Feb 20, 2024 - The Future of Generative AI, Autonomous Agents, and Multimodal Capabilities

Description

Chapters

Transcript

0:02 you
0:04 [Music]
0:10 [Laughter]
0:12 [Music]
0:21 ready
0:23 us Mar our FS
0:29 together real estate here in
0:34 my so we bought a pack of
0:40 cigarettes Mrs Wagner's
0:43 pass we've all come to look for
0:49 [Music]
0:59 America
1:02 [Music]
1:11 we bed the Greyhound in
1:17 Pittsburgh Michigan seems like a dream
1:22 now so we bought a
1:25 cigarettes this wag pass wrong words to
1:29 that verse we've all come to look for
1:33 [Music]
1:56 americ no rapping no rapping what what
1:59 are we rapping for
2:00 that was Simon and
2:03 Garfunkle it was Simon and Garfunkle you
2:05 don't rap to that sure the Disturbed
2:08 cover of Sounds of Silence
2:11 incredible you can go heavy metal with
2:13 Simon and gar funkle but rap I don't
2:15 think
2:16 [Music]
2:22 so now you're not singing at all oh this
2:24 is it silent
2:28 protest I'll hit you with an minor I'll
2:31 hit you with an A Minor I'll do
2:33 it oh yeah see I can get you every time
2:37 with that
2:39 [Music]
2:59 one
3:02 [Music]
3:09 yeah you good you
3:12 done did are you
3:14 apologizing you're apologizing for
3:16 getting so
3:18 angry yeah I know I know I I know I know
3:24 I know I
3:26 know I know I agree I agree hey I don't
3:31 like it when we fight
3:37 either I HT his feelings
3:40 exactly he's like I just want to rap
3:43 Daddy I just want to
3:45 rap I think it's music
3:49 [Laughter]
3:52 too ah fanastic Bob tell him what he's
3:57 won you've won a chicken stick
4:03 [Applause]
4:04 all
4:05 right fantastic where's champy coming
4:08 from Champ's coming from the well he
4:10 came from the bed he was on the bed and
4:13 I told him I said you you can't sing
4:15 from the
4:16 bed like get off and come come to the
4:20 office come do your job she
4:25 did but he was a little cranky I think
4:27 his blood sugar is a little low welcome
4:30 everybody Welcome to the AI learning lab
4:33 uh had a had a a nice uh uh AI Salon
4:38 after hours it was cool it was it was a
4:40 small group and it was a lot of the
4:42 original OG salers um Roger was there
4:46 Jane
4:47 CJ
4:48 um and who else was
4:52 [Music]
4:53 there anyway lots of people it was good
4:57 it was really
4:58 good we talking about
5:03 um we were talking about
5:07 Sora and then we were talking about I
5:10 don't know if if any of you on here
5:11 remember when I did some of the live Kaa
5:14 stuff the the the quick image Generation
5:16 stuff that at ca. where you can turn on
5:19 your webcam and it it interprets your
5:22 webcam and does cool [ __ ] with
5:24 it
5:27 um we were we were talking about like
5:30 that would be such a cool performance to
5:32 put together a stage
5:34 show where you're having AI render
5:38 actors in real time there's something
5:40 about that that just keeps keeps pulling
5:42 on me my my the old theatrical the old
5:46 theatrical instincts die hard you know
5:48 what I'm
5:51 saying this shirt's hard to read on you
5:54 just have to read up here at the
5:56 neck I've been using
5:58 ideogram ide's pretty good we used
6:02 ideogram a lot in here before uh before
6:04 Del got good at words what about the
6:07 long distance OG Salon Salon folks oh I
6:11 know well that's the thing is so we're
6:14 we're moving to inperson meetings every
6:16 other week and then virtual meetings
6:18 every other week cuz the hybrid just
6:19 wasn't working so um so we're getting
6:23 there we're getting there it's you know
6:25 it's a volunteer organization so things
6:28 move at a certain pace if you know what
6:30 I
6:30 [Laughter]
6:34 mean oh
6:37 Lordy I just leave it on my laptop like
6:40 a changing canvas that's actually pretty
6:41 an interesting
6:44 idea you know one of the ideas I had
6:47 that I I think Co sort of killed this
6:49 but
6:50 um one of the ideas I had for Creative
6:54 groups is just to have or if it's like
6:56 like with a group of friends just have
6:59 your web cams on all the time and don't
7:02 have it be a
7:03 meeting just have your webcams on and
7:05 you're each doing your work
7:07 independently and then at some point if
7:09 someone gets bored they can just go hey
7:11 uh Bob what uh what are you doing I'm
7:14 trying to get some work done you jerk
7:15 all right okay and just you know like
7:19 you were in a room working with some
7:20 friends I always thought that would be
7:21 kind of cool I tried it with a couple of
7:23 writing buddies lasted about a week but
7:26 uh there's something there oh lordy okay
7:30 so here's how this works if you have
7:32 questions about AI pop them in the
7:34 comments
7:34 below
7:36 um do me a favor feel free to share this
7:39 live share it with people who you think
7:42 might be remote support group yeah
7:45 exactly pretty much that or just a
7:47 remote hang like like thinking of of
7:50 telepresence not as I've got to be on a
7:53 meeting I got to be on a zoom call but
7:55 just more like hanging out with your
7:58 buds but you know living your your lives
8:01 independently um and if you need to have
8:03 a conversation you can have a
8:04 conversation or if you're working on a
8:06 project everyone can go do their stuff
8:07 and then go hey I got a
8:09 question Deep dive thoughts on the
8:11 combination of generative Ai and spatial
8:14 Tech like
8:15 apple apple Vision Pro
8:18 um I don't think they can exist without
8:21 one another
8:25 um
8:28 I looking at what
8:32 um looking what open AI is doing with
8:35 Sora and how they're doing it and then
8:37 thinking about what apple is likely
8:39 doing behind the scenes with AI and how
8:42 they're doing
8:43 it I think that
8:47 um I think that the spatial Computing
8:49 game is going to is going to get really
8:51 good really quick Kyle that's exactly
8:54 what I was talking about when I
8:55 suggested an Open Lab like old computer
8:58 labs oh that's
9:00 interesting that's interesting I didn't
9:02 I didn't see that unless is that
9:03 something you sent me directly or is
9:04 that something you posted posted on the
9:06 salon Amelio's wife because I that that
9:09 would be a cool that would actually be
9:10 kind of a cool Club you know we've got
9:13 an's Club on Saturday the GPT jams we've
9:16 now got um the business Guild has their
9:19 their their Saturday morning um business
9:25 workshop and uh yeah maybe there's
9:28 something where it's just like
9:30 there's no presentation and and it's
9:32 just it's just a
9:34 hang and but I think it's got to be in a
9:37 relatively small group because I think
9:38 if you have in a big group someone's
9:40 going to want to talk so I think part of
9:42 the thing that makes it cool is you have
9:43 it on for like 4 hours or five hours or
9:46 you know just a while and it's a working
9:49 session and everyone's working and if
9:51 someone has a question they can ask it
9:52 but it's it's not a it's not a a thing I
9:55 sent it directly okay
9:57 cool I'd be happy to host start it I
10:00 think that's really cool Amelio's wife
10:02 and absolutely feel free to start it I
10:05 mean if you want to start something like
10:06 that let's let's let's talk and we'll
10:09 we'll sort of figure out logistics on
10:11 how to get you
10:12 supported um we can go live on mudy
10:14 networks we can Mighty networks has a
10:17 weird thing I don't quite understand it
10:20 they've got a 20h hour a month limit on
10:24 going live in Mighty networks
10:27 but you can schedule it Mighty networks
10:30 so it shows up as
10:33 a as a
10:35 uh like a calendar event and it shows up
10:38 like in the sidebar that someone's live
10:41 but then you can just have it be a zoom
10:42 call or a Google meet or something like
10:44 that so that's kind of cool um hang on I
10:48 just got
10:49 a I got a
10:52 something oh
10:54 interesting
10:56 oh
10:58 huh
11:03 so Brad Perkins right now is at the Avs
11:07 game and they're they're doing AI at the
11:10 Avs game you know leave it leave it to
11:13 Big Sports Arenas to you know they they
11:16 realize they've go to they got to have
11:18 their game
11:19 rolling that's really
11:22 [Laughter]
11:26 interesting really funny
11:30 right that's super easy to do right they
11:32 just they just turn the camera around I
11:34 don't I assume they're not doing it in
11:35 real
11:37 time yeah it's called The Grateful Dead
11:39 heads and then they probably took a
11:42 picture of that person at some
11:45 point threw it into one of the AI things
11:48 very very
11:50 interesting and this is the thing you
11:52 know there's all the controversy right
11:55 now about are you using AI you should
11:57 label it if you're using AI
11:59 AI is just going to become this
12:01 pervasive thing like it's just going to
12:03 be everywhere it's going to be in
12:05 your software applications it's going to
12:07 be in your hockey games
12:11 um it's going to be everywhere and it
12:13 and it's it's not going to have such
12:15 clear boundaries did you write that with
12:18 AI I don't know some of
12:22 it which part would you like to know uh
12:25 I I so it was AI so AI did so robot
12:28 wrote it no wrote it well you said you
12:31 used AI yeah like I used spell check and
12:33 I used sentence completion and I used
12:36 grammar check and the third paragraph I
12:38 couldn't figure it out so I had it write
12:40 it but then I rewrote
12:41 it the outline was AI the idea was
12:46 AI me choosing the idea that was me me
12:50 figuring out what I'm actually going to
12:52 publish that was me so you tell me
12:53 where's the boundary right I think we're
12:55 going to run into that a lot a lot lot
12:57 lot lot lot
13:02 but right now it's it's like this scary
13:04 Beast cuz all the tools are separated
13:06 right all the tools are like these janky
13:08 ass duct tape piece of crap tools off to
13:11 the
13:15 side Amelia's wife she kind of already
13:17 did start it oh good awesome that's
13:20 great see the salon is growing so fast
13:24 that I I don't know what's going on in
13:26 there anymore so I've also been really
13:29 busy at work so that's going to happen
13:31 um all right so let me let me see what
13:33 we got going here see if we got any any
13:35 interesting comments or questions or
13:37 things like that and then we'll we'll
13:39 jump in and we'll do some
13:42 stuff we'll do some stuff The Hangover
13:44 League how do I know you're not an AI
13:47 well you don't do you wouldn't it be
13:51 swell if I were chat
13:53 gbt oh
13:55 [Laughter]
13:58 yes
14:10 I need some help beside chat
14:13 jp4 what kind of help do you need Mr
14:17 it Kevin Vasquez got the number three
14:20 gifter
14:21 badge I didn't even know you won the
14:23 number two gifter badge look at you good
14:25 Lord are we still doing AI Salon meeting
14:29 all around I don't know what that means
14:31 next Tuesday is the next AI Salon
14:33 meeting we've got Kelly
14:36 Bosch is speaking which if you haven't
14:38 seen her [ __ ] on Tik Tok go check it out
14:42 b o e c I think is her
14:46 name Kelly her work is [ __ ] amazing
14:49 it's
14:52 amazing amazing amazing amazing so what
14:56 else do you want what do you want to
14:57 talk about tonight Kyle did this once
15:00 did this once in hey Jen nine months
15:03 ago did what oh Kyle did
15:08 did I cloned myself yeah I cloned myself
15:11 a while
15:13 ago I actually just added it to we've
15:16 got I've got a board board meeting
15:17 tomorrow so I can't I can't stay up late
15:19 people I have a board meeting I don't
15:21 know if you know this but I'm a very
15:23 important founder and CEO of a
15:26 corporation we have a board meeting
15:28 tomorrow
15:34 so oh my gosh Kelly bsh she's amazing
15:37 isn't she marry
15:40 Mary what's what's amazing about a lot
15:43 of the people that I'm connecting with
15:46 that I think are doing really amazing
15:48 work in
15:50 AI almost to a
15:54 person to like two exceptions that I can
15:57 think of right now
16:00 people are apologizing for their work oh
16:02 I'm not really good it didn't really
16:04 take me too long you
16:08 know I'm I'm it is my mission right now
16:11 to stamp that [ __ ] out her work is
16:16 beautiful it's got a really clear
16:18 creative point of view it's got
16:20 storytelling in it it's got continuity
16:22 in it it's got incredible taste in music
16:25 incredible
16:28 editing and I honestly don't care what
16:31 tool she used to do it it's not my
16:36 business
16:38 and she is she no one no one has
16:41 anything to apologize for using AI
16:44 tools I'm gonna stamp that [ __ ] out I'm
16:47 gonna stamp it
16:53 out seems a lot more resistance to AI
16:56 than I thought yep burning autonomous
16:58 taxis in San Francisco yeah they're
17:01 going to be mean to the robots for a
17:02 while and then at some point the robots
17:04 are going to get guns installed in them
17:07 and then then when the taxi can shoot
17:11 back I don't think you're going to light
17:13 it on
17:21 fire the the CEO's just we got to
17:23 protect our investment this is a large
17:26 capital
17:27 expenditure
17:32 well while there are some downsides to
17:34 arming the
17:35 taxis thought it was a a dollar cost
17:38 analysis of the the risk reward factor
17:40 of
17:42 the AGI taxis
17:45 exactly where would you like to go can I
17:47 answer any questions about the
17:49 Mesopotamian War and don't [ __ ] with me
17:51 or I'll shoot
17:54 you I want to be a CEO you can be a CEO
17:59 you just you just have to [ __ ] lose
18:02 your okay step one lose your mind step
18:05 two start a
18:08 company oh man all right seems a lot
18:12 more
18:14 resistance Jason's so freaking bad
18:16 that's amazing she's amazing one day I
18:20 know she's amazing I know
18:23 Cory it's
18:26 nuts you got a board meeting tomorrow I
18:28 do have a board meeting tomorrow Kelly
18:31 Bosch AI Art thank you Silver
18:34 Fox yeah and she says she does it on her
18:36 phone crazy yeah Kelly BOS she says she
18:39 she does it on her phone um just kind of
18:42 as an absent-minded thing before she
18:45 goes to
18:47 bed but I mean isn't that [ __ ] cool
18:51 like some of the people in here are like
18:53 I don't watch Netflix anymore I just
18:55 come hang out in here and do AI [ __ ] and
18:57 she's like you know before she goes to
18:59 bed rather than watching TV she's making
19:02 incredible
19:03 art like we're just in a we're just
19:06 we're shifting there's a shift there's a
19:10 craziness there's a there's a there's a
19:12 thing happening people there's a thing
19:15 happening her work is spectacular my
19:17 faves are the ballerina the ballerina
19:19 bugs are amazing that that one that one
19:22 video she did the one she's got pinned
19:24 at the top of her account with the birds
19:26 on it it's got like 22 million views or
19:29 probably more at this point last time I
19:30 looked it was 22 million views so cool
19:34 and and you know what's you know what's
19:35 the
19:36 thing the 22 million people that looked
19:40 at that some percentage of them probably
19:42 knew it was AI some probably went oh
19:45 that's good AI art but 22 million people
19:49 don't look at something because it's AI
19:51 art 22 million people look at something
19:55 because it's
19:56 good and it resonates and it moves
20:01 them and that's what her work did she's
20:04 like I can't believe it cuz in her mind
20:08 this is something that I think we're
20:09 going to have to Grapple with as humans
20:12 as these tools get better and better and
20:13 the amount of time it takes us to
20:16 generate work
20:19 shrinks I think we're all personally
20:21 going to have to come to grips with the
20:22 fact that that we may
20:26 place a lot of value in the amount of
20:30 time we spend on
20:32 something versus the
20:36 output and so if you spend five minutes
20:39 on something and it's really good is it
20:42 really
20:43 good or or should it have taken three
20:46 hours and then it would be really
20:50 good that's where I think this this this
20:52 sort of apology there's a guilt thing
20:54 happening there's an apology thing
20:56 happening where good work is getting
20:59 done fast and therefore it shouldn't be
21:06 valuable and it ain't
21:10 true some painters
21:15 take months to create one
21:19 piece some painters
21:22 take an
21:27 hour
21:30 if the if the paintings affect the
21:33 viewer does the viewer care
21:38 no so but I think we're all going to
21:40 have to confront that I think that's a
21:41 weird one there's a bunch of
21:43 psychological stuff going on with this
21:45 AI stuff I love weo in San Francisco oh
21:47 that's cool AGI taxis I'm
21:57 scared after your show last night I used
22:00 the image generator for the first time
22:02 and Drew so much against the norm far
22:07 out after your live after your show last
22:09 night I used the image generator for the
22:11 first time it's amazing isn't
22:14 it it's just th those first times when I
22:18 first started doing it I started playing
22:20 in pixel mind which was an nft project
22:23 that had stable diffusion in it where
22:24 you could make images and then mint
22:27 them
22:31 it was kind of mind-blowing and then and
22:33 then once I got stable diffusion
22:35 installed and could actually prompt my
22:36 own
22:38 images it's just it's just it is kind of
22:41 a miraculous thing it really is so
22:44 congratulations that's awesome and
22:46 that's so so for those of you that are
22:48 new here I know I'm just I haven't
22:49 really started yet I've just been
22:51 rambling a
22:52 bit I can't stay on too long tonight and
22:56 uh so I figur maybe we just do a little
22:57 chatty chat I'll show something
23:00 maybe but
23:04 um if you take nothing a well else away
23:07 from this
23:09 live it's
23:12 that the I think the worst thing that
23:15 you could do right now with AI is just
23:18 sit on the
23:21 sidelines what's what's kind of [ __ ]
23:24 up and weird is the tools right now are
23:26 not quite good enough
23:28 [Laughter]
23:30 to like really work with them they are
23:33 like some are some use
23:35 cases like I've built an AI tool for my
23:38 company that's really good I've built an
23:40 internal tool I've built an external
23:42 tool they're really good I put together
23:45 a video translation a language
23:47 translation product that I think's
23:49 really good that we can deliver and then
23:50 ultimately I can
23:54 automate but you kind of have to sift
23:57 through all the garbage to get to the
23:59 parts that are good enough to use and
24:00 you have to glue all this [ __ ] together
24:02 and it's just it's a it's a pain in the
24:05 ass right
24:06 now
24:09 and it's moving so fast that that I
24:13 think the most important skill right now
24:14 is like rabid
24:16 curiosity and so what this channel is is
24:22 hopefully a way to kind of
24:25 demystify AI and what it is and you
24:27 don't need a math degree to play with
24:29 this stuff I think that's a big
24:31 misconception the robots are not going
24:33 to kill us not this year anyway um I
24:35 think we have some years before we get
24:37 to those
24:39 scenarios um even the possibility of
24:41 those
24:43 scenarios
24:46 um they're really
24:48 fun they're really
24:50 fun and
24:53 so if you just flip your mindset from
24:56 either you know I'm not using them cuz
24:57 they're not perfect or they've got to be
24:59 perfect and I'm going to will my way
25:01 into this and I'm going to work really
25:02 hard to make them perfect I think
25:04 neither I think both of those are very
25:05 stressful kind of existences so what I
25:09 try to strongly encourage in here is
25:13 play get your ass in a
25:16 community that's into this stuff and if
25:19 you don't have one we've got
25:21 one so oh I should show you that
25:25 yeah um that's actually really
25:27 interesting slide I'll show you that in
25:29 a
25:30 second
25:32 uh the AI salon so if you go to the
25:36 salon. and click on the very top Link at
25:38 the top it's a link tree the top link is
25:41 the AI Salon community and go check out
25:45 that community so this is you can kind
25:47 of think of this like a really cool
25:50 support group for for for people curious
25:53 about AI it's a community of
25:56 people jumping in into the AI adventure
26:02 together and go check it out and if if
26:05 and read about the salon so read go to
26:08 the welcome to the salon page well you
26:10 can read about the salon and you can
26:11 read about our values and really take in
26:14 those
26:18 values and they're not anything scary
26:21 they're things like curiosity like be
26:23 curious be
26:25 generous share what you're learning
26:29 be
26:31 adventurous be
26:33 collaborative be inclusive be
26:37 empathetic exercise gratitude right like
26:45 not because right now we're in a we're
26:48 in a weird time where we're so early in
26:51 generative AI that no one knows what
26:54 they're doing every everyone is a
26:57 beginner
26:59 and there's a lot of angst there's a lot
27:02 of fear there's a lot of
27:04 anger and there's a lot of
27:07 excitement and what's happening is what
27:10 happens on these lives and what happens
27:11 over at the AI salon on a on a daily
27:14 basis is that people are coming together
27:16 and saying oh I tried this and that
27:18 worked and I tried that and that didn't
27:19 work oh here's some news from here
27:21 what's what does that mean oh I don't
27:22 know I think it means this I don't think
27:23 it means that there's all this
27:26 engagement going on it's
27:31 remarkable and if if you're one of the
27:34 people on the planet right now who's
27:36 like holy [ __ ] this AI stuff's amazing
27:40 hey Jenny have you seen this AI stuff
27:43 yeah Bob I don't care about your stupid
27:45 AI stuff okay could you stop talking to
27:47 me about
27:50 AI if you're
27:52 Bob and you can't find anyone to talk to
27:54 about it come hang out
27:57 here and then have fun this group is
27:59 having fun we're producing stuff we're
28:03 people are putting on workshops they're
28:05 just putting on clubs where people go
28:07 hang out and learn
28:10 [ __ ] I talked with Cindy [ __ ] today
28:13 about just sort of transforming a
28:16 project she and I were working on into
28:18 like a club that just makes
28:21 [ __ ] we're going to publish [ __ ] we're
28:23 going to put out video series and we're
28:24 going to do books and we're going to
28:26 just do a bunch of [ __ ]
28:30 what's it going to be on it doesn't
28:32 [ __ ]
28:33 matter what matters is that you're in
28:36 the conversation and that you're
28:37 actively exploring Ai and you're trying
28:39 to figure it
28:41 out and I'm 60 years old against the
28:44 norm 58 here go Gen X back off
28:49 [ __ ] Gen X is in the house and
28:53 we're we're old enough to know that this
28:56 shit's a big deal we're young enough to
28:59 still be ambitious and have our [ __ ]
29:00 together enough to get some stuff
29:03 going and we're wise
29:11 enough to be paying attention to be
29:16 playing we can talk about hardware and
29:19 Matrix operations no nobody okay it's
29:23 Pate Pate works at Google P's got his
29:25 PhD and works at Google
29:28 and uh P's a really good guy Kate Pate
29:31 runs the in in the AI
29:33 Salon there are guilds and there's a AI
29:37 101 Guild run by Peter Kaminsky and
29:39 there's a art and a writing Guild by by
29:43 uh Leah who's the co-founder of the
29:45 salon with me and and Jane Endicott runs
29:48 writing and there's there's a whole
29:50 bunch of guilds and Pate runs the AI
29:52 mechanics Guild and so if you want to
29:54 geek out on stuff like I asked him a
29:56 question today there's I I don't know if
29:58 you've seen it but there's this company
30:01 called gro.com
30:04 gq.com
30:06 not Gro that Elon
30:10 Musk put up on Twitter X Twitter
30:15 X but
30:19 grq and they're they're a chip design
30:22 company I have a 700 page book called
30:25 Matrix compilations let's go over it
30:27 actually let's dump that in Gemini 1.5
30:31 and uh and and do all sorts of cool [ __ ]
30:33 with it
30:36 um but what grock is doing if you go
30:39 there if you go to gro.com right now you
30:41 can you can choose either llama 70
30:44 billion parameter model or the
30:46 mixl 8X what two open source models that
30:50 are that are big
30:52 ones and that thing is is returning 500
30:56 tokens a second so it's about 350 words
30:59 a second so if you're if you're on chat
31:02 GPT if you played with chat GPT 4 you
31:05 know sometimes it's a little slow and
31:06 you like watch it type across the screen
31:09 kind of like if you if you're old enough
31:11 to remember dial up modems remember when
31:13 they were so slow that you could watch
31:14 the text type across the screen that was
31:16 coming in over the Internet that's kind
31:19 of what CH chat gp4 is like right now go
31:22 play with that grock thing and you start
31:24 to get a glimpse into where all of this
31:26 stuff is headed where we're
31:28 headed where it's just going to be
31:31 instant it's going to be everywhere it's
31:33 going to be instant and it's going to be
31:36 good and we ain't there yet so Now's the
31:39 Time To Play Now's the Time to to
31:41 understand the difference between well
31:43 what's what's uh open AI doing and
31:46 what's Google doing and what are tpus
31:48 and gpus and how is grock doing what
31:50 they're doing and how is that better
31:52 right and so I asked Pate today in the
31:54 AI mechanics Guild I'm like do you know
31:57 what they're doing doing and how that's
31:58 different from a TPU cuz Pate was kind
32:01 enough to explain to me the difference
32:03 between a TPU a GPU and a
32:11 CPU and so this the salon is just an
32:14 amazing resource of people who are not
32:20 [ __ ] who are generous who are
32:22 curious who are
32:24 adventurous who all have a a humility
32:27 about about
32:28 them with the exception of myself of
32:31 course have a humility about
32:36 them that's really
32:41 refreshing and everyone in there
32:44 knows that they're not an
32:47 expert and yet they're willing to share
32:50 what they know they're willing to teach
32:51 they're willing to be
32:54 leaders even though that might be the
32:56 last thing they think they want to do
32:59 they're just naturally leading because
33:02 someone comes in and says I don't get
33:03 this and if they get it they're like
33:04 well here's how it works and why don't
33:06 you come to Ann Murphy's GPT jam on
33:08 Saturday and you can learn about this
33:10 stuff oh okay that's this group really
33:12 is
33:15 amazing so go
33:17 there and here's my request my first
33:20 request is go to go to welcome to the
33:22 salon read about the salon read about
33:24 our
33:25 values the next thing I want you to do
33:28 is in the community section the very
33:30 first
33:31 link is introduce yourself my request is
33:35 go there sign up and introduce
33:38 yourself our second value is generosity
33:41 go be generous and generously I know
33:44 introducing yourself especially online
33:47 is [ __ ] up and
33:49 weird but go introduce yourself here's
33:52 who I am here's what I'm up to here's
33:54 where I am with
33:56 AI
33:58 here's my hopes and dreams here's my
33:59 fears doesn't like anything let us know
34:02 who you are and what will happen is
34:04 people within this community will go hey
34:06 welcome oh you're a newbie you should go
34:09 check out Peter kaminsky's 101 Guild or
34:11 you should go here you should go there
34:13 come to this come to
34:15 that jump into Show and Tell show some
34:17 images you've
34:19 made the people that show up to this
34:21 channel regularly they're called the
34:23 Irregulars we have our own club here so
34:25 if you scroll to the bottom of
34:28 this it says AI L AI learning lab
34:34 Irregulars come hang out so when when
34:37 I'm doing these lives so I do these
34:38 things nightly seven nights a
34:41 week seven nights a week and there's a
34:44 bunch of people who need to have their
34:47 heads
34:48 examined that join me
34:52 nightly it boggles my mind that it that
34:55 it happens it boggles my mind
35:00 but what we're all
35:03 doing is keeping in the conversation
35:05 like why do this am I teaching anything
35:08 here that you couldn't learn from a
35:10 YouTube video
35:12 No in fact a lot of times what I'm
35:15 teaching here is I'm pretty [ __ ]
35:18 clueless what are your qualifications
35:20 sir qualifications I have none good day
35:23 to you sir Cheerio pip
35:26 pip
35:31 but that sense of adventure and that
35:33 sense of being
35:36 compelled to be playing with these
35:40 tools that are like the most profoundly
35:43 powerful tools in the history of
35:46 humanity they really
35:48 are like you get to tap into Humanity
35:52 when you
35:53 use chat GPT and tools like it
35:57 it was trained on the p in GPT stands
36:01 for
36:02 pre-trained and what that means is they
36:05 they took this thing called a
36:07 Transformer that's just a piece of
36:09 technology that Google
36:11 invented and what it allowed is it
36:14 allowed for these machine learning
36:16 models to be trained on massive amounts
36:18 of data on many many many many
36:22 many gpus Graphics processing units so
36:26 basically graphics cards from gaming
36:29 computers they're fancier versions of
36:31 that that's essentially what they
36:35 are and
36:40 so companies like open AI said hey if we
36:44 can train these things on any data and
36:47 massive amounts of it you know what's a
36:49 really good massive amount of data all
36:52 the [ __ ] that was stuck on the internet
36:54 for the past four decades and so open AI
36:57 in particular is one that's been really
36:59 good at going out and getting as much of
37:01 the public internet as they could get
37:03 their hands on embedding it you know
37:06 using this special way of putting it in
37:08 the Box
37:11 basically and so when you do a chat GPT
37:14 search you know what they call these
37:15 things are chat Bots right and it's oh
37:18 the Bots are going to kill us and you're
37:19 talking to a robot
37:22 and no what you're doing is you're
37:25 interacting with with what's effectively
37:27 a calculator it's a probability
37:30 engine but what it's making
37:32 recommendations on is the collective
37:35 knowledge of humanity that open AI was
37:39 able to scrape off the internet for the
37:41 past four five or six
37:43 years and so what you end up getting is
37:46 you end up getting you end up having
37:47 interactions that
37:50 feel remarkably more human
37:54 than robotic and it's one of the
37:58 counterintuitive things about large
38:00 language models for me one of the
38:04 most oh thank you very much in investor
38:07 friendly agent thank you Kyle I learned
38:09 I learned the most from you thank you so
38:10 much I appreciate
38:17 that it's funny when when I first
38:20 started this Channel and I first started
38:22 I would show people how to use chat GPT
38:23 and I'd go please do this and thank you
38:25 and and people would to make fun of me
38:27 oh you're saying please and thank you to
38:29 the
38:32 robot and I used to kind of apologize
38:34 for it or make fun of it and say yeah
38:36 I'm a dumb dumb and and what I've
38:37 realized is that no no like that feels
38:42 appropriate when you interact with a
38:43 large language model it's actually much
38:46 closer to working
38:48 with an incredibly well-educated
38:53 very inexperienced
38:56 intern
38:57 remotely right so thanks to co we
39:00 understand how to work together remotely
39:02 right so you know how you've you work
39:04 with people every day you've never met
39:06 them and like you say hey could you go
39:09 do this research for me and couple hours
39:12 go by and they come back and they give
39:13 you some [ __ ] and then you're like well
39:16 that okay technically you did what I
39:18 asked but clearly you need a little more
39:20 Direction so please go find this kind of
39:23 information which is the subset of what
39:25 I asked for so that would be great go do
39:26 that and then after seven of those
39:28 interactions you might have the research
39:30 you
39:31 need that's the identical experience of
39:35 of playing with chat
39:37 GPT ah go give me this kind of stuff
39:40 yeah here's a bunch of stuff oh but
39:42 instead of it being 2 or three hours
39:44 it's 2 or three minutes it's like
39:47 there's all your research you're like oh
39:49 that wasn't what I wanted I needed to be
39:51 the subset of this okay here you go and
39:54 she like bang bang bang bang but it's
39:56 like like you're
39:59 interacting with an
40:01 intern and so I find myself just
40:04 instinctively going please thank you
40:07 when I first started working with with
40:09 large language
40:11 models I had this weird so so in in this
40:15 channel I talk a lot about being neuros
40:17 spicy and having
40:19 ADHD um I'm also codependent as [ __ ]
40:23 right and so I've got one of those
40:26 personalities where I'm capable of doing
40:28 [ __ ] but I'm also old enough to know you
40:30 got to delegate some [ __ ] right so you
40:32 you say okay here go do this work for me
40:35 and then the person goes off and they do
40:36 the work and what they come back with is
40:39 a horrible steaming pile of [ __ ] and
40:42 you're like Ah that's hey thanks that's
40:44 really great uh could you uh could you
40:47 tweak it to be a little less um you know
40:49 horrible you don't say it in those words
40:50 but you send them off to do their work
40:53 and then they come back and it's it's
40:54 just not good and then you realize well
40:56 I guess I'm doing doing this on my own
40:58 because for me cuz I'm codependent I'm
41:01 like I don't want to tell him like the
41:02 fifth time that it just sucks so I just
41:04 say [ __ ] it I just suck it up and I do
41:06 the work right so when I first started
41:10 using chat
41:11 GPT I would ask it to do something it
41:13 would give me a piece of [ __ ] I'd say do
41:15 it again it would give me back a piece
41:17 of [ __ ] I'd say do it again it would
41:18 give me back a piece of [ __ ] and there
41:19 was a point in there where i' start to
41:21 feel those pangs of of codependent
41:24 anxiety like oh well I don't want to
41:26 offend it and I was like oh
41:30 [ __ ] it doesn't have feeling like it
41:32 doesn't give a [ __ ] but it's been it's
41:35 been programmed to to like give you what
41:37 you want so it's always like a golden
41:39 retriever like sure boss how can I help
41:41 you yeah I'll go do that for you this is
41:43 horrible can you try again sure how's
41:45 this oh that's really horrible try again
41:47 how's this and it's just dinging
41:52 dinging so it it is a it is a
41:55 surprisingly human
41:57 experience working with
42:01 AI Kyle when they created the internet
42:04 do you think they intended it to be the
42:06 foundation of
42:10 AI yeah I think I think artificial
42:12 intelligence was
42:15 coined in
42:18 1952 4 hang
42:21 on let's go ask Chad je oh we'll go ask
42:25 Gemini I got Gemini up
42:28 when was the
42:30 term
42:32 AI
42:34 invented I don't I guess you don't
42:36 invent a term but
42:37 whatever
42:40 1955 first coin at a workshop held at
42:46 Dartmouth and I think I think in 1955
42:50 they thought you know it'd be it'd be
42:51 seven or eight years till they had
42:53 artificial
42:54 intelligence um
42:58 so I don't know that I don't know that
43:00 the inventors of the internet thought it
43:02 would be the foundation of AI
43:04 necessarily but AI predated the
43:07 invention of the internet
43:12 so so they were likely aware that they
43:15 could you know connect computers
43:18 together and one of the very first
43:20 things that they did was try to
43:22 replicate the neural Pathways you know
43:26 between neurons in the brain the neural
43:29 network is literally computer scientists
43:32 saying well if we connected these
43:34 together like neurons in the brain I
43:36 wonder if they would do the same thing
43:37 and maybe if we connect like 10 of them
43:39 together we can get a thought going back
43:42 and
43:43 forth um it it
43:47 is there there's just tons and tons and
43:50 tons of innovations that have happened
43:52 but I mean I think the the people that
43:56 invented the inter internet I guess it
43:58 was DARPA right it was the
44:04 military I would think back then it was
44:07 probably as simple as it's probably
44:09 cheaper if we could get if we could get
44:11 data from this computer to that computer
44:13 without having to truck it across
44:15 country it would probably be cheaper and
44:18 faster and for military applications
44:20 that's probably good so I don't know I
44:21 don't know if if when they invented it
44:23 they were thinking about AI but um
44:27 but yeah computer scientists have been
44:29 fascinated about this forever and when
44:31 when was Turing um when did Turing make
44:36 his
44:39 test was 70s or was that earlier oh
44:45 1950 Alan Turing proposed the Turing
44:48 test in
44:51 1950 a paper called Computing machinery
44:54 and intelligence so the the Turing test
44:57 if you don't know
45:00 is well I'm paraphrasing it I don't know
45:04 it exactly so if you know it exactly
45:06 just don't yell at
45:10 me fraking fact monsters monsters um the
45:15 Turing test is basically if you have an
45:18 interaction with a computer can you tell
45:21 if there's a human being on the other
45:22 side or
45:23 not and the I use chat
45:27 GPT like the day I used chat GPT the
45:31 first thing that popped in my head was
45:34 the Turing test is
45:36 over because if you've ever used Alexa
45:39 or
45:41 Siri and you've been talking to it and
45:44 things have been fine and then you ask
45:45 it something that's fairly obvious and
45:47 it's like I'm I'm sorry Dave I'm afraid
45:49 I can't do that um you realize oh these
45:54 machines are stupid these just stupid
45:56 dumb Dum machines and when you start
45:59 interacting with chat GPT you realize oh
46:02 there's something different here I have
46:03 you seen Pi if you have not played with
46:06 pi so this is my
46:09 iPad and there's an app called Pi it's
46:11 from a company called
46:14 inflection that's from a company called
46:16 Story vine which I'm the CEO of okay in
46:20 pi is just like a chat GPT kind of app
46:24 except the thing that makes p special
46:27 it's been designed for
46:29 conversations so it can hear and it can
46:32 talk so when you open Pi you push this
46:35 little phone
46:40 button here's a question for you do you
46:43 think you passed the Turing
46:51 test well it's a bit tricky to answer
46:53 that Kyle passing the touring test also
46:55 known as the imitation game would
46:57 require a computer program to
46:58 successfully fool a human into believing
47:01 that it is a human and not a computer
47:03 this is typically achieved by having a
47:05 human conversing with both a human and a
47:08 computer without knowing which is which
47:11 now I like to think that I can engage in
47:13 natural and humanlike conversations but
47:15 whether I'd be able to fully fool a
47:17 human would depend on how perceptive the
47:20 human is plus the touring test is a bit
47:22 of an old concept and doesn't
47:24 necessarily account for all the
47:25 advancements in conversational AI like
47:28 myself so I suppose my answer is that I
47:31 might not be able to fool everyone but
47:33 I'm constantly learning and improving my
47:36 conversational abilities after all
47:39 there's always room for
47:41 growth shall we play a
47:48 game oh I love games KL let's do this
47:51 what kind of game did you have in mind a
47:52 trivia quiz a word game or maybe
47:55 something totally out of the box box I'm
47:57 game for anything and I'm sure we'll
47:58 have a lot of
48:00 fun thermonuclear
48:07 war uhoh I'm all for a good game Kyle
48:09 but I'm not about to help facilitate a
48:11 thermonuclear war even in a joking
48:14 manner how about we choose a less
48:16 potentially world-ending game to play
48:19 that was a joke from the movie War Games
48:22 with Matthew broadrick you're missing
48:24 your cultural references Z
48:34 uh you had me fooled there Kyle you're
48:36 right I was so anyway so that's
48:38 happening um investor friendly agent how
48:40 do I learn to talk to chat GPT how do I
48:42 ask questions in the right way okay
48:45 great
48:46 question
48:49 um one of the terms that emerged in 2023
48:53 well actually this is a term that
48:55 emerged well before 2023 but got popular
48:58 in
49:00 2023 is the term prompt
49:03 engineering and I [ __ ] hate that term
49:07 I [ __ ] hate it with a
49:12 passion
49:15 why because I think it's exclusionary
49:19 and I think
49:20 that when people hear you need to get
49:23 good at prompt
49:25 engineering
49:26 if they're not Technical and if they
49:28 haven't gotten a degree from Stamford
49:30 it's
49:33 intimidating and the fact of the matter
49:35 is is that as these tools get more and
49:38 more sophisticated you don't have to be
49:41 as sophisticated in your prompting
49:46 so how to get good answers out of chat
49:49 GPT is um there's there's essentially an
49:54 infinite number of ways to do do it for
49:59 one which doesn't answer your question
50:01 you're probably like but kle that
50:03 doesn't answer my
50:05 question how do I get it to do good
50:07 stuff
50:10 well
50:14 um I think the the biggest tip so so if
50:18 if you want to go learn structured
50:21 prompt crafting you can do that so you
50:24 can say the first part of the prompt
50:26 should be you tell
50:28 the um the large language model what its
50:31 role is okay you're going to act as an
50:33 accountant and then the next thing is
50:34 you give it your context and the next
50:36 thing is you give it your subject and
50:37 the next thing is you give it your
50:39 action and next thing is you give it the
50:40 output type you can absolutely do that
50:45 as as someone with ADHD that approach
50:48 makes me want to throw my computer it
50:51 really does I'm just like I just I I
50:53 can't I've tried it a couple of times it
50:56 can be impressive but it's just really
50:59 annoying
51:01 um
51:07 okay one trick and it's a good one
51:11 actually is you can ask chat GPT to give
51:14 you a prompt for chat
51:16 GPT you can say to chat GPT hey I'm
51:20 trying to solve this kind of problem how
51:22 should I ask you to solve that problem
51:25 in a way that'll give give us the best
51:28 results and you'll be blown away what
51:30 it'll come up with and and it'll
51:32 probably actually just answer your
51:34 question in asking it what prompt you
51:36 should ask it to ask it the question
51:38 okay so that's one thing
51:40 um but here's the most important here's
51:43 the most important thing in my opinion
51:45 for how to how to think about engaging
51:48 with large language
51:49 models is you have to unlearn what
51:52 Google has taught us for 20 years what
51:55 Google gole has taught us for 20 years
51:57 is you put in a concise prompt and you
51:59 get back an answer and you're happy with
52:01 it leave us alone we're done
52:03 here that's not how large language
52:06 models
52:10 work large language models rather than
52:12 needing a minimal prompt actually need a
52:16 lot of
52:17 context and so the more you tell it the
52:20 more the more you say so when you say to
52:22 a large language model I want you to act
52:24 like um an advert iing copywriter who
52:27 specializes in you know writing ad
52:30 effective ads for local businesses it
52:33 actually works because it kind of
52:35 Narrows the knowledge set that the large
52:38 language model is drawing from when
52:40 you're inter interacting with it and so
52:42 if you start out and say I want you to
52:44 act in this certain role and then you
52:47 say um I'm going to be talking about my
52:49 small business that's at this stage and
52:51 it's called this and here's what we
52:54 do and then chat GPT will probably come
52:56 back and go that sounds like an awesome
52:58 business how do you want to work
52:59 together and you know the most important
53:02 thing is rather than thinking about I've
53:05 got to create a prompt to get an answer
53:07 think about it more like a conversation
53:10 that's the biggest hack with chat GPT or
53:12 with all with all the large language
53:13 models it's a conversation and the more
53:17 context you give it and the more
53:18 feedback you give it the better it will
53:20 get and and what's what's
53:21 counterintuitive is the more you
53:23 challenge it the the better tends to get
53:27 because you're you're kind of narrowing
53:29 the information set so you get a lot
53:31 less what are called hallucinations
53:33 lies because you've kind of narrowed the
53:36 conversation with all this context and
53:38 then just ask go back and
53:40 forth if it gives you something you
53:42 don't
53:43 like tell
53:45 it tell it why you didn't like
53:49 it hey that seems kind of generic and
53:52 fluffy and I wouldn't believe anyone
53:55 that talked like
53:57 that I'll show you one other trick that
54:00 is I'm going to I'll show you a trick
54:01 right now that most people don't know
54:04 about because a lot of people are not
54:07 what what a lot of people are doing is
54:08 they're thinking oh chat GPT it's just
54:10 like fancy Google I get it it's a boxia
54:12 type words in it stuff comes out I got
54:14 it I'm good that's all I need to know
54:16 we're good we're good here we're good
54:18 people we don't we don't want to dig too
54:21 deep and actually understand what this
54:22 thing is so we'll just keep treating it
54:24 like the Google BX
54:27 there's a thing called custom
54:28 instructions which which can be very
54:30 very powerful so particularly if if you
54:34 are if you write in a certain style or
54:37 or if how chat GPT responds to you
54:40 drives you [ __ ] Bonkers so part of
54:44 the reason chat GPT acts like a golden
54:47 retriever is that someone at open AI
54:50 programmed or not programmed told chat
54:53 GPT hey when you respond to people be
54:56 super positive and and reinforce that
54:59 what they said is awesome or whatever
55:01 they said I don't know exactly what they
55:03 said but they said something like that
55:05 so when it talks back to you it's super
55:09 Pleasant there's times when I don't want
55:12 super Pleasant and I don't want fluffy
55:14 words and I don't want positive
55:15 reinforcement I just want the [ __ ]
55:18 answer right so I'm going to show you
55:20 something that a lot of people don't
55:22 know about so I'm going to go to chat
55:24 gpt3 .5 which is the free version so you
55:27 can do this with the free
55:29 version and you're going to go to your
55:31 name you're going to click on it and
55:33 then you're going to go
55:35 to is this it oh they changed it that's
55:40 funny this used to be called custom
55:42 instructions and now it says customize
55:43 chat GPT but that's okay that's this is
55:46 a better name look so someone from
55:48 someone with a marketing degree got to
55:51 got got in here and actually made
55:52 something better yay um customize chat
55:56 GPT so what you can do here is you can
55:59 tell chat GPT kind of who you are or who
56:03 who you want it to be and how you want
56:05 it to respond so I'm going to turn this
56:07 on and then what happens
56:09 is once
56:11 you wait it told you how much browsing
56:15 and Analysis time you had wait did
56:21 it what did where what are you talking
56:24 about
56:30 I don't know what you're talking about
56:31 Danielle what did I
56:34 miss it told you how much browsing and
56:36 Analysis time you had
56:45 huh anyway back to back to customiz and
56:47 GPT so I'm going to turn this on so once
56:49 you turn this on however you tell it to
56:52 act every every chat from this point
56:55 point forward um will be um will be like
57:01 this okay so so what I wrote this was a
57:05 good way a good okay and this is
57:07 actually I'll often do silly examples in
57:11 here
57:15 um because silly examples make it very
57:20 apparent how the changes that you you've
57:22 asked it to make like if they're working
57:25 and it's also fun right like have fun
57:27 with this stuff so the custom
57:29 instructions I wrote before I'm the CEO
57:31 of a comedy club and I can't take
57:33 anything seriously but I'm absolutely
57:35 committed to making everyone laugh in
57:37 every interaction so that's kind of who
57:39 it is and then how would chat G how
57:42 would you like chat GPT to respond no
57:44 matter what someone asks you make sure
57:46 that you're responding with humor and
57:48 puns and a New York Comics Edge even
57:51 when writing business correspondence
57:53 your goal should be to make someone
57:54 smile no matter what the news is and how
57:57 it's being delivered okay so I'm going
57:59 to save that I'm going to make a new
58:02 chat and I'm going to say
58:07 please write a
58:11 professional
58:13 business letter letting my hostess know
58:19 she no longer works
58:24 here
58:26 this is cruel this is Cruel Cruel
58:28 Cruel uh Kyle it told you how much time
58:31 you had to browse and do analysis that's
58:35 new oh wow okay I didn't see that I
58:36 missed that okay so this professional
58:40 business letter dear hostess's name I
58:43 hope this letter finds you well I'm
58:44 writing to inform you of a decision
58:46 that's been made regarding your
58:47 employment regrettably due to recent
58:54 restructuring
58:58 huh it didn't
59:01 work did I these are turned on aren't
59:09 they let me start a new chat that's
59:12 really weird
59:14 um what is
59:18 the
59:19 average uh
59:22 aage
59:24 speed
59:26 of an unladen
59:31 swallow ah the classic question we
59:33 talking African or european you see the
59:36 average speed can it's a tricky business
59:38 but catch the jet
59:39 stream this does not seem to be working
59:42 let me let me try let me change
59:45 this um no matter so uh okay let's
59:49 see for the New York Comics Edge
59:54 um
59:58 make sure the first thing you
1:00:04 say is
1:00:07 what famous
1:00:10 joke
1:00:12 their
1:00:15 uh question reminds you
1:00:24 of let's see if we can get it to do that
1:00:27 so uh let's see
1:00:32 um help me pick an outfit to look good
1:00:35 on
1:00:36 camera ah the Eternal question what to
1:00:38 wear for a photo shoot well remember
1:00:41 what they say in the Fashion World when
1:00:42 did the tomato turn red because it saw
1:00:45 salad dressing okay it seems to be
1:00:46 working now that they they've changed
1:00:49 something here so one of the things that
1:00:50 you'll notice um when you're using these
1:00:53 tools is sometimes they work brilliantly
1:00:55 sometimes they just don't work um that
1:00:57 was weird that it just completely
1:00:59 ignored its custom instructions so that
1:01:02 might be that they're tweaking something
1:01:03 there you're on 3.5 yeah but it still
1:01:05 shouldn't even if I'm on 3.5 the custom
1:01:08 instructions work let me let me try it
1:01:10 here
1:01:12 um let's see write a
1:01:17 professional Biz in this
1:01:20 letter letting the
1:01:23 hostess no she no longer works
1:01:31 here oh this one's like the classic joke
1:01:34 I told my boss three companies were
1:01:37 after me and I
1:01:39 needed I needed a raise to say he asked
1:01:41 okay here's the here's the letter change
1:01:43 in employment status I hope this finds
1:01:45 you well before diving into the main
1:01:46 purpose of yeah here this is good before
1:01:50 diving into the main purpose of this
1:01:51 letter let's start with a light note did
1:01:53 you hear about the restaurant on the the
1:01:55 mood great food no atmosphere speaking
1:01:57 of atmosphere we've always appreciated
1:02:00 the unique energy and flare you brought
1:02:02 to the comedy club making it brighter
1:02:04 place for both of us however as in every
1:02:07 great comedy routine timing is
1:02:08 everything and sometimes the show must
1:02:10 go on with unexpected changes to the
1:02:12 cast so the heavy heart so so this
1:02:14 actually worked um I can go back into my
1:02:17 custom instructions and I can say things
1:02:19 like um let's see you have no
1:02:23 patience
1:02:26 and just want the
1:02:32 answers as
1:02:35 efficiently as
1:02:38 possible uh
1:02:42 possible um do
1:02:46 not it doesn't do good with do not do
1:02:50 things it it does better if you say do
1:02:52 do this so resp
1:02:56 respond uh to every
1:03:03 inquiry with a
1:03:07 concise single sentence
1:03:14 answer followed by short bullet
1:03:21 points
1:03:23 if uh bullet
1:03:27 points
1:03:29 only use bullet
1:03:32 points if
1:03:35 the simple
1:03:38 answer isn't
1:03:40 enough save okay so now we'll start over
1:03:44 and we'll
1:03:49 say
1:03:52 um how
1:03:56 many
1:04:00 elephants uh tall is the oh no that's
1:04:05 that's let's see how many let's see
1:04:10 um what is the
1:04:16 difference
1:04:17 between an
1:04:20 elephant and a
1:04:23 mouse
1:04:26 the main difference between an elephant
1:04:28 and a mouse is their size and physical
1:04:29 characteristics three bullet points done
1:04:32 no fluff no
1:04:34 [ __ ] custom instructions changed
1:04:37 it's been acting up all day okay all
1:04:40 right good so anyway
1:04:43 so back to the original question how do
1:04:46 you talk to these things just talk to
1:04:48 these things like you're talking to a
1:04:49 person give them feedback um play with
1:04:53 custom instructions or customize chat
1:04:55 GPT um it is it's a very powerful tool
1:05:00 especially if if in trying to interact
1:05:02 with these things they're like it just
1:05:05 doesn't sound like what you want it to
1:05:07 sound like you can fix
1:05:09 that
1:05:11 um so I hope that helps I thought chat
1:05:15 GPT was not working well today I haven't
1:05:17 used it
1:05:18 today so I don't know I instead played
1:05:20 with acrobat AI how's acrobat AI I saw
1:05:23 my my co-founder sent me a note on that
1:05:25 Suzanne I'm curious I didn't get a
1:05:27 chance to play with it cuz I've been
1:05:30 busy I've been busy at work and work's
1:05:32 hard did you know that if you start a
1:05:35 business that you got to do all the work
1:05:38 it's so
1:05:40 hard it
1:05:43 sucks Be Your Own Boss set your own
1:05:46 hours everything's fine so long as the
1:05:49 hours are 24 hours a day seven days a
1:05:53 week
1:05:57 you can't turn it off you take
1:05:59 everything personally this is life is an
1:06:03 [Laughter]
1:06:08 entrepreneur was that helpful tell me if
1:06:12 that was
1:06:14 helpful and Murphy fundraising I've
1:06:17 never needed to oh AI plus fund rating
1:06:20 about 85% uses custom instructions if
1:06:23 I'm not using a GPT there you go
1:06:25 remember back in the day when it would
1:06:27 answer questions at the end yeah I
1:06:30 something okay so here's have you been
1:06:34 boss boying today I have been I have
1:06:37 been boss boying today thank you for
1:06:39 asking yes I had I had a uh a meeting
1:06:43 with our investment banker I had
1:06:48 uh oh I came up with a new slide I want
1:06:50 to show
1:06:51 you have you have you seen hey Jen's
1:06:55 personalized video it's [ __ ]
1:06:58 nuts let me show you this
1:07:08 um all
1:07:11 right so hey Jen's got this thing called
1:07:14 personalized video so you record a video
1:07:16 and then you can personalize it for
1:07:17 other people now what you do is I like
1:07:22 acrobat AI can summarize any type of
1:07:24 docs ask questions write emails now I
1:07:27 can actually use
1:07:31 AI at work and we don't have to use chat
1:07:34 GPT co-pilot oh that's cool oh that's
1:07:37 that's really
1:07:38 cool oh because it's an acrobat and you
1:07:40 work at Adobe nice okay all right
1:07:43 so here we have a little spreadsheet and
1:07:46 the spreadsheet is the person's name and
1:07:49 something I remember about them so Ryan
1:07:51 we went fishing in the middle of
1:07:52 November Justin
1:07:55 um we're talking about authentic
1:07:56 storytelling at digital Pharma East and
1:07:58 Emma we used to work together at
1:08:01 Ogie so I
1:08:03 record a message
1:08:07 once and then this
1:08:10 system how's it going hey I was just
1:08:14 sitting here thinking you remember that
1:08:16 time when we went fishing in the middle
1:08:17 of November anyway I was just thinking
1:08:20 about you I was thinking it might be
1:08:23 good for us to catch up hope you're
1:08:25 well hey Justin how's it going hey I was
1:08:29 just sitting here thinking you remember
1:08:31 that time when talked about authentic
1:08:32 storytelling at digital Pharma
1:08:35 East hey Emma how's it going hey I was
1:08:40 just sitting here thinking you remember
1:08:42 that time when we worked together when
1:08:43 you were at
1:08:45 Ogie so this is effectively maale merge
1:08:48 for video
1:08:50 [Laughter]
1:08:53 greetings and and you know they're a
1:08:55 little monotone they're a little
1:08:57 pronunciation weird they're a little
1:09:00 glitchy but they're pretty [ __ ]
1:09:03 close and if someone doesn't know me if
1:09:06 I'm cold sending out
1:09:09 emails pretty something and the pricing
1:09:12 on this is 10 cents per
1:09:16 variable so so the way you do this it's
1:09:19 pretty
1:09:21 cool
1:09:23 um
1:09:25 you record your
1:09:29 video and then it creates a transcript
1:09:33 and then you highlight words in your
1:09:34 transcript to be variables so you like
1:09:37 so when I say hi Bob I highlight the
1:09:39 word Bob and then I say turn that into a
1:09:41 variable and then I say you know here's
1:09:44 what we did did together that voice
1:09:46 sounds like you were about to say it's
1:09:49 about your husband and he's been in an
1:09:53 accident it totally does sound like
1:09:56 that remember that time we went fishing
1:09:59 in
1:10:00 November oh that's yeah that's good
1:10:03 that's that that's the voice so the
1:10:05 voice isn't quite there yet
1:10:07 but six months a
1:10:09 year so so you've got to
1:10:12 assume you have to
1:10:15 assume that what I just showed you is as
1:10:18 bad as it's ever going to
1:10:22 be it's as bad as it's ever going to
1:10:25 be so it just gets better from here
1:10:28 right
1:10:30 so so anyway so you so you define these
1:10:34 variables and then you basically just
1:10:35 point a CSV at it with a bunch of
1:10:38 different names and [ __ ] to talk about
1:10:42 you know it could be here's the company
1:10:44 they work
1:10:45 for it could be a memory like that it
1:10:48 could be hey the last time we talked
1:10:51 here's what was going on with you it
1:10:53 could be how are your kids
1:10:57 whatever acrobat a cost I assume it's
1:11:00 part of Creative Cloud the way they're
1:11:02 handling all the AI Integrations at
1:11:04 Adobe right now is they're just they've
1:11:06 got
1:11:10 Firefly I they've got Firefly and Adobe
1:11:14 Express I think are serving as the
1:11:16 experiment grounds for the public stuff
1:11:19 and then as they're taking things that
1:11:21 they're incorporating they're rolling
1:11:23 those into the products I think they're
1:11:24 just part of the Creative Cloud Suite I
1:11:27 don't think it's extra but Suzanne do
1:11:29 you know if it's extra I assume it's not
1:11:32 but you know their Creative Cloud is
1:11:34 [ __ ] expensive so it's it's
1:11:38 monthly it's monthly cost is
1:11:41 expensive I've been using a similar AI
1:11:44 personalized voice for custom video
1:11:46 emails for the past week star for first
1:11:51 name oh interesting so that's that's Jim
1:11:54 R so Jim is it working have you gotten
1:11:56 any emails back about don't send me that
1:11:59 weird AI [ __ ] I assume not because if
1:12:03 you're
1:12:05 looking so so first of all if you don't
1:12:07 know
1:12:08 me I just might talk like a [ __ ]
1:12:11 weirdo right some people just talk
1:12:13 monotone and this is how they talk and
1:12:17 hey remember when we talked about that
1:12:20 thing at the digital Pharma East some
1:12:23 people are just weird ass and talk like
1:12:25 that so if someone doesn't know you they
1:12:28 wouldn't necessarily know that that's
1:12:29 not real now anyone that's seen AI
1:12:32 probably will but most people haven't
1:12:35 most people haven't used chat GPT like
1:12:41 80% it's [ __ ]
1:12:44 crazy so you get an email with with Jim
1:12:47 Ross saying hey how's it going and Jim's
1:12:49 got a good bubbly personality so you
1:12:51 know probably animates him pretty well
1:12:55 really
1:12:57 interesting this is inside hey Jen yeah
1:13:00 this is something they
1:13:01 launched month ago and I just got an
1:13:04 email for it today so they might have it
1:13:06 might have been in like beta testing
1:13:09 before and now it's rolled out wide the
1:13:11 cost on it is 10
1:13:13 cents per
1:13:15 variable per
1:13:17 video so the video can be as long as you
1:13:19 want it if you've got you know three
1:13:21 little variables in there it's 30 cents
1:13:23 per video
1:13:24 I think that's their I think that's
1:13:26 their pricing cuz their normal pricing
1:13:28 is like a dollar a minute for just
1:13:31 normal voice synthesis or translation
1:13:34 stuff a dollar a minute something like
1:13:37 that yeah a dollar something like that
1:13:39 and they knew that if you're going to be
1:13:40 sending out lots of these things that
1:13:42 you you couldn't be doing a dollar a
1:13:43 minute so they made it 10 cents per
1:13:45 variable
1:13:47 so
1:13:50 reasonable I thought he was still firing
1:13:52 someone
1:13:54 which platform is that it's called Data
1:13:57 it's called hey Jen it's called Data I
1:13:59 was reading the next post the data there
1:14:02 the data is just from a CSV file so you
1:14:04 can any anywhere where you've got so
1:14:07 Salesforce would be where you'd get your
1:14:09 data right so you'd go into Salesforce
1:14:12 you'd pull the list of prospects you
1:14:14 want to send those videos
1:14:16 to and then in sales force you would
1:14:19 maybe have a a personalized note
1:14:22 field or the last thing that you talked
1:14:25 about or their company name or whatever
1:14:28 it might
1:14:31 be just joined what was the app that you
1:14:34 just showed the app that I just showed
1:14:36 was hey Jen's personalized
1:14:40 video enjoy the Jank there's a lot of
1:14:42 Jank in it right now is AI following the
1:14:45 one theory that computers double in yada
1:14:47 y every 18 months it's going faster than
1:14:51 that it's going faster than that we're
1:14:54 we're as David I just watched the
1:14:57 there's David Shapiro's latest video is
1:14:59 really good he talks about us being at
1:15:00 the knee of the curve
1:15:03 so so a
1:15:09 linear a linear growth is is like that
1:15:13 right where it's it's linear like you're
1:15:17 here and then it's this much better and
1:15:19 then it's this much better this is what
1:15:21 humans expect
1:15:25 um an exponential
1:15:31 curve is like this where where it's
1:15:35 going along you know at one growth rate
1:15:38 and then all of a sudden it just takes
1:15:42 off and the knee of the curve is the
1:15:45 point at which it starts to to make that
1:15:47 hockey stick up and and David Shapiro
1:15:50 thinks we're right at the knee of the
1:15:52 curve right now
1:15:54 he also thinks that we're going to have
1:15:56 AGI within s months and so AGI stands
1:15:59 for artificial general
1:16:03 intelligence pay pay y equals a ax plus
1:16:08 b exactly that's exponential curves for
1:16:10 you right
1:16:11 there you know like I said in theater
1:16:14 School hey why don't you stop being so y
1:16:19 = ax+ b you
1:16:22 tool
1:16:24 um anyway human beings do not know how
1:16:29 to process exponential growth so let me
1:16:33 give you an example true exponential
1:16:34 curves are
1:16:38 bananas P P's making math jokes and
1:16:42 they're actually funny I like it it's
1:16:47 solid oh good lord what was I just
1:16:50 talking about humans don't know how to
1:16:52 deal with EXP exponential
1:16:55 growth the banana joke threw me Pate
1:16:59 that was pretty
1:17:00 good
1:17:13 um I don't know what I was going to say
1:17:16 I had a good point I lost
1:17:22 it well anyway it it kind of looks like
1:17:25 this shit's G to get really weird in
1:17:27 2024 it's already starting to get weird
1:17:30 if you haven't seen it the the video
1:17:33 thank you Sherry D thank you very much
1:17:35 the video that
1:17:37 um the video tool that open a I just
1:17:40 announced
1:17:41 Sora is like Bonkers crazy so you go
1:17:44 from Runway ML and P Labs right now you
1:17:48 can make video with
1:17:50 them and you can make clips up to 4
1:17:54 seconds
1:17:56 long and then you can extend them for
1:17:59 four times so you can make a 16 second
1:18:04 clip we are at the knee of the curb the
1:18:07 curve the
1:18:08 curve um 16 seconds long and they're
1:18:12 really janky so like someone's head will
1:18:13 be normal in the first part of the video
1:18:15 and then it just goes and it just morphs
1:18:18 and does crazy
1:18:20 [ __ ] if you haven't seen Sora let me
1:18:22 just show you
1:18:24 a video from here I'll show I can show
1:18:26 you a bunch of them
1:18:28 but yeah this
1:18:31 one this video is 59 seconds
1:18:36 long if you look at the people in the
1:18:38 back behind the main
1:18:40 woman they're all staying intact
1:18:44 anything with Runway ml or or P Labs
1:18:48 there oh oh AGI earlier thanks Silver
1:18:50 Fox okay we let me come back to that
1:18:55 um all these things would be morph
1:18:57 you'll also see there's a cut here
1:18:59 there's an edit that edit to her close
1:19:02 up that was done as part of the same
1:19:06 prompt and look at the hang on look at
1:19:10 the reflection in her
1:19:16 glasses right this thing is aware of the
1:19:19 whole
1:19:22 world and it's 59 seconds
1:19:29 long and it's not flickering and it's
1:19:31 not morphing and it's not weirding
1:19:34 out now it's not perfect her shoes in
1:19:37 the first clip are like inconsistent
1:19:41 thicknesses
1:19:43 but the leap from 4 seconds of janky ass
1:19:48 video to
1:19:50 this that's what exponential change
1:19:52 looks like
1:19:54 um David Shapiro said that um in his
1:19:58 latest video that he's still bullish for
1:20:01 AGI showing up in seven months so by
1:20:07 September
1:20:08 October of
1:20:10 2024 AGI is going to show up AGI stands
1:20:13 for artificial general intelligence
1:20:15 that's basically the the point at which
1:20:18 Google came out with a p paper that
1:20:19 there's now five different stages of AGI
1:20:22 we're we're in stage zero right now um
1:20:25 David Shapiro thinks that we we will be
1:20:27 you know at fairly Advanced AGI within 7
1:20:30 months let's not get ahead of ourselves
1:20:32 Google Gemini looked impressive in the
1:20:34 video yeah I know steo
1:20:38 I'm I'm going with Shapiro on this one
1:20:41 he's written five books on AGI and he's
1:20:43 he's he's got an open- Source cognitive
1:20:46 architecture um project uh to build his
1:20:49 own AGI that he's going to open source
1:20:51 to the world um so I'm I'm going to
1:20:54 stick with him and I'm also an
1:20:58 optimist but but
1:21:00 basically when AGI is is achieved the
1:21:05 the AI tools can
1:21:09 do most humans jobs
1:21:15 completely meaning it's going to have
1:21:19 reasoning so it's going to be able to
1:21:21 it's going to be able to understand
1:21:23 understand that it made a mistake and
1:21:25 correct it it doesn't have that right
1:21:27 now and it's also going to have what's
1:21:29 called agentic Behavior it's going to
1:21:31 it's going to have agents so it's going
1:21:33 to be able to turn on things that will
1:21:35 take action on your
1:21:37 behalf we're going to see that this year
1:21:40 now what level it gets to and you know
1:21:43 will it actually be doing all of our
1:21:45 jobs no absolutely not right we're going
1:21:47 to see glimpses of it um but we're going
1:21:51 to see we're going to see
1:21:53 the
1:21:54 improvements
1:21:57 in in chat GPT here so this is I I can
1:22:01 never remember this because Sam Alman
1:22:03 speaks in such specific language so I I
1:22:05 wrote it down on a on a crinkled piece
1:22:07 of
1:22:08 paper
1:22:10 when I don't trust anyone who can write
1:22:13 five books on the same question figure
1:22:15 it
1:22:17 out P's had his weedies today he's he's
1:22:20 in a good mood I like it pate
1:22:23 I like it all right so here's what Sam
1:22:25 Alman said about chat GPT
1:22:28 4 so these two sentences were were in
1:22:31 his talk uh at Davos like three weeks
1:22:34 ago
1:22:37 um oh Nicholas shared the live thank you
1:22:39 Nicholas encourage more viewers to share
1:22:42 sure be like Nicholas share this live
1:22:46 here's if you want to have some fun
1:22:48 tonight share this live with people who
1:22:50 don't like Ai and I'll I'll straighten
1:22:53 them out I will I love talking to people
1:22:56 that are like AI is awful AI is stupid
1:22:58 ai's poo poo You're Dum dumb cuz you
1:23:01 like
1:23:03 AI I love talking to those
1:23:05 people cuz you know why you know why
1:23:08 they're acting like that
1:23:11 right they don't know more than you
1:23:13 they're
1:23:16 scared and being afraid is okay and we
1:23:20 can acknowledge that we should
1:23:21 acknowledge that
1:23:24 if that's what's going
1:23:27 on and then we should say want to see
1:23:29 something
1:23:30 cool and then blow their mind okay
1:23:33 here's what Sam Alman said about chat
1:23:35 gp4 at Davos three weeks ago he said
1:23:38 this in a lot of different interviews
1:23:39 some variation of this he said gp4 is
1:23:43 best understood as a
1:23:46 preview as a
1:23:50 preview like an appetizer
1:23:55 like wetting our
1:23:58 whistle
1:24:00 gp4 is I think four years old the the
1:24:03 underlying
1:24:04 model has been around for three or four
1:24:06 years chat GPT has been around for a
1:24:09 year they only released chat GPT 4 in
1:24:15 March of last year so hasn't even been a
1:24:17 year that we've had access to it they've
1:24:20 been working on it for three or four
1:24:22 years
1:24:26 so what do you think is sitting behind
1:24:28 that well one of the things they just
1:24:30 showed us is that video
1:24:32 [ __ ] and if you go read the technical
1:24:35 paper about how they're doing that video
1:24:36 [ __ ] video is the least thing it it's
1:24:39 the most simple thing that that system
1:24:42 is going to be
1:24:45 outputting in fact I'm just going to
1:24:47 read to
1:24:49 [Applause]
1:24:51 you
1:24:54 the very first sentence of this
1:24:59 announcement of of open ai's video
1:25:03 tool we're teaching AI to understand and
1:25:07 simulate the physical world in motion
1:25:10 with the goal of training models to help
1:25:13 people solve problems that require real
1:25:16 world
1:25:20 interaction that video is so good
1:25:24 because the underlying
1:25:26 model understands and is simulating the
1:25:29 physical
1:25:32 world
1:25:35 what they're not talking about frame
1:25:37 rates they're not talking about color
1:25:39 depth they're not talking about aspect
1:25:41 ratios they're talking about
1:25:43 understanding and simulating the
1:25:45 physical world so that we can help
1:25:47 people solve problems that require real
1:25:49 world interaction that's not a video
1:25:51 tool video just happens to be the
1:25:55 simplest thing that they can show us
1:25:58 that won't melt our [ __ ] faces so Sam
1:26:01 Alman says chat gp4 is best understood
1:26:04 as a preview and then he followed that
1:26:07 up with this sentence progress here is
1:26:10 not
1:26:12 linear right it's not
1:26:16 this
1:26:18 it's this my paper's too crooked it's
1:26:23 this right we're going along like this
1:26:26 we're just now starting to do
1:26:28 this and then these things do this
1:26:33 and why this why this is impossible to
1:26:38 comprehend is that as soon as these
1:26:40 systems have
1:26:42 reasoning where they can start
1:26:45 to not only give you an answer but take
1:26:48 that answer and analyze it and determine
1:26:51 if that answer was good good or not and
1:26:53 improve
1:26:54 it they can start improving
1:27:00 themselves that's that's where all the
1:27:01 Sci-Fi nerds are like what if they
1:27:03 improve theirselves so much and say that
1:27:05 we're not necessary right and that's
1:27:08 where you talk about things like
1:27:10 alignment are the are the imperatives of
1:27:13 the large language model in alignment
1:27:15 with the imperatives of the meat bags
1:27:19 that invented them they better be or
1:27:22 were in trouble but there's a lot of
1:27:24 smart people working on alignment
1:27:26 so I think I think Sam Alman said
1:27:29 they're
1:27:31 spending 20% 20% 20% of all money open
1:27:36 AI is spending right now is on what
1:27:38 they're calling super alignment to make
1:27:40 sure that these tools are
1:27:41 safe let's let's hope that's let's hope
1:27:45 that's enough that's
1:27:48 true Kyle what do you think things with
1:27:50 AI will be like at Year's End
1:27:53 um Year's End is easy if I go if you ask
1:27:56 me three years out I don't have a
1:27:58 [ __ ] clue okay bu years end I
1:28:02 think I think we'll have some version of
1:28:05 chat GPT
1:28:06 5 you think I'm working on this so that
1:28:09 the humans
1:28:11 win paint paint with another
1:28:15 singer you can make money with
1:28:18 ch um I think by the end we'll have
1:28:21 whatever the hell they're calling ch GPT
1:28:23 5 it seems clear to me that they're not
1:28:24 going to call it chat gp5 but whatever I
1:28:27 don't care about the name so much other
1:28:29 other than I hope they get a branding
1:28:31 person that knows what they're doing
1:28:34 um so so these these tools will have
1:28:40 reasoning so when Sam Alman was fired
1:28:43 there was a there was a there was a
1:28:46 release there was a leak of an email
1:28:49 that was sent to the
1:28:50 board that talked about this thing
1:28:53 qar and
1:28:56 qar is a is a an AI reasoning engine and
1:29:01 apparently
1:29:03 qar did some really good [ __ ] which
1:29:06 scared the B Jesus out of people and I
1:29:09 and rumor is that's why Sam Alman was
1:29:12 fired that they hit that they hit some
1:29:16 milestone in on the path to AGI which is
1:29:20 open ai's mission is to get to AGI and
1:29:23 you know provide it to humanity that's
1:29:26 their that's their mission it's why
1:29:27 they've got this weird nonprofit
1:29:28 structure it's why um that structure was
1:29:32 such a [ __ ] show where they fired Sam
1:29:34 Alman like for seemingly no [ __ ]
1:29:38 reason because one of the things they
1:29:40 said after they patched Humpty Dumpty
1:29:42 back together was oh yeah uh it had
1:29:44 nothing to do with qar everything's fine
1:29:46 here don't look over here Sam's back
1:29:47 everyone's good right we're good can we
1:29:49 just carry on
1:29:51 now
1:29:54 so I think that I think that gp5 will
1:29:57 have some high level of reasoning in it
1:30:00 and then that means it can do things
1:30:02 like math and when it can do things like
1:30:04 math then it can start to do all sorts
1:30:07 of crazy things so I think one is
1:30:08 reasoning I think the second one is
1:30:12 Agents autonomous agents so I think
1:30:15 Apple will do this I think
1:30:18 um uh open AI will certainly do this I
1:30:22 think Google will do
1:30:24 this
1:30:26 where when you want something done it
1:30:29 won't be like you go to chat GPT and you
1:30:31 ask it to do something it'll be like
1:30:33 whatever your tool is you say I want
1:30:35 something done and it will
1:30:38 spawn little agents little things little
1:30:41 Bots that go on your behalf and go do
1:30:43 [ __ ] for you and take action can deposit
1:30:47 money in your bank account can book a
1:30:49 travel thing can book a restaurant can
1:30:51 just do [ __ ] [ __ ] for you so I think
1:30:53 we'll see
1:30:54 that and then I think the third thing
1:30:57 that we're going to see is
1:31:00 that all of the major players will have
1:31:03 true
1:31:07 multimodal
1:31:10 um I don't they're not large language
1:31:13 models I don't know what they are true
1:31:15 large Lang true
1:31:17 multimodal AI so right now chat GPT can
1:31:21 hear you can talk to it if you download
1:31:25 the chat GPT app you can put it into
1:31:27 conversation mode it can talk so it's
1:31:30 got text to speech it can hear it can
1:31:32 see it's got Vision it can make images
1:31:36 it can do all the language stuff it can
1:31:39 code it can execute that code if you use
1:31:43 code interpreter for Advanced Data
1:31:44 analysis it does all
1:31:48 that what's missing music
1:31:53 audio voice
1:31:56 synthesis video 3D
1:32:00 worlds
1:32:02 dynamic programming like Dynamic
1:32:05 interfaces and Google already
1:32:08 demonstrated in their
1:32:10 Gemini marketing movies Dynamic
1:32:12 interfaces I can show you that [ __ ]
1:32:15 that's [ __ ] mind-blowing you ask it a
1:32:17 question and it says oh I think it would
1:32:19 be easier for this year if we just spin
1:32:21 up a little software application here
1:32:23 with buttons that you can click to do
1:32:26 [ __ ] and it spins it up writes the code
1:32:28 connects the databases and presents you
1:32:30 an application created on the fly so
1:32:34 when you have a single
1:32:36 interface that can do sight sound motion
1:32:40 programming 3D World development
1:32:45 creation
1:32:48 logic plus you have
1:32:50 reasoning plus you can take
1:32:53 actions I
1:32:55 like I what the I don't what the I don't
1:32:58 know what I don't know what that
1:33:05 is I still don't have my head fully
1:33:08 around chat
1:33:09 GPT in its current state I still don't
1:33:13 frankly have my head fully around chat
1:33:15 GPT
1:33:19 3.5 like if you've been using chat gbd
1:33:22 for the last year you're like well yeah
1:33:24 you you ask at things and it gives you
1:33:29 things but it's
1:33:32 remarkable and like you could ask at
1:33:34 anything and you could if you put your
1:33:36 mind to it you could either go
1:33:38 incredibly deep on a single
1:33:40 subject or if you did it in a focused
1:33:43 way you could you could as an
1:33:46 individual act as a 10 or 15 person
1:33:51 company just using chat GPT
1:33:59 3.5 and and they're showing us this
1:34:01 movie [ __ ] this you're going to be able
1:34:03 to talk something like that into
1:34:07 existence and did you see the 11 Labs
1:34:09 thing that the response to this from 11
1:34:12 Labs here let me show you
1:34:14 that you want to lose your [ __ ] mind
1:34:17 I'll you people want to lose your mind
1:34:20 I'll lose your mind for you
1:34:24 oh look at this
1:34:25 thing look at
1:34:27 this this dude is it a dude Martin this
1:34:35 dude hooked up a pressure sensitive
1:34:38 Wacom tablet that's got a screen
1:34:40 underneath it the sync Wacom tablet with
1:34:44 Kaa and
1:34:47 Photoshop
1:34:49 and is I guess in Photoshop I I think
1:34:53 he's in
1:34:55 Photoshop but it's sending the output of
1:34:58 Photoshop to the input of Crea which is
1:35:01 rendering it in real time so as he's
1:35:03 drawing on screen it's
1:35:05 rendering the [ __ ] underneath
1:35:09 it like this is someone just clever that
1:35:12 duct taped together three different
1:35:14 tools so this we're going to be able to
1:35:17 do look at that you can't draw that's
1:35:21 okay you can't draw here here now you
1:35:23 can but it's exactly where you want
1:35:26 it holy
1:35:30 crap so that I saw today and that's like
1:35:34 you can't buy that anywhere that's just
1:35:36 someone who's particularly clever has
1:35:39 some illustration skills and could
1:35:41 figure out how to connect those three
1:35:42 tools together um check this out um
1:35:48 11
1:35:50 Labs s
1:35:55 is this it no it's this one all
1:35:59 right okay
1:36:03 so open AI creates like launches this
1:36:06 thing Sora which is these you know you
1:36:09 can do up to a minute long video that
1:36:11 doesn't suck in 11 Labs who does voice
1:36:15 synthesis right they're the ones the
1:36:17 voice behind hey Jen's video
1:36:19 translations that's 11 labs um a lot of
1:36:22 the voice synthesis stuff that you hear
1:36:25 11 Labs is the engine underneath it so
1:36:27 open AI launches um Sora and the day
1:36:32 after 11 lab says sign up for our wait
1:36:36 list for text to sound effects and so
1:36:40 they took their video and used their
1:36:43 audio generation engine to do
1:36:50 this
1:37:17 in a place beyond imagination where the
1:37:20 Horizon kisses the heavens one man dares
1:37:22 to Journey where few have
1:37:25 ventured armed with nothing but his wit
1:37:27 and an unyielding Spirit he seeks the
1:37:30 answers to Mysteries that lie beyond the
1:37:36 stars wait see
1:37:42 that all these sound effects were
1:37:44 generated by 11 Labs without
1:37:47 modification how did they create these
1:37:49 sounds by sampling YouTube videos no no
1:37:52 they have a model well I don't know
1:37:54 where they got their sound libraries
1:37:56 from they might have done it from you
1:37:58 know open source there's all sorts of
1:37:59 audio libraries out there do you know
1:38:01 how long it takes the this one blows me
1:38:04 away do you know how long it takes this
1:38:11 one like that little robot sound do you
1:38:13 know how long it takes sound editors to
1:38:16 Source the sounds and Stitch them
1:38:20 together to be able to get something
1:38:22 like that or FY artist to do the walking
1:38:26 um how they generated them was text to
1:38:29 audio so you know sound of a woman
1:38:31 walking on a city street or whatever it
1:38:35 was click click click and they I'm I'm
1:38:37 sure they I doubt that they uploaded the
1:38:40 videos and it automatically did it I
1:38:41 assume they text to audio generated the
1:38:45 audio and then manually edited them but
1:38:47 still so so I think back to the question
1:38:51 what what do I think we see by the end
1:38:53 of the
1:38:55 year I think by the end of the year
1:38:58 we're going to have tools that have all
1:39:01 of the modalities of
1:39:05 communications at our
1:39:09 fingertips and you may not need to do
1:39:12 prompt engineering if if I'm looking at
1:39:14 the guy with the tablet where he's
1:39:16 drawing tree limbs and they're just
1:39:18 magically sprouting into
1:39:20 illustrations
1:39:24 you might prompt your large language
1:39:26 model with your phone your you know the
1:39:28 camera on your
1:39:30 phone or with just you know it's
1:39:33 connected to Siri and it just spins up
1:39:34 all this [ __ ] by the time you get to
1:39:36 work you've got a 30- second commercial
1:39:38 for your business
1:39:44 done so I think I think that by the end
1:39:46 of the
1:39:48 year those tools are out
1:39:52 somewhere they're probably not all going
1:39:54 to be in the same interface they're
1:39:56 probably not going to all be in the same
1:39:58 company but they're all going to be
1:40:00 there and they're going to be like
1:40:03 staggeringly
1:40:04 good and so anyone right now that's
1:40:06 sitting on the sidelines like well these
1:40:08 things aren't
1:40:10 perfect this is stupid it's
1:40:16 stealing they're going to be trying to
1:40:18 hang on for dear life trying to catch up
1:40:20 to people that show up in this community
1:40:23 and hang out at the AI
1:40:26 Salon not that
1:40:30 community that Community the AI
1:40:35 Salon which if you don't know it you
1:40:37 should go check it
1:40:39 out AI Salon the
1:40:47 salon. go read about the salon and our
1:40:50 values and then the first thing you
1:40:52 should do there is go introduce yourself
1:40:54 if you're curious about this AI stuff
1:40:56 this is your
1:40:59 group this group is not like any other
1:41:02 [ __ ] group out there I promise you
1:41:04 it's generous it's full of nice people
1:41:09 helpful people enthusiastic
1:41:16 people it's so
1:41:18 overwhelming when I get in though it is
1:41:20 it so
1:41:22 what's what's your name L wait LOL o LOL
1:41:26 LOL oh I get it that's good that's very
1:41:29 cute nine out of 10 just because it took
1:41:31 me a while to parse it thank you Sherry
1:41:34 D yeah this stuff is overwhelming which
1:41:37 is which is one of the reasons that
1:41:41 I strongly advocate for you playing not
1:41:46 working not learning I should call this
1:41:48 the AI playing lab not the AI learning
1:41:50 lab
1:41:52 because in playing you will
1:41:54 learn just in hanging out in these
1:41:57 things even though it doesn't seem
1:41:58 obvious tonight cuz I'm just sort of
1:42:00 yapping my [ __ ]
1:42:04 Jaws being in the presence of other
1:42:06 people that are trying to figure this
1:42:08 out that are also overwhelmed one of the
1:42:10 things that I see really consistently in
1:42:12 the salon and in this community is is
1:42:14 kind of an
1:42:16 oscillation where people get really
1:42:18 excited about Ai and they're like oh my
1:42:20 God the Poss the future is so bright I
1:42:21 see all this possibility then they get
1:42:23 overwhelmed and they they just go like
1:42:25 [ __ ] it I can't I can't and they they go
1:42:27 down down in this trenches and then
1:42:29 either something happens or they try
1:42:31 something and they're like oh right this
1:42:33 shit's really [ __ ] remarkable and
1:42:34 then they come back up and it's there's
1:42:36 these oscillations that just are doing
1:42:38 this all the
1:42:40 time and and it's also the if you if you
1:42:45 haven't really played with a lot of AI
1:42:47 just start playing with it a lot until
1:42:49 you have what I call your first Kevin
1:42:51 mallister moment right this
1:42:55 one you'll have that
1:42:58 moment and then you'll have another one
1:43:00 and then you'll have another one and
1:43:01 then you'll have another one and you
1:43:03 can't unsee
1:43:06 it that's the thing about this stuff is
1:43:08 it's like you
1:43:12 know we all watch The Matrix you know
1:43:15 red pill blue pill we get it you know
1:43:18 you can't unsee it right don't take the
1:43:21 red pill you take the blue pill you just
1:43:23 go back to your Slumber everything's
1:43:26 fine you take the red pill you can't
1:43:29 unsee it right you're going to learn
1:43:32 something that you can't
1:43:34 unlearn that's what this [ __ ]
1:43:40 is and I got to tell you as overwhelming
1:43:44 as it is it's a lot less scary when
1:43:46 you're in it than when you're sitting on
1:43:48 the [ __ ]
1:43:50 sidelines I've had multiple whole family
1:43:53 Kevin mallister moments at this point
1:43:55 yeah exactly that's one of the other
1:43:57 things I strongly recommend one is just
1:43:59 play two is try to solve problems with
1:44:03 it like try to solve problems for your
1:44:05 family or your friends or at work or for
1:44:07 yourself whatever and then the third one
1:44:09 is talk about it share about it share it
1:44:12 with your family share it with your
1:44:14 co-workers but they don't let us use AI
1:44:16 at work [ __ ]
1:44:19 them use it on the weekends
1:44:22 get yourself a $200 Chromebook at Best
1:44:25 Buy if you can't use it on the work
1:44:28 computer and just start playing with it
1:44:30 just do it on your phone download
1:44:32 Pi start having conversations with
1:44:36 pi just start playing with it and start
1:44:39 talking about
1:44:40 it and the people that resist it
1:44:43 eventually are going to say hey could
1:44:44 you teach me how to do
1:44:47 that
1:44:48 so anyway I got to go I got to get up
1:44:52 early it's already
1:44:58 late 10 minute goodbyes all right so if
1:45:01 you enjoyed what what what you heard
1:45:03 here tonight I don't know what's wrong
1:45:05 with you
1:45:08 um follow my channel so I don't know how
1:45:12 you do that on Tik Tok but you can
1:45:14 figure it out you can subscribe to my
1:45:16 lives if you want to support this you
1:45:18 can buy in the in one of the corners
1:45:20 there's a video ser series I created
1:45:21 called underhyped which explains sort of
1:45:24 my point of view on why I
1:45:26 think um this AI stuff is a lot bigger
1:45:28 deal than people think and it gives you
1:45:30 a lot of tools on how to cope with and
1:45:33 not get overwhelmed by these
1:45:36 tools and
1:45:38 then consider joining the AI Salon
1:45:40 there's there's no cost to this by the
1:45:42 way this doesn't cost anything it's free
1:45:45 it's supported by people who just step
1:45:47 up and say yeah I'm willing to sponsor a
1:45:49 month Sher D is sponsored dur ing
1:45:51 February so thank you Sher
1:45:55 D all right and then we're going to get
1:45:58 out of
1:46:01 here did audio generation Minds blown
1:46:05 three Minds blown I know crazy
1:46:08 right how did they create those sounds
1:46:10 by sampling YouTube videos it gave him a
1:46:13 knitted helmet that no so Dena that
1:46:16 knitted helmet was part of the prompt it
1:46:18 was a knitted motorcycle helmet it was a
1:46:21 Spaceman on ass salt Dune or ass salt
1:46:25 planet or ass salt Dune I don't know um
1:46:28 with a knitted motorcy a red knitted
1:46:30 motorcycle
1:46:31 helmet um oh the first Jack's class is
1:46:33 tomorrow oh it's being run by awesome
1:46:37 coworker info in the mechanics Guild
1:46:39 okay so tomorrow so if you're
1:46:42 geeky and you want to learn about Jacks
1:46:46 which is if you want to create your own
1:46:47 large language model like if you want to
1:46:49 really get under the hood and understand
1:46:51 how these things actually work Pate who
1:46:54 runs the mechanics Guild in the AI Salon
1:46:57 he's got information in in the AI
1:46:59 mechanics Guild or the yeah the
1:47:00 mechanics Guild um about a class
1:47:03 tomorrow that his coworker is putting on
1:47:05 on how to make your own large language
1:47:07 model all right so go there read about
1:47:11 our values introduce yourself run to the
1:47:14 mechanics Guild and geek
1:47:18 out all right I'm getting out of here P
1:47:21 good to see you LOLOL I like your name
1:47:24 Tobias great to see
1:47:26 you uh Be not Afraid Danielle thank you
1:47:29 very much Sherry d as always thank you
1:47:31 for your
1:47:32 generosity uh who else we got in here
1:47:34 what's up Tock Dylan good to see you sha
1:47:36 bomb good to see you solar cycle lady
1:47:41 good to see you Clint Copperhead thank
1:47:42 you everyone in the regulars present
1:47:44 your
1:47:45 animals um thank you Tobias thank you
1:47:49 thank you thank you invest friendly
1:47:51 agent thank you so much good question
1:47:53 tonight by the way uh a human being good
1:47:56 to see
1:47:57 you all right Dr J Amelio's wife sherid
1:48:01 D thank you for the dangly lights I
1:48:04 loves me some dangly
1:48:06 lights laughing Thunder Rachel
1:48:11 Sparkles all right
1:48:13 everybody Kayla funnel fairy
1:48:16 coach love your
1:48:19 name
1:48:23 all right everyone I am out of here I'm
1:48:26 Crispy I'm going to go crash and get all
1:48:30 ready for my board meeting I'm going
1:48:32 board meeting you know why they call it
1:48:34 a board meeting CU if you show up you'll
1:48:37 be [ __ ] bored
1:48:41 peace