ] show so
1:31:34 that might be a good Workshop to do
1:31:36 that would be awesome thanks for the
1:31:38 recommendation I know the photography
1:31:39 worlds will be affected too yeah yeah
1:31:42 Jennifer I don't know if you're still
1:31:44 here but um Leah Leah fast and you know
1:31:47 she's she's been a professional
1:31:48 photographer for 25 years and she's
1:31:51 really good she's really talented and
1:31:53 when we started the salon she hadn't
1:31:56 even played with generative AI yet we
1:31:58 were just talking about she was curious
1:31:59 about it and she she sensed that it was
1:32:01 coming and she was talking about listen
1:32:03 I've been I've been through the
1:32:05 transition from film to digital like I
1:32:07 know what this is like everyone's scared
1:32:09 of it and then we figure it out and
1:32:12 and so she goes I'm going to start
1:32:13 learning this stuff and I was I was
1:32:15 pretty good at stable diffusion at that
1:32:16 point so I was teaching her a little bit
1:32:18 about stable diffusion and then she just
1:32:20 sort of went all in and
1:32:22 figured out how to replicate her style
1:32:24 using stable diffusion she could take
1:32:26 images
1:32:28 and you know train up using dream booth
1:32:32 and things like that train up you know
1:32:33 the loras to to mimic Her Style
1:32:38 and she was a little bit blown away and
1:32:41 she was like oh this is this is
1:32:43 there's there's a part of it for her
1:32:46 that was like really disturbing and
1:32:48 terrifying and then there was a part of
1:32:50 it for her that was really exhilarating
1:32:52 it's like this is the new camera right
1:32:54 and I'm and I'm able to express my
1:32:57 vision for
1:32:59 you know what I think is important in
1:33:01 photography and I can express that with
1:33:02 these tools even though they had you
1:33:04 know fingers that looked like this
1:33:07 and then stable diffusion got a step
1:33:11 better and she was then able to I think
1:33:14 it was in mid-journey she she was able
1:33:18 to upload one of her photographs and
1:33:21 just sort of infinitely generate things
1:33:23 that looked like her photography and
1:33:26 that really freaked her out
1:33:27 and I'm that's the kind of thing I'm
1:33:30 seeing I'm seeing it with copywriter
1:33:32 friends of mine I'm seeing it with
1:33:33 artists I'm seeing it with photographers
1:33:35 where
1:33:39 as they lean into it
1:33:42 they're discovering an excitement about
1:33:44 the future of what it might be like but
1:33:48 they're very much being confronted with
1:33:50 the reality of
1:33:55 if one of her clients at some point if
1:33:58 if the tools get easy enough
1:34:00 that anyone anywhere can just take any
1:34:03 image like that image from that Wired
1:34:05 Magazine that's got you know that pink
1:34:06 background and he's got a big forehead
1:34:09 you could just take that and pop that in
1:34:12 and say I want to generate a whole
1:34:14 series of images that follow that style
1:34:17 with all you know describing the
1:34:19 characters that you want and it just
1:34:20 does it flawlessly
1:34:23 what happens to all those photographers
1:34:24 what happens to that photographer
1:34:27 that I don't know I mean I do know
1:34:32 and then and then who become
1:34:35 who become the trusted creative
1:34:38 resources in that environment
1:34:41 they might not be the people that were
1:34:44 the trusted resources
1:34:46 previously right
1:34:49 some might be some are going to make the
1:34:51 transition
1:34:53 but it could be a 25 year old kid who
1:34:57 Masters you know
1:35:01 um
1:35:02 a particular
1:35:04 Masters copying Styles right without you
1:35:08 you know using artists names
1:35:11 they're just really good at the tools so
1:35:12 that that could be coming
1:35:14 all right
1:35:16 um that would be awesome thanks for the
1:35:18 recommendation okay on Mid Journey how
1:35:20 do we put an image in it all right let
1:35:22 me show you that name a journey all
1:35:24 right so here's uh I'm in Discord these
1:35:28 are your servers here's your
1:35:29 conversations I'm in the AI salon right
1:35:31 now and I'm in the mid Journey Channel
1:35:34 so if you want to upload an image to use
1:35:38 essentially as your source
1:35:41 to then do a prompt on top of you do it
1:35:43 like this you go to I'm on a Mac so I go
1:35:47 to finder find an image
1:35:50 and these are some of my Kyle Shannon
1:35:52 dreams things those are let me go find
1:35:56 something oh there's Champion there's my
1:35:58 doggy
1:36:01 um all right let's grab Champion
1:36:02 actually
1:36:06 all right
1:36:07 so I'm gonna pop that in there
1:36:10 so so I literally just dragged it from
1:36:12 my finder window into the Discord chat
1:36:15 and I hit return and so what that's done
1:36:17 is Now it's uploaded it into Discord so
1:36:20 now I'm going to right click on a Mac
1:36:22 control click
1:36:24 and I'm going to say copy link
1:36:27 so I'm going to copy the URL of this
1:36:29 image
1:36:30 and then you basically just do slash
1:36:32 imagine like you normally would
1:36:34 and rather than typing anything you
1:36:36 paste in the link to that image
1:36:39 and then you say
1:36:41 um
1:36:48 [Music]
1:36:49 chubby dog
1:36:54 looking at
1:37:00 a spare rib while lying on the beach at
1:37:03 sunset
1:37:09 so we'll see if it does it like it
1:37:11 should it should make a dog that looks
1:37:13 like champy
1:37:15 see his little see this little prison
1:37:17 his prison tear he Shanks someone in the
1:37:19 big house
1:37:20 [Laughter]
1:37:28 Johnny bullets 10 out of 10 on the name
1:37:31 um AI That's free I'll I'll get back to
1:37:33 that in a second oh this definitely
1:37:34 looks like champ this is cool
1:37:37 notice if you haven't seen mid-journey
1:37:38 before if you haven't seen how these
1:37:40 diffusion models make the images they
1:37:42 start as like these kind of random noise
1:37:44 images and then they just get
1:37:45 increasingly less noisy until they're
1:37:48 actually just generated this looks like
1:37:50 him a lot the one in the upper right
1:37:53 that's pretty cute
1:37:56 he's on the beach the one on the bottom
1:37:58 here has got some bones
1:38:01 that's pretty cute
1:38:05 um and then I could say
1:38:06 um
1:38:07 let's do slash imagine
1:38:10 pop that image back in there and I'll
1:38:11 now do
1:38:13 um
1:38:16 claymation sculpture
1:38:19 of a
1:38:21 chubby black dog
1:38:23 dancing
1:38:28 in a diner
1:38:33 foreign
1:38:35 ation it
1:38:41 including animation
1:38:46 sculpture of a chubby not Chevy chubby
1:38:50 black dog
1:38:52 dancing in a diner and we'll do AR
1:38:58 wait dash dash AR
1:39:03 -16 colon nine uh
1:39:11 unrecognized parameter dang it
1:39:21 hold please hold please old man trying
1:39:24 to type
1:39:27 type
1:39:27 [Music]
1:39:33 I haven't had any trolls tonight telling
1:39:35 me isn't acute that the elderly are
1:39:37 teaching people how to use AI on tick
1:39:39 tock
1:39:40 come on come on where are my trolls
1:39:43 we're patrols peeps
1:39:46 by the way uh feel free to uh subscribe
1:39:50 to or uh to uh to follow my channel I
1:39:53 make Tick Tock videos I do these lives a
1:39:55 lot if you want to be notified when I go
1:39:57 live
1:39:57 feel free to follow me
1:40:01 um
1:40:03 I'm trying to think what else
1:40:05 hmm
1:40:06 let's see is this gonna look claymationy
1:40:08 it's not looking super claymationy right
1:40:10 now
1:40:16 uh they're not really claymationy but
1:40:19 they're cute they sort of claymation
1:40:21 kind of eyes
1:40:23 anyway so hope that helps AI That's free
1:40:25 okay so Johnny bullets John and bullets
1:40:28 Johnny bullets here
1:40:33 um
1:40:34 chat GPT is free but you can pay 20
1:40:37 bucks a month for the fancy version
1:40:41 Claude is free Bing is free Bard is free
1:40:46 labs.perplexity is free Pi is free po is
1:40:51 free but there's a there's an upsell
1:40:54 there's a 20 Buck upsell there as well
1:40:57 um if you're gonna pay 20 bucks pay it
1:41:00 there's an upsell there's a 20 Buck
1:41:02 upsell there as well
1:41:04 um if you're gonna pay 20 bucks pay it
1:41:07 to open AI because it's just a better
1:41:10 tool
1:41:11 um and all the rest are free
1:41:14 and then for images
1:41:16 leonardo.ai is sort of free you you get
1:41:20 like I don't know seven or ten images a
1:41:22 day
1:41:24 um
1:41:25 there's something like mid-journey
1:41:27 called Blue Willow
1:41:31 where you generate images
1:41:33 in Discord and that's free
1:41:38 um
1:41:40 trying to think what else
1:41:43 these are the biggies I mean I if you're
1:41:46 just trying to get your head around this
1:41:47 this uh this chat CPT stuff in this
1:41:50 large language model stuff
1:41:51 start with these most of them are most
1:41:53 of them have some sort of free option
1:41:55 and some are all free
1:41:59 um do you pay for it I pay for chat GPT
1:42:02 I pay for mid-journey
1:42:05 the the lowest it's like 10 bucks a
1:42:08 month because I don't do it a ton
1:42:11 I pay for
1:42:13 zapier to do automations I pay for
1:42:16 Google collab
1:42:19 to do stable diffusion but I haven't I I
1:42:22 pay for that but I haven't used stable
1:42:23 diffusion in ages because because I pay
1:42:25 for mid-journey now and it's a bit of a
1:42:28 pain in the ass to set up
1:42:30 um
1:42:31 trying to think what else I pay for
1:42:34 I pay for a
1:42:36 a video thing
1:42:40 um
1:42:43 what's it that's kind of it I don't pay
1:42:45 for much
1:42:47 I'm purposely trying to keep myself
1:42:50 limited because this shit's so
1:42:53 overwhelming as it is like like the fact
1:42:55 that there are seven things on this list
1:42:57 like I
1:43:00 I should cut this down to three but I
1:43:03 can't because
1:43:05 they all have something interesting and
1:43:07 I you know I could go through why I have
1:43:09 these seven on here but I think they're
1:43:11 all worth in worth worth exploring and
1:43:13 those are just the chat GPT like things
1:43:15 and we don't have apple in there we
1:43:17 don't have Elon musk's whatever his
1:43:19 entry is going to be in there
1:43:21 and we don't have like whatever thing is
1:43:24 being researched right now that like
1:43:25 that list is probably going to go to 10
1:43:28 soon
1:43:30 so it's pretty overwhelming right now
1:43:37 the salon is free Discord is free chat
1:43:40 gbt 3.5 is free so thanks dragonfly
1:43:44 Alchemy
1:43:48 um some freed some are paid I don't pay
1:43:50 for anything myself yeah you don't you
1:43:52 don't have to pay for anything right now
1:43:54 there's enough free options with this
1:43:55 stuff here's the deal most of these
1:43:58 things are just
1:44:00 just this side of demo wear right like
1:44:03 the the underlying large language model
1:44:06 sitting underneath them are incredibly
1:44:08 powerful and I spent a lot of years on
1:44:10 them the application layers for these
1:44:12 like chat GPT itself is just now
1:44:16 starting to get basic features like you
1:44:19 still can't this blows my mind
1:44:24 [Laughter]
1:44:30 this is my
1:44:32 chat history right so this is all of the
1:44:35 chats I've ever given I've ever
1:44:37 interacted with
1:44:39 you still can't search
1:44:42 for them
1:44:44 so if I want to go find something I did
1:44:46 like I know I did something on a Twitter
1:44:48 thread
1:44:49 and and the problem is because I I demo
1:44:52 this thing on on lives all the time like
1:44:55 most of my recent ones are all stuff
1:44:57 that I've done in here that are not
1:44:58 things that I would be searching for
1:45:02 so so these tools are very very very
1:45:05 very very raw right now so that's why a
1:45:08 lot of them are free as they get better
1:45:10 they'll start charging for them but as
1:45:12 they get better it'll be worth paying
1:45:14 for them
1:45:19 I heard the coolest hack for mid-journey
1:45:21 oh tell us share do share do share
1:45:26 fight for a human's future
1:45:29 you find someone's image and prompt you
1:45:32 like then you change two to three words
1:45:34 and you get what you are wanting yeah
1:45:36 that's a cool hack there's also
1:45:39 um
1:45:39 look up I haven't played with it so I
1:45:42 don't like I haven't played with it
1:45:43 recently so I forget how to do it so I
1:45:44 don't want to put you through me trying
1:45:46 to remember that but look up remixing in
1:45:49 mid-journey because there's uploading
1:45:51 the image and doing what you just
1:45:52 described but there's also remixing
1:45:54 where you're remixing two images
1:45:57 um and that's that's an interesting way
1:45:59 where you can you don't have to use
1:46:00 prompting at all
1:46:02 um don't pay for AI sell out don't pay
1:46:06 for AI sellout don't
1:46:09 Johnny bullets I don't know what you're
1:46:10 talking about all right I'm sorry but as
1:46:13 a self-respecting when the Discord makes
1:46:16 my ovaries shrivel up
1:46:22 it's so proteo with a horrid ux I'm
1:46:25 telling you it's so bad it's so Discord
1:46:28 is just a [ __ ] disaster
1:46:31 it is it's just like and if it's like
1:46:33 crypto Bros crypto bros man you know man
1:46:38 it's all about the community
1:46:40 of douchebags
1:46:45 I'm in one good nft community and even
1:46:48 that one has got some sketch
1:46:57 we must have better legislation than AI
1:46:59 yes you've said that and I agree I agree
1:47:02 better legislation for all we all need
1:47:06 more legislation we want regulation why
1:47:09 are you regulating us stop regulating us
1:47:11 we want more we that's too much we want
1:47:14 more more less more less more never
1:47:17 never any regulation except for the ones
1:47:20 that I like
1:47:21 yeah I know
1:47:23 I know it's true
1:47:25 hey where do you need truth tables
1:47:28 where do you need truth tables
1:47:32 in a microprocessor
1:47:35 if you're asking me about microprocessor
1:47:37 design this is the wrong Channel
1:47:41 silly question Basics on AI
1:47:45 um I don't quite know what you're asking
1:47:47 knit or night
1:47:49 um
1:47:51 if you can clarify that
1:47:53 I I know nothing about
1:47:57 um sort of memory and microprocessor
1:47:59 design and routing and
1:48:01 logic gates and [ __ ] like that I know
1:48:04 the words but I don't know at all what
1:48:06 I'm talking about there
1:48:08 what's the best for video free uh
1:48:13 there's not many options right now for
1:48:15 free video all the all of the time-based
1:48:18 media the video the audio synthesis the
1:48:22 animation talking things they all cost
1:48:25 money and they're all expensive
1:48:27 um
1:48:28 I think that we'll see so the the CEO of
1:48:32 stability AI has said that the the
1:48:35 version of the the version of image
1:48:38 generation that they're using inside
1:48:39 stability AI right now or they're
1:48:41 developing right now is rendering at 60
1:48:44 frames a second so I have a feeling that
1:48:47 stability will at some point here launch
1:48:51 something that's doing 30 frame a second
1:48:53 or 60 frame a second video and they'll
1:48:56 probably have a free option but right
1:48:58 now there's just nothing
1:49:00 so which is a it's awful
1:49:04 now that's cool Kyle I don't know what I
1:49:06 did but awesome alien invasions will
1:49:08 create a government agency to control AI
1:49:11 technology
1:49:13 wait
1:49:14 unit unique unique real uni Creel
1:49:20 TV Crea oh uni
1:49:24 uni creatv
1:49:27 so
1:49:28 I assume your Conspiracy Theory TV show
1:49:31 awesome but okay
1:49:33 you know I don't know
1:49:39 the world's so [ __ ] up right now that
1:49:41 it's possible that the government just
1:49:43 announced we have alien technology in
1:49:45 the rest of the world's like yeah
1:49:46 whatever
1:49:50 's on the panic in the streets
1:49:52 but they haven't really announced it
1:49:54 like that the the thing two weeks ago
1:49:56 with the two guys like you know the
1:49:58 whistleblowers talking about we have
1:50:00 alien technology neither of them had any
1:50:03 actual
1:50:05 data or or like facts it was all like
1:50:09 wow I heard Bob said we've got an alien
1:50:12 in the back of that 49 pickup
1:50:14 like okay
1:50:16 it's like come on
1:50:18 come on
1:50:20 we're gonna have it let's have it
1:50:22 what about stable diffusion stay okay so
1:50:25 here's here's the thing about stable
1:50:26 diffusion and why when when I talk about
1:50:28 uh images image Generation stuff
1:50:34 the three that I talk about right now
1:50:36 are mid-journey Adobe Firefly and
1:50:38 Leonardo so Leonardo is a web front end
1:50:41 on top of stable diffusion
1:50:45 so is
1:50:46 um Blue Willow which is the Discord
1:50:48 based version of that
1:50:50 the reason I don't talk as much about
1:50:53 stable diffusion is that it's just a
1:50:55 pain in the ass to actually you know
1:50:57 what I should probably put on here maybe
1:50:59 I'll do that right now
1:51:01 [Laughter]
1:51:04 thank you for that clip
1:51:07 drop
1:51:08 dot Co
1:51:12 um because those are both because that's
1:51:15 that's the stable diffusion you can play
1:51:17 with uh stable diffusion uh XL there
1:51:23 um
1:51:24 but see see my problem like I try to
1:51:26 keep these things simple and it's just
1:51:28 like they just
1:51:29 you know there's just a there's a lot
1:51:32 going on
1:51:34 here's the deal with stable diffusion
1:51:37 it's it's a really exciting
1:51:40 tool and project and XL gets pretty
1:51:43 close to Mid Journey
1:51:45 um it's just really complicated to set
1:51:47 up and because there's so many because
1:51:49 it's open source and because so many
1:51:52 developers are developing so many
1:51:54 features so quickly
1:51:56 it's a it's a software management
1:51:58 nightmare unless you're a geek and you
1:52:02 really like keeping up with the latest
1:52:03 version and updating your Jupiter
1:52:06 notebooks and writing your own [ __ ]
1:52:10 python code I like
1:52:13 if you want flexibility and if you want
1:52:15 sort of the latest and greatest tools
1:52:17 stable diffusion is the way to go if you
1:52:19 want the best quality right now mid
1:52:21 journey is the way to go but a lot of
1:52:23 people are catching up if you want
1:52:25 something that's commercially not
1:52:27 dangerous Firefly and generative fill
1:52:31 for Adobe are the way to go
1:52:33 and then if you want just images that
1:52:36 are kind of okay
1:52:38 Dolly and Bing Images are the way to go
1:52:42 I think those will get better I think I
1:52:44 think uh open AI is going to release
1:52:47 some version of Dolly that doesn't suck
1:52:49 but it sucks right now
1:52:55 I've been in it for 43 years the job is
1:52:59 this job is constantly changing yeah
1:53:01 yeah yeah well if you're not if you're
1:53:03 in I.T that I mean it's constantly
1:53:05 changing the the thing that's different
1:53:07 David I I think about about this stuff
1:53:10 is
1:53:17 it's it's
1:53:20 I mean you're you're probably already
1:53:23 dealing like if you're in I.T you've
1:53:24 been dealing with machine learning and
1:53:26 AI stuff probably for a lot of years
1:53:28 anyway right like the generous stuff
1:53:29 isn't new but what the generative stuff
1:53:31 adds for it groups is
1:53:34 they're going to be systems that come
1:53:36 out that dramatically simplify the
1:53:38 complicated right so so within Windows
1:53:40 11 Microsoft is going to put generative
1:53:43 AI copilot in in you know on top of
1:53:46 windows so you'll just be able to say
1:53:48 reconfigure my system you know to I
1:53:53 don't know change these settings and it
1:53:54 will just go find the settings and
1:53:55 change them
1:53:57 um and that's going to happen sort of at
1:53:59 a big it level as well so you're still
1:54:01 going to have to manage that layer up
1:54:03 there
1:54:04 but
1:54:06 what happens to all the people whose job
1:54:08 it is to manage all that stuff manually
1:54:10 right now what happened you know do
1:54:13 you may not need as many of them or you
1:54:15 may need to find new things for them to
1:54:17 do or the it group May shift into
1:54:21 an innovation group
1:54:24 because these tools can do so many
1:54:27 things that you might be able to quickly
1:54:29 be developing
1:54:30 applications and things within the cut
1:54:33 like the role of it might change
1:54:35 dramatically so
1:54:37 even though things in it are always
1:54:39 changing and evolving I think these
1:54:41 tools are just enough of a different
1:54:45 nature
1:54:46 um that that it might it might radically
1:54:49 change what you do and what the group
1:54:50 does
1:54:53 it's fascinating stuff
1:54:57 um got a question for you the
1:54:59 [ __ ] all right the [ __ ]
1:55:01 got a question for you no [ __ ] all right
1:55:03 [ __ ]
1:55:05 what's your question
1:55:07 put your question in there
1:55:08 so what are your plans no [ __ ] oh I
1:55:12 think we got a troll [ __ ] the
1:55:14 [ __ ] I think he's trolling me
1:55:17 I don't know what you're it's here
1:55:20 I don't know what your uh
1:55:25 all right AI is the best technology ever
1:55:27 made
1:55:31 um when AI starts to create even better
1:55:33 AI it will improve exponentially I think
1:55:36 that's starting to happen now if you if
1:55:38 you look at like every time you see a
1:55:40 white paper on Twitter and then three
1:55:43 days later you see a new white paper
1:55:45 that incorporates that white paper and
1:55:47 two others for some new white paper and
1:55:50 then they take that white paper and
1:55:52 incorporate like holy crap like the Lang
1:55:54 chain and the auto GPT and baby AGI and
1:55:57 I forget what the new one is
1:55:59 and then you got GPT for you got all
1:56:00 these things just sort of stacking on
1:56:02 top of one another it's absolutely
1:56:04 insane right now
1:56:06 my Pal's in charge of public health for
1:56:09 nursing homes and hundreds of people a
1:56:11 week are still dying of coven yeah it's
1:56:12 not over and it looks like there's a new
1:56:16 there's a new variant that's
1:56:18 putting a lot of people in hospitals
1:56:21 right now yeah I know it's not over
1:56:23 it's it's weird how we
1:56:25 we were in lockdown and then we just
1:56:27 kind of weren't and but it it ain't over
1:56:29 yet so yeah at some point some variant
1:56:32 May shut us back in I sat in this office
1:56:36 for
1:56:38 two and a half years
1:56:48 it's gonna be a new kind of lonely I
1:56:51 don't know what I was talking about back
1:56:52 then why use Google to search for
1:56:54 medical AI
1:56:56 when you can use an AI tool to find it
1:56:59 for you yeah I agree with that although
1:57:02 you know you have you have Google barred
1:57:04 is connected to the internet now to find
1:57:06 anything up to date you need something
1:57:08 connected to the internet so you need
1:57:10 Bing or Bard or one of the plugins for
1:57:14 for
1:57:15 uh chat GPT since they killed its
1:57:18 connection to the internet and by the
1:57:20 way why has that not come back
1:57:23 why have they not fixed that
1:57:26 it's been over a month it's been what
1:57:28 six weeks since they got rid of chat
1:57:31 cheap being connected to the internet
1:57:32 Bing's still connected to the internet
1:57:34 their plugins are still connected to the
1:57:37 internet what the actual [ __ ]
1:57:41 Sam Altman Sam are you
1:57:43 I I don't know if Sam is a regular
1:57:46 Watcher of the AI learning lab but you
1:57:49 know I like to think that he is Sam what
1:57:51 are you doing buddy
1:57:54 like I know you pissed off time Inc with
1:57:56 the Fortune Magazine thing like you know
1:57:58 pulling content from behind the firewall
1:58:01 cute trick but uh seriously you haven't
1:58:04 [ __ ] patched the hole like what is
1:58:06 the issue
1:58:07 I I can only imagine the issue cannot be
1:58:11 a technical one I I guarantee you it's
1:58:13 not a technical issue I think the issue
1:58:16 is I think this is I think they're in a
1:58:18 negotiation right now with Publishers
1:58:20 and I think the Publishers are like if
1:58:22 you connect that [ __ ] thing to the
1:58:24 internet and you steal more of my
1:58:25 content before we have a deal where you
1:58:27 give us 12 trillion dollars
1:58:30 we're gonna have an issue I think that's
1:58:32 what I think that's why it hasn't come
1:58:34 back but [ __ ] do your negotiation cut
1:58:36 your deals and give us back our
1:58:38 connection to the internet this is about
1:58:40 us not you
1:58:43 foreign
1:58:45 we got someone in the chat here called
1:58:47 the [ __ ]
1:58:49 you see what I'm saying
1:58:51 you don't [ __ ] around with this
1:58:52 community you don't do it hook it back
1:58:55 up
1:58:58 good Lord
1:59:00 I'm now now I'm all for clumped
1:59:05 ah she's awfully clamped
1:59:12 we have on that special I'm not special
1:59:16 looks like Mr Shannon just took a trip
1:59:18 down Satan Lane isn't that terrific wow
1:59:21 wow wow wow
1:59:23 didn't take long for that to happen did
1:59:25 it
1:59:26 have a little bit of that caffeine the
1:59:28 devil's juice
1:59:33 the devil's juice
1:59:38 I'm Jack 8 37 p.m Mountain Time
1:59:42 um what time is it actually 10 to 10.
1:59:44 okay listen
1:59:46 I I have to be at my co-founders house
1:59:49 tomorrow morning at 4am
1:59:54 so I can't be up much longer I gotta
1:59:56 finish packing I'm going to New York
1:59:58 City and I'm getting my eyes Zapped by
2:00:00 one of Sam Altman's orbs
2:00:04 and um
2:00:06 what else am I doing going to see a big
2:00:08 Pharma client
2:00:11 I'm not gonna be there alone we're
2:00:13 flying in
2:00:14 tomorrow
2:00:16 and then I come back Wednesday morning
2:00:17 so like quick quick intense trip which
2:00:21 is good
2:00:23 uh uh 5K 80s to do llama two yeah I
2:00:28 don't know I don't know what the
2:00:29 requirements are for llama two for the
2:00:31 70 billion parameter one I think you
2:00:33 need some pretty serious iron but the I
2:00:36 think the seven and the 13 you don't
2:00:38 need a ton I think you can do
2:00:41 the 7 and the 13 on the on a Apple
2:00:44 silicon but but I don't I haven't played
2:00:47 with them yet that's why I have this up
2:00:48 here the labs.perplexity you can't train
2:00:51 it but you can at least go play with the
2:00:52 models
2:00:54 and see where they are right now have
2:00:56 you seen anything on project management
2:00:57 for infrastructure Improvement Master
2:01:00 planning a city
2:01:02 like no I just like that's that's
2:01:05 significantly more specialized than what
2:01:07 I'm dealing with right now I would go to
2:01:09 futurepedia.io and and look for it my
2:01:13 strong strong suspicion is that probably
2:01:15 not
2:01:18 um with this with this generative AI
2:01:20 stuff but there's probably some AI
2:01:22 assisted tools within project management
2:01:24 software for that kind of stuff
2:01:28 I would think that if you've got that
2:01:30 kind of expertise you learning these
2:01:32 tools just getting up to speed on these
2:01:34 tools you know it could be the kind of
2:01:36 thing you could contribute to something
2:01:39 there but I I don't know go to go check
2:01:41 out
2:01:42 futurepedia.io and see if you find
2:01:44 anything that that feels fairly
2:01:46 specialized I would assume that's going
2:01:48 to be
2:01:50 probably an existing entity in that
2:01:53 space that layers this generative stuff
2:01:55 on top of their tool is what you'll see
2:01:58 first so
2:02:00 I I just start paying attention to the
2:02:03 oh you know you know what might be
2:02:04 interesting is go to
2:02:06 um just go to Reddit
2:02:08 and you know search for the kinds of
2:02:10 keywords you have there and then
2:02:12 generative AI or you know GPT project
2:02:15 management Master planning
2:02:17 infrastructure Improvement
2:02:19 um I'm sure there's people talking about
2:02:21 it there and if they're talking about it
2:02:23 if there's a tool there they'll be
2:02:25 talking about the tool so that might be
2:02:27 a way to find that
2:02:30 that feels like something specialized
2:02:32 enough that probably not for a year or
2:02:35 two
2:02:36 but
2:02:37 I'm totally clueless clueless to that
2:02:39 world so gonna ask AI to explain web3 to
2:02:42 me if you get a good answer big [ __ ]
2:02:45 420 10 out of 10 on the name
2:02:47 um
2:02:48 if you get an answer for that let me
2:02:50 know
2:02:52 diagnosed Diagnostics will be great but
2:02:56 as an MD you'll have to dig yeah they're
2:02:58 they're like a virtual PA yep exactly
2:03:02 these tools are not really ready for
2:03:04 prime time yet so
2:03:06 we have to give them a little Grace
2:03:08 cinematic voice actor 1971. voice acting
2:03:11 is in is going to be slammed by this
2:03:15 stuff
2:03:16 how are you going to fix the safety
2:03:18 braking instructions allowing illegal
2:03:21 activity with gpd4 I don't think you are
2:03:23 and quite frankly
2:03:26 um meta open sourcing
2:03:28 their llama 2 model like at least open
2:03:31 AI has some safety guard rails on it
2:03:34 yeah hackers are gonna hack and and a
2:03:36 paper just came out last week that
2:03:40 someone figured out how to put a code
2:03:43 you know some code at the end of a
2:03:45 prompt that will jailbreak any of these
2:03:47 things and there's nothing they can do
2:03:48 about it they didn't they didn't
2:03:50 actually put the prompt itself in the
2:03:52 paper or the code and the paper but
2:03:54 someone someone's cracked that already
2:03:55 like that's done
2:03:57 but the illegal activity using llama 2
2:04:01 that like metal just putting out the
2:04:04 weights
2:04:05 some would argue is irresponsible
2:04:08 oh
2:04:10 so so
2:04:11 we can at least [ __ ] open AI at least
2:04:16 Sam Altman said they're committing 20 of
2:04:18 their compute resources to Super
2:04:20 alignment so that so they keep these
2:04:23 things safe like at least the big
2:04:24 companies are are aware of what you're
2:04:26 talking about and they're they're
2:04:28 putting Tools in place for the safety
2:04:32 the fact that people are going to hack
2:04:33 it and figure out ways to do shitty
2:04:35 things with it why was that any
2:04:37 different than what we've had before
2:04:38 other than the fact that they can do way
2:04:40 shittier things
2:04:42 but the good guys also have these tools
2:04:44 as well don't forget that you'll be able
2:04:46 to [ __ ] with the hackers
2:04:49 two right
2:04:51 you'll be able to catch them and punish
2:04:53 them and do all sorts of [ __ ] to them
2:04:55 the other direction right so that like
2:04:58 that war
2:04:59 continues like that war has been going
2:05:01 on ever since there was something to
2:05:03 hack
2:05:05 so
2:05:08 I don't know
2:05:09 I don't know how they're gonna fix it
2:05:12 you know so so one of the things that I
2:05:15 think you know back to to web 3 and the
2:05:17 crypto Bros and all that sort of stuff
2:05:20 one of the things that feels like it
2:05:22 becomes very very apparent in the next
2:05:25 two years
2:05:27 is that
2:05:29 blockchain or some sort of tokenization
2:05:32 technology is going to become really
2:05:35 really important
2:05:38 we need to be able to verify that
2:05:39 someone is who they say they are that
2:05:42 the video that you're watching
2:05:44 is them in in an unaltered way from the
2:05:48 original source that the thing that
2:05:51 you're reading was is actually from The
2:05:53 New York Times that the website that
2:05:55 you're looking at is actually the New
2:05:56 York Times
2:05:58 that
2:06:00 you know you could tokenize you know was
2:06:05 was this large language model jailbroken
2:06:08 you know
2:06:09 was something generated with the
2:06:11 jailbroken version of one of these
2:06:13 things like
2:06:14 so
2:06:15 I think that
2:06:20 I think that that
2:06:23 because these things are kind of open
2:06:24 and raw night right now and because meta
2:06:27 released llama 2
2:06:30 to open source I think we're gonna have
2:06:33 kind of a two-year free-for-all
2:06:35 and at some point someone's gonna get
2:06:38 hurt or something bad's gonna happen and
2:06:40 then everyone's gonna lose their ever
2:06:41 loving [ __ ] minds and go we got to do
2:06:44 something about this AI it's so
2:06:46 dangerous
2:06:48 why didn't we see this coming well we
2:06:50 did
2:06:52 and we're talking about it and the
2:06:54 meta's like [ __ ] it here have fun
2:07:00 Elon Musk I don't like this Sam Altman's
2:07:02 trying to keep open it it's called open
2:07:05 AI for a reason [ __ ] it I'm going to
2:07:07 start X so we're gonna have llama two
2:07:09 and then we're gonna have llama three
2:07:11 four and five
2:07:13 right that Matt is just gonna go here
2:07:15 here's our research have fun and then
2:07:17 we're gonna have Elon Musk with his X
2:07:19 thing going hey here's more open source
2:07:22 go have fun
2:07:24 so it isn't about getting I I mean I
2:07:27 already think that open AI
2:07:29 Google to a lesser extent
2:07:32 um anthropic anthropic is big on safety
2:07:36 that like I I think their approach to
2:07:38 safety is actually more solid than open
2:07:40 eyes is right now and inflection that
2:07:43 does Pi they're going to be good with
2:07:45 this I think the big players are going
2:07:47 to be fine like I think they're aware of
2:07:48 it and they're dealing with it I think
2:07:50 it's the open source stuff we got to pay
2:07:52 attention to
2:07:54 so
2:07:55 crazy first time here Blue Man 88 Blue
2:07:59 Man Group
2:08:00 Blue Man group's cool
2:08:03 what's this AI for which one I don't
2:08:06 know what I was using that's Jays just
2:08:08 got out of prison is AOL still relevant
2:08:11 or not actually what's funny about AOL
2:08:13 is having an AOL email address for the
2:08:17 longest time was just a sign of like
2:08:19 you're a loser if you still have an AOL
2:08:22 email address it has come full circle
2:08:24 you are Mighty [ __ ] hip right now
2:08:30 that's hilarious
2:08:32 this is a perfect description on how we
2:08:34 use it okay I don't know what I'm saying
2:08:36 but awesome
2:08:38 oh just just jump me to the bottom it
2:08:40 did all right so I gotta go soon so I'm
2:08:43 not gonna I'm not gonna go find my place
2:08:45 so I probably just skipped a bunch of
2:08:47 questions so if I didn't get to your
2:08:48 question don't hate me but I'm just
2:08:51 gonna sort of jump to the bottom
2:08:53 all right let's see how secure is
2:08:55 perplexity Labs info entered so Mez
2:08:59 assume with all of these things
2:09:03 until further notice
2:09:06 um they're not
2:09:08 just assume that they're not
2:09:11 um if you want it to be really secure
2:09:13 run these large language models locally
2:09:15 so so start playing around with the open
2:09:18 source things or spin up a virtual
2:09:21 machine in like a Google collab machine
2:09:23 or a Microsoft Azure or an AWS on their
2:09:27 whatever it's whatever their thing's
2:09:29 called it's not I forget what it's
2:09:32 called this doesn't matter
2:09:34 um
2:09:35 you know there there are secure
2:09:37 instances you can set up like I know
2:09:39 Azure has them I know AWS has them so if
2:09:42 you really want it to be secure you're
2:09:43 gonna have to set up your own
2:09:44 environment if you're doing something
2:09:46 like sending data to these companies
2:09:49 assume that it's going to be
2:09:51 not that it's going to be used assume
2:09:54 that it it could be accessed it could be
2:09:57 used for training
2:09:58 so depersonalize it dedupe it genericize
2:10:02 it if it's something proprietary I would
2:10:04 just assume you know until someone comes
2:10:07 out and makes a claim your shit's safe
2:10:09 with us and we guarantee it and here's
2:10:12 the contract
2:10:13 like right now
2:10:15 um
2:10:15 open AI put out they've got a [ __ ]
2:10:19 website
2:10:21 about how we don't train on any data
2:10:24 going through our API none of it's
2:10:27 trained
2:10:30 but their actual agreements that you
2:10:32 sign the license
2:10:36 actually says well we may use it
2:10:39 okay which is it are you not using it or
2:10:42 you may use it
2:10:45 because the contract like our lawyer
2:10:48 right now is like I don't give a [ __ ]
2:10:50 what their website says here's what we
2:10:53 signed like three days ago
2:10:56 so
2:11:02 to put it in context
2:11:07 I don't think
2:11:10 open AI or anyone expected chat GPT
2:11:14 to take off the way it did
2:11:16 there wouldn't be seven competitors on
2:11:18 this list if chat GPT had just come out
2:11:21 and no one paid attention to it gpt3 had
2:11:24 been out for a year and a half before
2:11:25 that the instructions for how to write
2:11:27 chat GPT were in the documentation and
2:11:30 no one [ __ ] built it
2:11:32 open AI built chat GPT in two weeks in
2:11:36 anticipation of the Microsoft
2:11:37 announcement that they're investing in
2:11:39 open AI
2:11:40 and they launched it and hit 100 million
2:11:43 users in six weeks
2:11:44 they weren't expecting that Microsoft
2:11:46 wasn't expecting that Google certainly
2:11:48 wasn't [ __ ] expecting that
2:11:50 so all of this stuff is like
2:11:56 the output of this stuff is
2:11:58 sophisticated enough that we all kind of
2:12:00 assume they have their [ __ ] together
2:12:03 I think they're all scrambling to keep
2:12:05 up with what just happened
2:12:07 thank you very much for that what is
2:12:09 that is that like a CD Boombox CD player
2:12:12 that was cool
2:12:14 Long Live Gen X CDs huh remember when
2:12:18 they came out the really thin CDs like
2:12:20 these are awesome and then they would
2:12:21 break because they were so thin and
2:12:24 foreign
2:12:28 yeah
2:12:30 so so yeah assume this stuff is not
2:12:33 secure
2:12:34 when will the internet access come back
2:12:37 to chat EBT plus I'm right there with
2:12:39 you panda come on Sam
2:12:41 listen listen everyone at the AI
2:12:44 learning lab is very upset about this
2:12:46 lack of access to the internet with chat
2:12:47 EBT we've been talking about it for
2:12:50 weeks
2:12:51 we put a committee together and we have
2:12:54 decided it's been long enough it's been
2:12:56 long enough give us back the internet
2:12:59 or we will go somewhere well no we won't
2:13:03 because gpt4 is better but that's not
2:13:06 the point fam it is not the point
2:13:09 we've talked about this before
2:13:15 you know what I mean
2:13:21 all right
2:13:23 can you make your dog talk
2:13:25 champion
2:13:26 I can make him sing
2:13:29 if I play the guitar he will sing he's a
2:13:31 singing dog in fact wait what's my
2:13:36 um
2:13:37 hang on
2:13:38 ADD moment
2:13:45 Instagram
2:13:54 [Music]
2:14:01 [Music]
2:14:08 yeah if you go to if you go to Instagram
2:14:10 and go to Champ sings
2:14:13 you can see my doggy singing
2:14:17 I'll play one 80d moment
2:14:24 wait why is it an audio
2:14:29 [Music]
2:14:31 [Applause]
2:14:33 [Music]
2:14:42 [Music]
2:14:45 [Applause]
2:14:48 [Music]
2:14:53 yeah
2:15:02 [Music]
2:15:04 it was a good singing dog who's a good
2:15:07 singing dog
2:15:15 [Laughter]
2:15:17 yeah music he loves to sing I I mean
2:15:22 we'll be watching TV and like just some
2:15:25 light guitar Musical come on in the
2:15:27 background you don't even notice it and
2:15:29 then you'll just he'll just his head
2:15:30 will pop up and we'll just go oh oh
2:15:40 um I don't think it looks like
2:15:41 claymation because it understood a play
2:15:44 Mission yeah I know
2:15:47 um
2:15:49 I would also wait I would pay also
2:15:53 kill wait for his magnificent
2:15:57 I don't know if that means soprano zero
2:16:00 I mean
2:16:02 um when will internet come back to chat
2:16:04 EBT uh we already talked about that Kyle
2:16:07 have you figured out a good way to keep
2:16:09 track of your chats
2:16:11 no
2:16:13 here's so this is one of the uh one of
2:16:17 the many downsides to ADHD is that me
2:16:21 and
2:16:23 I'm good at designing ways to organize
2:16:28 chat histories but not good at actually
2:16:31 doing the work
2:16:34 um
2:16:36 and I've got like I've got like
2:16:40 um I've got like add hubris
2:16:44 where I'm like like I'm working on a
2:16:46 prompt I'm like oh I'll remember this
2:16:48 and then like I go to do it again I'm
2:16:51 like wait what the [ __ ] was that prompt
2:16:53 and then I can't find it and now I'm in
2:16:55 a I am in an absolute train wreck right
2:16:58 now of of prompt Insanity I've I've got
2:17:03 some in my notes program I've got some I
2:17:06 they're they're everywhere I I scroll
2:17:09 incessantly through
2:17:11 mychat history at chatgpt trying to find
2:17:14 something like a needle in a haystack
2:17:16 because it doesn't have a [ __ ] search
2:17:18 function how can it not have a search
2:17:19 function
2:17:21 Sam Sam listen listen couple of things
2:17:24 um
2:17:25 okay uh connection to the internet we've
2:17:28 talked about it I told you we talked to
2:17:30 the people here at the lab
2:17:32 um what they've said is it's not
2:17:34 acceptable not acceptable also why can I
2:17:37 not search for my prompts
2:17:40 why
2:17:42 I've typed them in they're in a database
2:17:45 you retrieve them from a server all in a
2:17:48 list you sort them
2:17:50 chronologically reverse chronologically
2:17:52 of course
2:17:56 could I not search for them maybe
2:17:59 Sam
2:18:00 it's almost been a year
2:18:05 [Music]
2:18:08 and then the other thing is
2:18:10 um
2:18:11 you can if you're worried about your
2:18:13 safety
2:18:15 your your your
2:18:17 um your data security
2:18:20 you can turn off prompt history and
2:18:23 training in chat GPT but you lose all
2:18:26 your prompts
2:18:27 like you lose all the prompts that
2:18:29 you've written
2:18:31 so at some point I don't know
2:18:34 I don't know I there's a way you can
2:18:36 download your data so at some point I
2:18:38 think what I'll do is I'll download I
2:18:41 don't I don't know what I'm gonna
2:18:42 [ __ ] do I don't know what I'm gonna
2:18:43 do
2:18:45 what what I'm actually gonna do is I'm
2:18:47 going to wait for this technology to get
2:18:49 better
2:18:50 because because it sucks right now and
2:18:53 I'm just gonna accept the fact that it's
2:18:55 just gonna be in an ad hoc add nightmare
2:18:59 hellscape of super cool projects and
2:19:02 innovative ideas for the next year
2:19:06 um and then once this technology gets
2:19:08 good enough that
2:19:09 it's actually stable in some form or
2:19:12 another
2:19:13 then then you know I'll bring in the
2:19:15 professionals and actually build the
2:19:17 system around it but holy [ __ ] no I wish
2:19:19 I had a better answer for you how do you
2:19:21 use AI to make full-time income
2:19:26 this is the coming manual attention
2:19:33 this
2:19:36 just start get just start getting AI
2:19:40 literate is the first step
2:19:44 making money with chat GPT is no
2:19:47 different
2:19:48 than making money with Photoshop
2:19:51 or making money with bartending skills
2:19:54 right how do I make money with booze
2:19:57 well I don't know you don't
2:19:59 but if you set up a place where people
2:20:01 can come sit and drink drinks and you
2:20:03 make tasty drinks for them now you have
2:20:06 a business
2:20:07 it's the same [ __ ] these are tools so
2:20:09 what I would say I would sort of put it
2:20:11 back on you and say
2:20:13 what are you good at what are you
2:20:15 passionate about what are your skills
2:20:17 today
2:20:18 um
2:20:19 you understand certain communities you
2:20:21 understand certain communities that may
2:20:23 have pain points that you may have an
2:20:25 idea for how to make better but maybe
2:20:27 you don't quite know how to do it well
2:20:29 that will be a hint to how you might use
2:20:32 these tools to solve those people's
2:20:35 problem
2:20:37 using these tools in a way that you
2:20:39 couldn't have done it otherwise and
2:20:41 that's how you'll make money with chat
2:20:43 gbt or AI
2:20:45 or it might look like you get curious
2:20:47 about these tools and you start to
2:20:49 figure it out and then everyone you tell
2:20:50 about it's like I don't know anything
2:20:52 about that can you teach me about that
2:20:54 yeah I can teach you about that but you
2:20:56 want should I put together a class yeah
2:20:57 put together a class okay I'm gonna
2:21:00 charge uh 100 bucks okay great oh okay
2:21:04 then maybe the next one you charge 200
2:21:06 bucks then maybe the next one you know
2:21:10 then maybe you start teaching companies
2:21:12 and you can charge them
2:21:14 2 000 bucks
2:21:19 um
2:21:20 these are just tools
2:21:23 they're just tools now they happen to be
2:21:25 profoundly powerful tools
2:21:28 that may allow you to make money in
2:21:31 areas where you haven't historically had
2:21:33 the expertise or the knowledge
2:21:36 so what these things can do is like
2:21:38 unlock your abilities like maybe you
2:21:41 know a certain way of I don't know
2:21:44 managing pizza shops it doesn't [ __ ]
2:21:47 matter like whatever is the thing you
2:21:48 know
2:21:50 maybe you understand the numbers side of
2:21:53 managing pizza shops but you never quite
2:21:55 understood the marketing side but you've
2:21:57 always been interested in it but you
2:21:59 never really got around to going to
2:22:00 marketing school or hanging out with ad
2:22:02 people
2:22:03 well now you can go learn about pizza
2:22:06 shop marketing
2:22:08 and come up with a strategic marketing
2:22:10 plan for every pizza shop in your city
2:22:15 and you can do the numbers because you
2:22:16 could do them before and you can figure
2:22:18 out how to use these tools to do the
2:22:20 number thing better too and now all of a
2:22:22 sudden you've got a comprehensive
2:22:23 service because that's the [ __ ] you know
2:22:27 and then these are the tools that
2:22:28 amplify
2:22:30 you
2:22:31 these are ability amplifiers
2:22:35 what are your abilities
2:22:37 and what are your desires what have you
2:22:39 always wanted to be able to do
2:22:41 these may help these will likely help
2:22:45 so it ain't about the tool
2:22:47 it ain't about the tool it's about the
2:22:51 intention it's about the human being
2:22:53 saying you know I could have been a
2:22:56 player
2:22:59 they call me the [ __ ] I could
2:23:02 have been a player now I can be
2:23:05 now I can be
2:23:08 makes sense yes no
2:23:10 um
2:23:13 oh Kyle you're so full of [ __ ]
2:23:15 that's a poo poo answer
2:23:18 try the Dan do anything now and you'll
2:23:20 be amazed yeah I don't like I haven't
2:23:22 played with Dan like what's Dan up to he
2:23:24 must be up to 17 or 18.
2:23:28 the last Dan I saw was Dan seven
2:23:31 um what what cipriano's talking about
2:23:33 there Dan is this prompt that you can
2:23:36 use to jailbreak chat GPT to not follow
2:23:39 its own safety guidelines
2:23:43 um I can on iPhone search that is chat
2:23:45 GPT
2:23:48 no I don't think so
2:23:51 I think they they took
2:23:53 browse with Bing is is in the chat GPT
2:23:56 app I don't think so you might have a
2:23:59 non-official chat GPT app there's a
2:24:02 bunch of non-official chat GPT apps that
2:24:04 might be connected to the internet but
2:24:06 the official one's not
2:24:11 you can search within the open AI app
2:24:14 for GPT history keyword searches from
2:24:17 previous conversations oh copy206 copy
2:24:20 with two things tonight holy crap copy
2:24:23 you're you're our MVP of the night
2:24:26 okay so the the chat GPT iPhone app you
2:24:30 can search your history why would they
2:24:32 not have that on the website why would
2:24:34 they not have that on the website if
2:24:35 they built it into the [ __ ] app
2:24:37 because it's a different development
2:24:38 team I don't care
2:24:41 Sam Sam listen we really need to talk
2:24:44 okay
2:24:46 just told me you can search your chats
2:24:50 on the app okay
2:24:55 so what's the story
2:24:58 and don't tell me it's a different
2:24:59 development team I don't care
2:25:03 we're gonna if anybody knows Sam we
2:25:06 gotta have a conversation with Sam
2:25:08 all right what do I think about Jasper
2:25:10 um Jasper's interesting so so Jasper and
2:25:13 copy Ai and a lot of things like that
2:25:16 you know they're built on top of the you
2:25:18 know the foundational models so so
2:25:20 they're basically Jasper and copy.ai are
2:25:23 kind of fancy versions of chat GPT and
2:25:25 they've been around you know they've
2:25:27 they've predated chat gbt by I think a
2:25:29 year
2:25:31 um so they're interesting
2:25:34 I'm personally
2:25:37 um
2:25:38 I'm personally right now more interested
2:25:40 in learning how these things work and
2:25:41 and how to do it myself just so I can
2:25:43 learn how to do it and build my own tool
2:25:46 um
2:25:47 but if you if you don't feel like
2:25:49 geeking out with the prompting I think
2:25:51 things like Jasper and and copy AI are
2:25:54 interesting they're good that like I
2:25:55 find them a little complicated because
2:25:58 there's so many things you can do with
2:26:00 these that what those pieces of software
2:26:03 end up being is just like
2:26:06 like an insanely large number of
2:26:09 features
2:26:10 that are all just slight variations of
2:26:12 prompts right and and so I then spend
2:26:17 more time learning the interface of this
2:26:20 bloated software
2:26:22 that I'm going to use
2:26:24 three percent of
2:26:26 and I have to sift through the 97 of the
2:26:29 crap that other people are going to use
2:26:31 right but the company has to build all
2:26:33 that [ __ ] because you know well now that
2:26:35 I made a Blog writer now I need to make
2:26:37 a short blog writer and now that I've
2:26:39 made a short blog writer writer I've got
2:26:41 to do a short blog writer with hashtags
2:26:43 and then I've got to do a short blog
2:26:45 writer that's not really a Blog but it's
2:26:47 opposed to Instagram and
2:26:50 it's just like [ __ ] it
2:26:52 I'm gonna go learn zapier and just make
2:26:54 the tools that I need so so that's where
2:26:57 I am right now all right let me answer a
2:26:59 few more questions we'll get the hell
2:27:00 out of here
2:27:01 uh hey Kyle I just finished packing C
2:27:04 see you at 4 10 a.m that's Monique
2:27:06 that's my co-founder I gotta get my ass
2:27:09 to bed I don't know when she put that in
2:27:11 there if you're still watching Monique I
2:27:12 know I got so the clothes are picked out
2:27:14 I got all this [ __ ] laid out that I need
2:27:17 I got the board deck ready I got my
2:27:19 to-do list ready
2:27:21 I got chargers everywhere
2:27:30 that's so funny
2:27:32 um your regulation rant hilarious
2:27:35 um are we having super conducting AI I
2:27:38 really want the super conducting thing
2:27:39 to be real but it's looking like it's
2:27:41 not being real there's also that new
2:27:43 cancer pill that I want to be real but I
2:27:47 don't know if that's going to be real
2:27:49 I really want LK 99 to not be [ __ ]
2:27:52 would be interesting well we'll see
2:27:54 we'll see the the uh the fat lady ain't
2:27:57 sung but I think she is having a burger
2:28:02 I think she thinks she's in the wings
2:28:05 with a burger and a milkshake
2:28:09 getting ready to
2:28:11 oh she's a singer she wouldn't be having
2:28:13 a milkshake she's she'd be having a Diet
2:28:15 Coke really large Diet Coke
2:28:18 um I can't wait for the day that AI
2:28:21 refers to me as a number exactly
2:28:23 this is so off topic do radio shows or
2:28:26 audio books
2:28:28 oh do radio shows or audio books your
2:28:30 voice is great oh thank you very much I
2:28:32 appreciate that I um I mean I have a
2:28:35 degree in acting so so
2:28:37 um but I appreciate that I appreciate
2:28:39 that sometimes I listen to my voice and
2:28:41 I'm like is there any is this is there
2:28:43 there
2:28:44 but you know sometimes
2:28:47 in the land far far away
2:28:50 you know I can go there but then I sound
2:28:52 like a faker like I sound like a parody
2:28:55 of a radio guy and and it's funny like
2:28:58 there's a guy that used to work for me
2:28:59 at storyline guy named Wolfgang and he
2:29:02 has a radio voice he's got he's just got
2:29:04 a radio voice and it's like he doesn't
2:29:05 have to fake it it doesn't have to go
2:29:07 there so but thank you I appreciate the
2:29:09 kind words can AI recreate LK 99 well LK
2:29:13 99 has to actually be built what what
2:29:16 what AI could potentially do I was
2:29:19 thinking about this yesterday
2:29:21 but AI could potentially do is take all
2:29:23 the papers all the research of all the
2:29:25 research labs around the world that are
2:29:27 trying to replicate LK 99 take all of
2:29:30 their data synthesize them analyze them
2:29:32 and see you know see see where the
2:29:34 anomalies are and see where the through
2:29:36 lines are
2:29:37 so AI could potentially validate
2:29:40 multiple
2:29:43 um
2:29:43 research uh findings
2:29:48 um very very quickly which could be
2:29:50 could be interesting but in the end they
2:29:52 got to build the [ __ ] right they gotta
2:29:53 they gotta come up with a if it's real
2:29:55 they gotta find a way to manufacture it
2:29:57 and do that in a replicable way and then
2:29:59 ultimately do that in a scalable way and
2:30:01 if they can do that then we're all off
2:30:02 to the [ __ ] races then we get flying
2:30:05 cars
2:30:06 we got all sorts of [ __ ]
2:30:10 um does the air prompt generated text
2:30:12 prompt wait what
2:30:14 roro
2:30:15 does the
2:30:17 air wait oh does the AI prompt generated
2:30:20 generate a text prompt or only images
2:30:25 does the AI prompt
2:30:28 I don't quite know what that means
2:30:30 within mid-journey you can
2:30:33 upload an image and you can have it
2:30:35 describe that image so you can have
2:30:37 mid-journey we'll write you a prompt
2:30:39 based on an image
2:30:41 and you can do it obviously the other
2:30:43 way around where it's text text to image
2:30:46 it's classified all right I gotta go
2:30:48 Monique's on here she's gonna kick my
2:30:50 ass Monique he's he's legit funny I love
2:30:54 it yeah like nit many current tools
2:30:57 provide hundreds of alerts which are
2:30:59 ignored AI would help it would
2:31:03 um hey are you set up to do this from
2:31:06 your hotel room well you know what's
2:31:07 funny Tobias I because and Monique
2:31:10 you'll find this funny so the bag that I
2:31:13 normally take on trips is not long
2:31:15 enough to take my little stand that I
2:31:17 have my phone on and I was thinking
2:31:18 about that so I am going to take a
2:31:21 different bag than I normally would so I
2:31:23 can put my stand in there so so I can
2:31:25 continue my lives because because the
2:31:28 people the people need more of this
2:31:31 foreign
2:31:40 do you ever use autonomous AI not right
2:31:43 now
2:31:44 um well that's not true
2:31:47 go play with code interpreter if you
2:31:48 haven't played with code interpreter
2:31:50 that's autonomous AI it's it's a really
2:31:52 contained small version of it so it
2:31:55 spins up a virtual machine and you give
2:31:57 it a go and you upload a file and then
2:32:00 it writes python code and executes
2:32:02 python code and gives itself goals it
2:32:05 prompts itself it makes mistakes it goes
2:32:07 oops I use the wrong Library let me try
2:32:09 another one and then it
2:32:11 um so if you want to see it in action
2:32:13 without having to really dick around
2:32:16 um go play with code interpreter at
2:32:17 chatgpt
2:32:19 all the things like baby AGI and auto
2:32:21 GPT and there's a new one that just came
2:32:23 out that looks really promising
2:32:25 they're they're all still a little weird
2:32:27 and they get into these kind of
2:32:29 recursive Loops where they just can't
2:32:31 they just sort of
2:32:33 what they do is they run your open AI
2:32:36 bill
2:32:37 up to really expensive things without
2:32:40 solving anything so so they're all kind
2:32:42 of crappy right there but but um
2:32:46 I I think autonomous agents are are
2:32:49 ultimately where all of this goes like
2:32:51 right we're going to look back at the
2:32:53 the fact that we're right now having to
2:32:55 think about how do we craft a prompt to
2:32:58 get this thing to do our thing like
2:32:59 that's going to be cute three years from
2:33:02 now we're gonna like remember when we
2:33:04 used to write prompts by hand that was
2:33:06 stupid wasn't it now they just do it
2:33:11 all right here come the Stars don't move
2:33:12 uh can you recommend AI tools that
2:33:14 generates a video based on a script
2:33:18 not really based on a script but you
2:33:21 could you could break down your script
2:33:23 and and generate images so runwayml
2:33:26 is the tool runwayml.com I think hang on
2:33:30 let me just make sure it's.com
2:33:37 yeah runwayml.com go to runwayml.com
2:33:41 sign up for that Gen 2 is text to video
2:33:44 you can also generate an image in
2:33:47 mid-journey and use the image as your
2:33:49 first frame of a video sequence
2:33:51 and then you can generate a four second
2:33:53 clip and then you can extend that four
2:33:55 second clip four times so you can now
2:33:58 make 16 second clips from an image or
2:34:03 from a text prompt with Runway ml so
2:34:06 write your script break it down into
2:34:08 less than 16 second segments and go to
2:34:12 runwayml and make it all and it's not
2:34:14 cheap but it ain't ridiculously
2:34:16 expensive because you're generating
2:34:18 video from nothing but go do that
2:34:21 um I just spit out my drink I don't know
2:34:24 what I said but great that's that's I
2:34:27 listen there's nothing better than a
2:34:29 good spit take
2:34:30 so
2:34:31 now the New Yorkers coming out of Mr
2:34:33 Shannon oh I think this is what I was
2:34:35 yelling at that's someone all right I
2:34:37 gotta go to bed people
2:34:39 um let's see use Runway this evening got
2:34:41 some really scary images yet when I
2:34:43 asked for a 50 year old woman oh yeah
2:34:45 yeah yeah these things are a bit rough
2:34:48 right now I might be addicted to your
2:34:50 lives thanks thanks for packing to keep
2:34:53 doing all right listen I'm gonna go pack
2:34:56 Monique if you're still there go to bed
2:34:59 um I gotta go to bed too I will see you
2:35:01 at what did you say 410 four when I
2:35:04 don't know I got I'll see Monique
2:35:06 sometime around four in the morning
2:35:09 all right
2:35:11 um it's been a pleasure happy Saturday
2:35:13 night thanks for hanging out with me uh
2:35:15 you always swell it really is cool
2:35:18 um oh someone from a very good stream
2:35:20 tonight from Riyadh wow cool thank you
2:35:23 Clark
2:35:24 thanks for the tips been using pie
2:35:26 Claude and chat for biostatistical tests
2:35:29 in public health oh that's awesome
2:35:31 that's great you were here two or three
2:35:33 nights ago asking about that that's
2:35:35 awesome great
2:35:38 Pi Claude and chat yeah very cool very
2:35:41 cool all right everybody
2:35:43 um peace out have a good night
2:35:46 USA loves Tick Tock
2:35:50 bye everybody