aiLL 9/27/2023 - From DOS to Conversational AI: How Far We’ve Come and Where We’re Headed

Description

Chapters

Transcript

0:01 foreign
0:23 good evening welcome welcome welcome
0:28 hey Ted looks at life what's happening
0:29 good to see you
0:33 user
0:36 575-217-03-8542 good to see you nice
0:39 memorable name I like it it's solid
0:41 solid if you had a couple of numbers in
0:44 a row it would have thrown me the fact
0:45 that they're all different they're all
0:47 uh it's with that's Fibonacci Sequence I
0:50 like it solid it's just it's like uh
0:53 it's like a mathematician's dream
0:55 yeah yeah
1:01 what's happening
1:03 what's happening jockster it's going
1:05 down what should I cut
1:07 Lots going on
1:10 the internet is back in chat GPT except
1:12 not in my [ __ ] Chef jpt
1:16 those sons of [ __ ] I got nothing
1:19 everyone's like I can play with images I
1:22 can play with this the internet's back
1:23 and I'm looking at mine it's like
1:29 welcome everyone to the AI learning lab
1:31 this is Larry that's Larry from Larry's
1:34 uh
1:35 Larry's hedgehogs Lawn Lawn Care Service
1:39 Landscaping service
1:42 we made him last night
1:45 um let's see what else we got here what
1:48 else we want to look at yeah we'll be a
1:51 wheel let's see am I late
1:55 AI learning lab this is the AI learning
1:58 lab my name is Kyle Shannon Paul Z what
2:01 is up KS what's going down Bigfoot
2:05 Electro hello from Virginia
2:07 welcome to Virginia
2:11 what was that line it was a line of a
2:13 play I was in
2:14 called The Amazing Adventures of tense
2:16 guy
2:17 oh and one of the actors Mark Ashmore
2:20 he's [ __ ] hilarious he misread one of
2:22 the lines the line was
2:24 um get me Langley Virginia right give me
2:27 the give me the CIA headquarters or FBI
2:31 CIA FBI and he goes he goes get me
2:34 Langley Virginia he thought Virginia was
2:37 like the secretary's name pretty funny
2:40 oh man working on the comics with Dolly
2:44 and mid-journey
2:45 nice do you have do you have access to
2:47 Dolly three yet jockster I don't have
2:50 nothing I got nothing
2:53 gotta have them Winters for for for Bing
2:56 chat jpete
2:58 Chad jpt you know one of the things we
3:01 talk about here in AI learning lab is
3:07 so AI is just like nfts
3:11 except
3:12 if you give send a hundred thousand
3:15 dollars to my to my world ID with in
3:18 World coin and then uh your chat GPT
3:22 you'll make money with this is solid
3:24 gold I made this out of uh nugget a Doug
3:27 in the last Yukon
3:31 all right
3:33 so we'll let some people get in here a
3:35 lot of folks in here I learned about hey
3:37 Jen from you on the previous stream the
3:39 demo blew my mind go check out my latest
3:42 you want to see something really [ __ ]
3:45 wild go check out my latest video in my
3:48 channel is this it no that's not it hang
3:50 on look at this thing this is just this
3:53 is just getting dumb at this point man
3:56 man
3:57 all right look oh wait you can't
4:01 so
4:02 yeah so this is just a Story vine video
4:05 storyvine's my company but then look
4:07 that there's Chinese Mandarin Chinese
4:15 what's this
4:21 is
4:38 now here's one weird thing so notice
4:41 that my hand is inside my face
4:45 because what they're doing is they're
4:47 kind of synthing the lower half of your
4:49 face and then laying it on top of the
4:52 video so so they're not masking out the
4:54 hands if they go over and I'm a hand
4:56 talker so so if I'm gonna if I'm gonna
5:00 record any of these that actually you
5:02 know that I actually use for anything
5:03 other than a demo I need to not do that
5:05 but still even with this it's like
5:07 totally forgivable because it's you know
5:10 it's it's like it's like fully fluent
5:14 Chinese you know with lip movement and
5:17 it sounds like me
5:19 stories
5:38 so that's hey Jen so anyway welcome
5:41 everyone this is the AI learning lab
5:45 and my name is Kyle Shannon and we'll
5:49 get started here so generally what we do
5:51 here if you're new here
5:53 um I'm a little bit about my background
5:55 I'm an entrepreneur of
5:57 I've started enough companies that I'm
5:59 now officially not hireable anywhere
6:03 and my current gig is oh yeah this like
6:06 all story mine tonight right uh I'm the
6:09 founder and CEO of a company called
6:10 storyvine I even have a hat here look
6:13 look see this is
6:17 solid
6:19 um
6:22 so so
6:24 um
6:25 so that's my day job I've been uh kind
6:28 of deep diving into this generative AI
6:30 stuff for about oh I don't know what
6:33 two years now and and ever since chat
6:37 GPT uh I've lost my [ __ ] mind
6:41 so this channel I go live every night
6:43 here I'll answer any questions you have
6:46 um
6:47 this channel is really about the
6:49 inflection point that started on
6:51 November 30th 2022 so the launch of chat
6:53 gbt prior to that point machine learning
6:57 and AI was really the the domain of you
6:59 know data scientists and mathematicians
7:02 and researchers and engineers and after
7:04 that point
7:06 it's for the rest of us to figure out
7:08 like what are we going to do with this
7:10 stuff what are we going to do with these
7:11 tools now that these tools are in our
7:13 hands November 1st marks
7:16 um General availability of Chachi BT
7:19 inside office 365. so you're gonna have
7:21 chat EBT within word and PowerPoint and
7:25 Excel
7:26 um it's already in Windows 11. it's
7:28 already in GitHub co-pilot it's already
7:30 in all the Google uh things not gpt4
7:33 there there's his Bard but all the
7:36 Google office docs have access to this
7:38 thing called Duet which is kind of a
7:39 shitty version of it but it's going to
7:40 get better quick
7:42 um
7:44 chat GPT has gone multimodal where it
7:47 can see images now it can see
7:49 I saw a thing today a guy took a
7:51 screenshot of a computer interface and
7:55 asked chat GPT to analyze the components
7:57 of the computer interface and write the
8:00 code for the different components and it
8:02 did it and it was good
8:04 it's like wow so
8:08 we have no idea zero idea what it means
8:13 to have a tool like chat GPT when it can
8:16 now speak and listen and see and draw
8:20 and make words
8:22 and you know within some short amount of
8:24 time it'll be able to do music and video
8:26 too and you know once once we have all
8:29 of the all of the media
8:32 um you know all bets are off what what
8:34 do we do with that I don't know
8:37 um
8:38 I'm a gen xer look I have an AARP card
8:42 I uh that just came today
8:47 um
8:48 I'm a little bitter
8:50 chat GPT hasn't given me access to the
8:52 new toys yet I flip and advertise their
8:55 their silly little company every night
8:58 for like four months and they don't give
9:00 me access to the new toys really chat
9:02 gbt really
9:04 really I I made a button for you
9:09 nothing
9:11 so anyway
9:13 um if you have questions pop them in the
9:14 comments below I'll do my best to answer
9:16 them I'll go through some of these uh
9:18 some of the if you want to get started
9:20 where to start but but let's just jump
9:21 in and start answering some questions
9:26 Winston joined Dave Winston there's pate
9:31 good to see you sir good day sir good
9:35 day how are you
9:36 well I'm rather fine I do Wonder Mr
9:39 Shannon what are your qualifications no
9:42 well well now we're asking for
9:44 qualifications are we well it's I think
9:46 it's quite reasonable to wonder you know
9:47 if I'm at a learning lab what are we
9:49 learning at the lab and if you're
9:50 teaching at the lab what are your
9:52 qualifications seems a relatively
9:54 straightforward question does it not
9:56 well I suppose for you it does
9:59 all right
10:01 let's let that one hanging right there
10:03 laughs
10:07 all right all right all right I did not
10:10 see your play uh which play I do every
10:14 night here is a play
10:15 oh I learned about hey Jen for me on a
10:17 previous one we talked about that
10:19 started at 38 000 now at 20 000 in the
10:22 queue this free version is really going
10:24 to take four days to run for render I
10:26 got to pay up I know my um my Mandarin
10:29 one the the one I just showed
10:32 I got in it was like 83 000 and I think
10:35 mine took five days so yeah yeah it took
10:38 it took a while I forgot about it and
10:40 then I got an email today this morning
10:42 that said here's your Mandarin
10:44 um yeah I'm gonna pay up for that as
10:46 well because I got I got to do it for my
10:47 company because uh this is this is too
10:50 important technology not to not to have
10:53 not not to be building stuff on
10:58 um so anyway yeah exactly learning how
11:01 important using the right words for
11:03 prompts is yeah it's you know it's it's
11:06 using the right words that you
11:08 have having
11:12 being experimental and having a good
11:14 sense of
11:15 language logic is definitely an
11:18 advantage
11:20 um but one of the things that you can
11:22 either do also do
11:24 um uh Harrow Herod Herod is is one of
11:29 the Irregulars here
11:31 um and and he came on the other night
11:33 and was talking about the fact the other
11:35 thing that you can do is you can have
11:36 chat GPT help you write prompts for
11:39 itself
11:40 so if you're getting stuck on the if
11:43 you're you know if you're getting
11:44 twisted up on the on the language stuff
11:45 and I don't really know how to ask this
11:47 and what to do
11:48 ask Chachi PT how how
11:51 what's the best prompt that you could
11:54 give it to get a certain
11:57 um result
11:58 and let it write the prompt or or you
12:00 write a prompt and say instead of
12:02 answering this prompt I want you to
12:04 critique it and tell me how you would
12:06 improve it to get a stronger answer or a
12:09 more clear answer or whatever that is
12:11 let it write it
12:13 let it write it all right
12:16 this one feels vastly more complex than
12:19 Anglo-Saxon European languages
12:20 impressive I know well that's so so
12:23 that's the thing about
12:25 these the Transformer models and the way
12:28 all of this technology works it's
12:30 [ __ ] magic it's not it's it's all
12:33 mathematics
12:35 um but the result right the Arthur C
12:38 Clarke quote any sufficiently advanced
12:41 technology is indistinguishable from
12:44 Magic
12:45 so they train these models on the
12:48 different languages
12:50 and whether it's word languages or
12:53 spoken languages
12:55 and
12:58 you know or or uh you know voice
13:00 synthesis right or you know music or
13:03 images
13:04 and then when you put in your prompt and
13:06 they marry that large language model to
13:09 that other whether it's a diffusion
13:10 model or or you know the audio stuff
13:13 and it just I don't know
13:16 it mythically pulls the [ __ ] out of The
13:19 Ether mathematically it's it's it's it's
13:21 it's absolutely incredible
13:24 you know it's it's not I think from a
13:27 machine learning point of view
13:28 I don't think I I'm curious if pay
13:32 you're in here
13:33 is is is is training a large line or not
13:37 a large language model is is training a
13:39 machine learning model
13:40 if we're talking about audio you know
13:43 synthesis stuff here is it any harder to
13:45 train it on Mandarin Chinese than it is
13:48 on Italian or it is on English I assume
13:50 it's the same thing
13:52 it's more complicated for us because we
13:56 have these tiny [ __ ] filet mignons in
13:59 our head
14:01 we can't take everything in at once
14:05 oh man mods in Irregulars present
14:09 Turtles
14:10 good evening Irregulars I hope you're
14:12 all well tonight
14:14 curious about this hello Kyle hello
14:16 Apocrypha how are you
14:19 I just started using docchio Turtle Army
14:23 throwing all my privacy concerns out the
14:26 window but I can find files you know I
14:29 listen I'm I'm being careful with my
14:32 with my uh my client data
14:35 um
14:36 but I but at this point I just feel like
14:39 I'm just putting it out there so yeah
14:41 I'm right there with you
14:44 I figure you know I'm not doing anything
14:46 all that particularly interesting right
14:48 now like from from a privacy standpoint
14:51 so like I feel like the value of what
14:55 I'm contributing to the world of machine
14:57 learning and prompt engineering is low
15:00 value enough that you know open AI ain't
15:02 ain't siphoning my ideas off you know if
15:06 that's even a concern right so I you
15:09 know
15:10 this is this is all like everything
15:13 we're doing these days is temporary
15:15 anyway right all the because like like
15:17 as these new models change it's it's
15:19 gonna get
15:20 um
15:21 you know the the thing that that we
15:23 thought was the perfect prompt three
15:26 weeks ago is not is going to be useless
15:29 now because there's it's just a
15:31 completely different thing I keep
15:32 checking my settings here to see if I
15:35 get any extra I got nothing
15:37 I got no I got no brows with Bing I got
15:41 no image upload
15:43 nothing my my uh open AI Chachi PT app
15:47 doesn't have the voice thing in it I got
15:50 nothing people nothing
15:52 anybody here have it anybody here got
15:55 the the extra good and chewy new stuff
15:59 how would we know if we have dally three
16:02 access I don't know policy I'll tell you
16:05 I got the new co-pilot on Startup today
16:08 and it had chosen a turtle for a
16:10 background that's [ __ ] awesome silver
16:13 fox
16:14 that's great so you got very suspect you
16:16 so you got
16:18 um that's co-pilot for Windows 11 right
16:20 I'm curious what that does
16:23 uh oh cool I've been keeping that tab
16:26 open too I was afraid to close it yeah I
16:28 think it's tied to your account and it's
16:30 I assume what they're doing is just sort
16:32 of batching you know batching accounts
16:35 with the with the new you know they're
16:36 probably doing 10 000 Accounts at a time
16:38 you know every 15 minutes they'll do
16:40 another I mean I don't who knows how
16:42 they're rolling that [ __ ] out but that's
16:43 probably what's going on
16:45 qualifications AI picked me out of a
16:49 lineup to study my neural Pathways
16:51 exactly
16:54 um I just booted up not sure yet what's
16:57 up oh okay that's cool
17:02 um but apparently Silver Fox you can do
17:04 things like you know
17:06 put my system into dark mode or don't
17:09 make my system go to sleep you know
17:11 after five minutes make it 15.
17:14 it can do
17:17 all the [ __ ] that you normally have to
17:18 go set up in your settings it'll just go
17:21 do that
17:24 there's a new project out called open
17:26 interpreter which is like code
17:28 interpreter that you can install into
17:30 your system
17:32 and I've installed it in mine but it's I
17:34 haven't really played with it
17:36 I'm waiting I'm waiting for the the kind
17:38 of sexier commercial version of this
17:39 [ __ ] I feel like George drill sergeant
17:43 when I do that Turtle present turtles by
17:47 the way if you're wondering why uh
17:48 people are talking about turtles it's
17:50 because they force me on a nightly basis
17:54 to play
17:56 a video
17:58 oh do I not have internet
18:00 um this looks suspect
18:03 I wonder if my edge browser has crashed
18:08 oh Microsoft
18:12 what just happened
18:16 huh I think I quit it all right Edge
18:28 all right let's see I like
18:35 uh
18:36 [Music]
18:38 this is for all of you
18:41 back here live at the waterfront village
18:43 with my friend the zombie Jonathan
18:45 you're looking good Jonathan just got an
18:47 awesome face paint job what do you think
18:49 I like turtles all right you're great
18:52 zombie and good times here
19:02 not really Kyle which is your turtle
19:04 Army that makes me feel better you
19:07 weirdos y'all weirdos I love it I like
19:11 it I like it I like it all right all
19:13 right all right all right let's see what
19:16 other questions we got here
19:18 do a regulars truly exist if when there
19:21 are no turtles there to perceive them uh
19:24 we don't get that philosophical here
19:25 Winston we can't when we start asking
19:27 questions we don't want to know the
19:29 answer to it goes south really quick
19:32 we just we just we just stumble through
19:34 life here at the AI learning lab we
19:36 don't want to be conscious of anything
19:38 right now
19:39 actually you know that's kind of funny I
19:41 love I love how it makes you giggle
19:43 every time it does you know it's it is
19:46 it is the kid because it's so unexpected
19:49 but it's the reporter trying to keep it
19:51 together that is the thing that just
19:53 busts me up every time okay good zombie
20:00 ah
20:01 you can't that's you can't teach that in
20:03 acting class man that level of
20:06 spontaneity is just that's gold
20:09 all right let's see
20:13 so your hunch is right it's not easier
20:15 ah okay pate it's not easier to
20:18 translate to languages that are related
20:21 yeah so
20:24 so to the machine learning models like
20:26 once these things go into Vector space
20:29 it's actually quite a simple operation
20:32 well I don't know how simple it is
20:33 technically but it's simple conceptually
20:36 you write a prompt and it turns into you
20:40 know
20:41 a probability within this thousand
20:44 Dimension Vector space and some point in
20:47 that Vector space is the highest
20:49 probability thing and whether that's a
20:51 word or whether that's a
20:53 I don't know points to some sort of uh
20:56 noised up image in a diffusion model
21:00 um it just does its thing it's it really
21:03 is remarkable how this Tech works
21:06 is there an art renderer for 3D design
21:09 well there there are some quasi ones
21:13 um and and there have been a bunch of
21:15 white papers on taking objects like
21:18 mid-journey images and turning them into
21:20 3D objects
21:21 there hasn't been like a big like
21:24 mid-journey type tool released yet that
21:27 automatically generates 3D objects that
21:30 I know of what I would do if I were you
21:33 because it's very possible that
21:35 something has been released and I don't
21:37 know about it so if anyone in here knows
21:38 about it something like that let me know
21:41 but what I would do also is go to
21:46 futurepedia.io that's like a list of
21:49 like I don't know it's got to be four
21:51 thousand forty two hundred forty five
21:52 hundred tools at this point those are a
21:54 bunch of AI tools just go there I'm sure
21:57 there's a 3D section or I'm sure there's
21:59 a a something section
22:02 um there are likely applications that
22:05 that
22:06 um let you do natural language input
22:08 that
22:10 um generate 3D objects in a in a program
22:13 like Unity
22:15 that probably exists but if you're
22:17 talking about just make me an apple and
22:19 it makes you an apple like like
22:20 mid-journey but it's a 3D object I don't
22:23 think that's there yet but I could be
22:24 very wrong
22:25 what company do you own sir it is a
22:28 company called storyvine in fact
22:31 I got the little URL right there right
22:34 there it is storevine.com oh yeah
22:38 I bought that in the in the in 1990s
22:43 news Vine was big at the time
22:46 and I had an idea for the company that
22:50 became storybine and I thought storybine
22:52 that's really good and I bought that I
22:55 paid it off over two years
22:58 so
23:00 fancy that
23:02 um
23:03 [Music]
23:06 meta announcements today VR glasses and
23:09 headset oh VR glasses yeah the Ray-Ban
23:12 glasses the headset
23:14 which I can't I cannot believe I heard
23:16 one little clip of the meta thing and
23:19 the dude from meta said we're gonna have
23:22 the world's cheapest spatial Computing
23:25 glasses
23:27 meta used apples terminology
23:32 they were everyone was calling it The
23:34 metaverse
23:36 meta changed their [ __ ] name to meta
23:40 and today a dude from meta
23:44 called
23:45 the thing that those Goggles do spatial
23:49 Computing goggles
23:53 oh my god do they not have a PR for like
23:56 just does Zuckerberg like need a loan
23:59 for a PR firm
24:02 I don't know maybe they cut some deal
24:04 with Apple or something like that but
24:05 that just blew my mind that they did
24:07 that that guy said that
24:09 uh yeah the meta AI assistant and the
24:12 Tom Brady bot and there's like there's a
24:14 yeah they yeah meta today announced all
24:17 these like
24:18 um personality Bots like one is the
24:21 dungeon master and it's Snoop Dogg I I
24:23 don't know that D and D players are
24:26 thinking Snoop Dogg although probably be
24:27 kind of cool to play D D with Snoop Dogg
24:30 but
24:32 oh man I don't know man meta seems like
24:36 they're all over the place I I just I I
24:39 keep every time I see their product
24:41 launches I'm like who is their target
24:44 audience it's like it's like they don't
24:46 have one and it's like our target
24:48 audience is everyone you know here's a
24:51 consumer toy here's this here's that
24:53 here's a piece of important technology
24:56 uh but yeah Big announcements today big
24:58 announcements for meta
25:01 um this Chachi PT stuff's rolling out
25:04 the Microsoft stuff is coming the Gemini
25:06 stuff is coming soon
25:09 um it's gonna be a wild October November
25:12 December I think
25:13 crazy crazy crazy it's gonna be crazy
25:17 um listen if you're new here if you have
25:19 not played with chat GPT
25:22 what we talk about here is generative Ai
25:25 and kind of the the little poster child
25:27 right now
25:29 of that technology is chat TPT
25:33 they're the ones that got this ball
25:34 rolling down the hill really fast so
25:37 chat.openai.com that's the official
25:39 website right there just go there and
25:41 start playing while while I'm rambling
25:43 on here look at Harper Chrome extension
25:46 plug-in for YouTube videos transcript
25:48 and mind map oh I I saw that today I saw
25:52 a tick tock video on that today of
25:53 someone that showed that mind map thing
25:55 yeah it's pretty cool that's pretty cool
25:57 we'll we'll see how uh
26:00 how like once once these tools get
26:02 multimodal and and I have a sneaking
26:05 suspicion that Gemini Google's new bard
26:10 under under engine underbelly I have a
26:13 sneaking suspicion it's going to have
26:14 some pretty tight YouTube Integrations I
26:17 I have a feeling it was trained on
26:18 YouTube
26:19 which is why it's probably really good
26:21 um what do you mean paid it off over two
26:24 years I thought it's a monthly payment
26:25 no no no some dude owned storyvine.com
26:31 and I agreed to pay him seventy five
26:33 hundred dollars and I didn't have
26:34 seventy five hundred dollars so I said I
26:37 forget what I paid him like 200 bucks a
26:40 month for
26:41 I don't know however long it took
26:42 whatever or whatever two years is into 7
26:46 500 bucks
26:48 um
26:49 so yeah that was that that was that I
26:53 forget the name but ten thousand percent
26:54 yes not sure what that's about but cool
26:57 Luna stick hola what's happening
27:05 Zippy zambo good name I like it what's
27:09 the best domain registry to purchase a
27:11 domain name website safely surely I
27:15 don't know I buy my ship from
27:17 GoDaddy or Network Solutions just
27:19 because I'm old Network Network
27:21 Solutions was like this
27:23 quasi-governmental thing for a long time
27:25 and then it turned into that and then
27:27 GoDaddy I hate because it's it it was so
27:32 like commercial and weird for a while
27:33 but it's it's kind of normalized now so
27:36 I don't really [ __ ] around with them I
27:38 bought
27:39 um I bought storyvine.ai from whatever
27:41 the [ __ ] registry you've gotta bait pay
27:45 the 200 bucks for the AI
27:47 top level domain
27:50 um
27:51 I don't know I I'm not big into that
27:54 game I don't I don't know that game at
27:56 all I just I just wanted that domain
27:58 because I had an idea for a company and
28:00 then I went and started the company and
28:01 now it's an 11 and a half year old
28:03 company storybine we've been around for
28:05 a while
28:07 we are out there
28:12 all right oh Zippy zambo that is awesome
28:16 thank you very much I appreciate that
28:19 Meadow throws out lots of stuff hoping
28:22 something will stick to the wall then
28:23 they run with it yeah that makes sense I
28:25 suppose it just it it feels like that
28:27 you know this channel is like chat add I
28:29 feel like meta is like product add it's
28:32 like it's like here's 30 things you want
28:34 one
28:35 uh
28:37 okay
28:39 got Windows 11 co-pilot today it's Bing
28:43 on your toolbar very cool
28:50 all right so here's what I'm gonna do
28:52 I'm gonna go
28:54 I'm gonna walk through some things here
28:55 so if you're new to the channel
28:58 I'm gonna I'm gonna walk you through a
28:59 list of
29:01 tools that I think are worth your time
29:03 to explore
29:05 and if you if you'd be so kind
29:09 um go ahead and share the share their
29:11 lives so we can get the word out there
29:12 too bit more people some more people
29:15 that would be good
29:17 and you know tap the screen if you've
29:19 got some Irregulars in here start
29:22 playing your game see if we can get the
29:23 get some folks in here
29:25 um so so what I'd suggest if you're if
29:28 you're brand new to this stuff
29:29 screenshot I'm gonna go through I think
29:32 it's four or five slides here
29:34 um of just tools that I think are worth
29:36 playing with right now
29:38 um this is what I call the adult table
29:40 so these are the top three tools that
29:42 I'm using pretty consistently now
29:44 although Bard probably not as much I use
29:46 the top two a lot
29:48 uh the top one is chat GPT
29:52 um depending on oh thank you very much
29:55 that was a cool one I don't think I've
29:56 seen that one before
29:58 um depending on if you uh pay for uh
30:01 chat GPT plus you may or may not have
30:04 access to their new
30:06 um chat gptv which is V is for vision
30:09 where it can see stuff you can upload
30:11 images to it
30:13 um
30:14 and they also brought back browse with
30:16 Bing internet connection today so so
30:19 um I don't have I can't show you either
30:21 of those because I don't have them yet
30:23 but
30:25 this is about to get
30:27 um much more sophisticated really
30:29 quickly Claude is by a company called
30:32 anthropic who's a bunch of ex-open AI
30:34 people who started a competitor and they
30:37 just got invested in by Amazon to the
30:40 tune of four billion dollars well 1.4
30:42 billion 1.5 billion with a four billion
30:45 dollar you know a possible total
30:47 investment
30:48 and then Bard is Google's entry into
30:51 this space it's connected to the
30:53 internet and
30:55 they've got this thing coming out called
30:57 Gemini which I assume is going to sit
30:59 underneath that thing and they claim
31:02 it's like five times better than gpt4
31:05 and like eight times bigger and I don't
31:07 know they're they're all swinging there
31:11 we Willie winkies at each other
31:15 mine's bigger than yours though mine's
31:17 bigger than yours well mine can do
31:19 multimodal now mine can do multimodal
31:22 there's a lot of that going on right now
31:23 so the good news is we're the
31:25 beneficiaries of it but it's a little
31:27 exhausting okay so that's that page oh
31:29 and then prompts.chat that's just a
31:31 little document it's like a little
31:32 primer document that talks to you about
31:35 how these large language models work how
31:38 to talk to them it gives you some
31:39 prompts you can copy and paste it'll
31:41 give you a lot of ideas
31:43 um so so that's that stuff
31:44 okay
31:46 um this is what I call the kids table so
31:49 these are are ones that I think are a
31:51 little uh
31:54 less useful on a day-to-day basis but
31:57 still really interesting
31:59 so bing.com is in the doghouse that's
32:02 why it's dark gray and you can barely
32:05 see it
32:06 um because the safety guard rails on it
32:08 just make it a steaming pile of poo
32:10 right now
32:12 um perplexity.ai is a really interesting
32:14 site and service labs.perplexity is free
32:17 hosting of meta's llama 2 models so
32:20 that's worth going and playing around my
32:22 mine looks like a rocket exactly Silver
32:25 Fox exactly mine looks like a rocket
32:28 well I have a very tall rocket I'm
32:30 flying to Mars
32:32 it's exhausting
32:34 um
32:36 Pi Pi dot a I
32:38 um that's a company called inflection
32:40 they just raised 1.3 billion dollars so
32:43 that's going to be a significant player
32:44 if you download the PI app
32:47 um you can put it into conversation mode
32:49 and just talk to it it's connected to
32:51 the internet and you can talk to one of
32:53 these GPT models it's designed for
32:55 conversation it's really interesting for
32:57 that and then Poe is one where you can
33:00 play with multiple models in the same
33:01 interface and it's got a bot maker and
33:04 and so so that's it so those are the
33:06 seven
33:06 large language model things I think are
33:08 worth playing with right now this list
33:11 the you know these two lists
33:14 um are probably going to get shaken up
33:16 here in a week or two because just
33:19 shit's changing
33:21 so that's the that's the word stuff now
33:23 the picture stuff
33:25 um if you want to make pitches
33:27 again this is going to change
33:29 probably in a week this list will not be
33:32 the same so if if you want this list
33:35 screenshot it now because it's gonna go
33:37 away this could be a collector's item we
33:40 could turn it into an nft you can buy an
33:43 nft and then
33:46 exactly you see that's what this channel
33:49 is really about what this really is is a
33:51 multi-level Market I mean a pyramid
33:53 scheme
33:54 and you level up and when you level up
33:57 your downline gets into your up upline
34:00 as your the upline is the in the
34:02 downline and anyway all right
34:05 so
34:07 there's people in here going is this
34:09 really a multi-level marketing scheme oh
34:12 Winston Winston knows how to pay Winston
34:14 knows how to get make it rain in here
34:16 you see what I'm saying
34:18 see what I'm saying this is not this is
34:20 not a game people
34:22 it's not a game
34:24 always be prompting
34:29 I I got nothing tonight just just bear
34:32 with me
34:33 um okay
34:35 um mid-journey in terms of quality is
34:38 the best of these of these image
34:39 generation tools ideogram
34:42 is not great at the image generation but
34:44 it's really good well not really good it
34:47 is the first of these large language
34:49 models that can do words like that
34:51 and that and that and that and that and
34:55 that and that and that all these were
34:57 done in ideogram
34:59 all of the tools
35:01 in AI right now are janky pieces of [ __ ]
35:05 thank you to one of the Irregulars who
35:06 made me that it's lovely
35:08 that was made in ideogram
35:12 um
35:13 leonardo.ai is is a really good
35:15 interface if you want a good powerful
35:17 interface on top of stable diffusion but
35:20 you don't want to figure out how to
35:21 install automatic 1111 in a Google
35:24 collab notebook because you didn't get a
35:27 computer science degree and you're like
35:30 can I just make pictures you can at
35:32 Leonardo
35:35 um heygen.com that's the one that I
35:38 showed earlier that uh you know I'm
35:41 speaking Chinese
35:43 foreign
35:50 .com actually labs.hagen.com and there's
35:54 a video translate tool there
35:56 um and then
35:57 um Adobe Firefly and Photoshop
35:59 generative fill if you've got Creative
36:01 Cloud licenses they have now generally
36:03 released their generative generative AI
36:05 tools so they're available for
36:07 commercial use and they haven't been
36:09 trained on copywritten work so if you're
36:12 in a company if you're in a corporation
36:14 Adobe Firefly and and Photoshop
36:16 generative feeling good
36:18 openart.ai promptbook thank you
36:21 apocryphal Apocrypha
36:24 um thank you very much
36:26 um
36:28 The Prompt book thing there will teach
36:29 you how to talk to these image models
36:32 it's designed for stable diffusion but
36:34 it'll work for all of them because
36:35 they're all basically the same
36:37 technology
36:38 um
36:40 I'm also the one of the co-hosts of a
36:43 thing called the AI Salon which is a
36:45 community of AI adventurers and we've
36:48 got a Discord channel so if if you want
36:50 a place to hang out and talk to people
36:52 that are into this AI stuff and you want
36:55 to ask questions last night at the we we
36:58 have a meeting every two weeks last
37:00 night was our our uh our bi-weekly
37:02 meeting and we launched this thing
37:04 called guilds so they're they're sub
37:06 communities within the salon so we've
37:08 got a writing Guild and a making Art
37:12 Guild and a AI 101 Guild it's really
37:15 cool so go to the salon.ai the links are
37:18 there for that
37:20 um what else oh I've been invited by
37:23 another content creator named Rhonda
37:26 Phillips who's got a community of
37:29 creative professionals she's a voice
37:30 actress and a creative coach
37:32 and so I'm doing a
37:36 um Workshop that actually starts a week
37:38 from tonight called the AI launch pad so
37:41 it's it's a it's a workshop for AI for
37:44 creatives
37:46 um over three weeks it's two hours once
37:48 a week on a Wednesday night starting
37:50 next Wednesday
37:52 and what time is it seven to nine
37:55 Eastern so that's five to seven
37:58 that's five to seven mountain time where
38:00 I am so I'll still I'll still do my Tick
38:02 Tock lives even though I'm doing
38:03 workshops
38:05 um starting next week on Wednesdays
38:08 um and it's gonna be good we're gonna do
38:09 kind of an intro to AI it's the first
38:12 Workshop applying generative AI to your
38:15 business is the second workshop and then
38:17 Advanced AI like you know voice
38:19 synthesis and animation and video and
38:22 things like that is going to be the
38:23 third Workshop so it should be a lot of
38:24 fun so the information's there at
38:26 Rhonda's voice.com events and then what
38:30 else I think that's about it that's good
38:32 that's good that's a good place to start
38:34 if you if you have questions about AI
38:36 pop them in the comments below I'll do
38:38 my best to answer them we've got some
38:39 some mods in here we call them the
38:42 Irregulars they just show up here night
38:44 after night and harass me they don't
38:46 harass me they they they they are
38:48 incredibly supportive and and awesome
38:50 and so thank you all whoever's modding
38:52 tonight thank you for doing that
38:55 um sorry had an add moment had to throw
38:57 some garbage out I'm back
39:01 chat add
39:07 um I like taters taters tasty
39:11 they call it a sling blade
39:14 oh that's Ally That's Ali right there
39:18 Ali the AI learning lab
39:20 she's so cute
39:23 all right we'll put her right there
39:26 ion China good evening Bigfoot Electro
39:29 that's hilarious
39:31 Amway Amway AI exactly exactly this is
39:36 this is the multi-level every time you
39:39 do a prompt every time someone in your
39:41 downline does a prompt you get a tenth
39:44 of a penny now if you just had to make
39:45 10 million prompts you'd have a million
39:48 bucks oh no you wouldn't you'd have a
39:50 hundred thousand I don't know I'm not
39:51 good at math but Chad gbt could tell you
39:54 that
39:59 um
40:01 all right now I just sent out to my
40:04 friends and said not a money thing
40:09 that's really funny and then they come
40:11 in that's it's one of my joys in life is
40:13 is when I do one of my stupid characters
40:15 just knowing that someone is sort of
40:17 flicking through Tick Tock and just
40:18 seeing like how do you work this thing
40:20 March is this is this tick tock on
40:23 foreign
40:25 oh man
40:27 all right let's see it's so exciting to
40:30 see the changes happening so fast pretty
40:31 much daily yeah it's you know I kind of
40:34 feel like
40:37 kind of from mid-july to August things
40:40 kind of stabilized like there weren't it
40:43 wasn't like this daily barrage I feel
40:45 like
40:46 you know January through March we're
40:48 just [ __ ] insane where it was just
40:50 daily daily announcements of just crazy
40:53 crazy stuff and then it lulled for a
40:56 little bit and then May and June got big
40:58 again and then kind of July and August
41:00 it got a little normal and boy
41:04 yeah
41:05 man the uh what's coming like September
41:08 was decently quiet but but you heard we
41:10 started hearing some Rumblings of what
41:12 was going on
41:13 and uh
41:16 yeah I just you know the thing for me
41:19 about about what's going on with all
41:21 these changes is just
41:24 I feel like all of us like myself
41:26 included like I kind of had my head
41:29 around
41:30 okay so so given what these things are I
41:33 kind of get what they are and I kind of
41:35 get where their limitations are and I
41:37 kind of get how I can use them in my
41:38 business
41:40 and now I'm like okay I no longer know
41:43 where the limitations are and no longer
41:45 know where the boundaries are when
41:48 Gemini comes out from Google if indeed
41:50 it's even as good as chat gpt4 much less
41:54 better than chat gbt4
41:56 if it's as good as
41:58 well where is it as good as is it is it
42:02 better at coding is it better at
42:04 language is it better at you know doing
42:06 basic math does it hallucinate less does
42:09 it hallucinate more
42:11 is it is it tightly integrated with
42:14 YouTube which could completely transform
42:16 the kind of applications you could
42:18 create like like so I just feel like you
42:21 know I've mentioned before that I feel
42:23 like the the stage of the industry we're
42:25 at right now is this primordial soup
42:27 right and it's and it's like that thick
42:30 sort of bloop you know and and I think
42:33 we're just like right now we're just
42:34 about to have like one of the bubbles
42:36 sort of come up and go bloop and just
42:38 shoot [ __ ] everywhere and just kind of
42:41 upset upset what what seemed like
42:43 equilibrium for a moment there so yeah
42:46 it's just crazy just crazy so
42:49 um it's one of the other things that I
42:50 talk about a lot on this channel is
42:53 if what you're looking for right now is
42:56 what's the best course I should take or
42:58 what's the best like who's the expert
43:00 who's the this who's there are no
43:02 experts right now like even the workshop
43:03 that I'm going to teach starting in a
43:05 week
43:06 like I'm thinking long and hard about
43:08 what I actually teach there because if I
43:10 teach anything too Tactical
43:12 it's going to be useless in six months
43:15 right but if I teach if it's too
43:17 esoteric and just like you know be Zen
43:20 with with your exploration of AI then
43:23 that's completely useless I I need to
43:25 sort of you know figure out how to how
43:27 to walk that thread but I would say be
43:31 very very skeptical of anyone right now
43:33 that claims to be an expert at this
43:36 stuff no one is an expert at this and
43:38 then on the flip side of that coin if
43:41 you're feeling like oh I'm so far behind
43:43 I don't know what I'm doing and I'm not
43:45 technical that it doesn't matter
43:48 you're not behind we're all behind right
43:51 these these tools are coming at us
43:53 they're coming at us hard
43:55 what they make possible we have no idea
43:59 what five years out looks like much less
44:02 you know even a year out
44:05 um so you're not behind and the whole
44:08 purpose of this channel is just to say
44:10 welcome welcome to the madness it was
44:13 funny I got a I got a direct message
44:15 from someone today that said
44:18 um you know I stumbled on your channel
44:19 and you know I was really resistant to
44:22 AI so like she was in a she was in in
44:24 this sort of demonization place of like
44:26 ai's evil ai's bad and she said she's
44:30 slowly coming around to oh maybe this is
44:32 just a tool and I need to start
44:34 experimenting with it and you know it's
44:36 this isn't so black and white it's not
44:38 black and white these things are
44:40 profoundly powerful and very flawed
44:42 they're there there's things that they
44:44 do really horribly just bad they're just
44:47 not good and then there's other things
44:49 that they do that are face meltingly
44:51 powerful
44:53 the thing about them is that
44:56 as
44:58 as flawed as they are right now
45:01 how they are today is as bad as they're
45:03 ever going to be
45:06 they only get better from here
45:09 so start
45:11 understanding what they are and how they
45:14 work and it will serve you I'm telling
45:16 you I feel like it's the next 18 months
45:20 anyone anyone right we we've had a
45:23 couple of people in here uh more than
45:26 one person that have come in and like
45:27 I'm in retirement but I'm think I'm
45:29 coming out now like I'm into this let's
45:31 let's do something
45:33 like there's a lot of people right now
45:34 rediscovering a passion for something
45:39 because we're at the birth
45:42 of a thing that I think we all
45:44 instinctively know is gonna
45:45 fundamentally transform how we do
45:48 business the definition of work
45:51 the definition of productivity the the
45:55 the it's going to completely upset the
45:57 apple cart in where the value is in the
45:59 value chain
46:03 and so we're all just trying to figure
46:04 it out it's tremendously exciting it's
46:07 tremendously terrifying like I you know
46:10 I
46:12 it's funny you know you talk to some
46:13 people and they're just like absolutely
46:15 scared or they're absolutely pissed off
46:17 and then you talk to other people and
46:18 they're absolutely excited and they're
46:20 all like ah far out it's gonna stay yeah
46:22 I tend to lean more toward that side but
46:25 where I find myself is I'm both right
46:29 like it's like this is tremendously
46:31 exciting and holy [ __ ] is that scary
46:35 what does it mean if the advertising
46:38 industry
46:40 loses the time and materials model that
46:44 their entire industry is founded on
46:47 what does that mean
46:50 I don't know
46:53 does it mean the big holding companies
46:54 right now
46:56 flail if they don't adopt fixed pricing
47:01 Maybe
47:04 how long can they hold on to that model
47:08 of here's 50 copywriters sitting in a
47:11 pen working on your Social Media stuff
47:13 we're going to charge you you know 120
47:15 ah dinosaur that's very good that's very
47:18 very good exactly
47:22 how many Kodak moments
47:24 are happening right now I they're
47:28 happening in universities they're
47:30 happening in companies
47:33 you know the Kodak moment right they
47:35 [ __ ] invented the digital camera
47:41 and the engineers came in and said
47:43 he has a digital camera sirs bah humbug
47:48 pixels we're in the silver halide
47:52 business
47:56 10 years later 20 years later they're
47:58 out of business
48:00 how many moments are gonna happen you
48:03 know
48:04 this year next year the following year
48:07 where
48:08 arrogance and hubris of well we're us
48:11 this won't affect us
48:16 and then some Punky little
48:18 20-somethings or 30 somethings or a pile
48:21 of gen xers are going to get into a room
48:23 together and they're gonna huh what if
48:25 we started a 10 person company
48:28 just to [ __ ] with the ad industry
48:33 you have a bunch of EX ad execs that
48:35 that know how to Market and how to [ __ ]
48:37 with the ad they know where all the soft
48:38 spots are on the armadillo
48:42 I mean sounds kind of fun to me
48:48 and then how many of those 10 person
48:49 shops are going to turn into 100 person
48:51 shops and then 500 person shops and that
48:54 500 person shop
48:56 can do the work at the same or better
48:59 quality than a 50 000 person holding
49:02 company
49:07 everything is relevant that you're doing
49:09 I'm going through the YouTube videos oh
49:11 awesome Obi-Wan yeah so oh that I forgot
49:14 to say that so
49:15 um
49:17 I'm making archives of these lives
49:20 because you know it's funny these these
49:22 are I mean
49:24 they're at least preserving moments in
49:26 history right like like like
49:28 everything before tonight
49:31 um Chachi PT was just text and
49:33 everything from kind of tonight forward
49:35 Chachi pitian just text anymore like
49:39 it's going to be interesting to be able
49:40 to go back and say what were we talking
49:42 about when chat GPT was just text oh
49:44 yeah we were talking about prompt
49:46 engineering that seemed like a big deal
49:47 huh we don't even think about prompting
49:49 anymore now the tool the fairies just go
49:52 off and do our work for us right uh
49:56 madcap thank you Kyle for doing this I'm
49:58 sure you helped much more than you know
50:01 by your
50:04 honesty oh thank you very much I
50:05 appreciate that yeah I try listen I one
50:09 of my pet peeves in life and in business
50:11 is the gratuitous use of Technology
50:14 um I hate it I think the This Crowd
50:20 I just want to smash my phone every time
50:23 I hear that and see that
50:26 um
50:28 you know this this is not this is not
50:30 black and white right now it's not ai's
50:33 great or it's not ai's evil
50:36 and it's not generative ai's perfect and
50:38 it's not these things are gonna it's
50:40 it's not any of that it's just that
50:44 if you pay attention to what the tools
50:46 are doing and how they're doing it
50:49 in my lifetime
50:52 from my understanding of computers I I
50:56 wrote my first line of basic
50:59 in uh
51:01 seventh grade in
51:04 1977.
51:06 and so I kind of knew about computers
51:08 before that and then I got into them
51:10 then
51:13 and everything about everything that I
51:15 know about computers
51:20 requires
51:22 human beings
51:26 to think logically
51:28 about
51:31 how to get the machines to do what they
51:33 want and they have to think of the logic
51:35 and they have to write the logic and
51:37 they have to Branch it and they have to
51:38 do these hierarchies
51:42 and this generative AI stuff is not that
51:47 it's some of How It's trained is a lot
51:50 of that but
51:53 the the fact that we can use our our
51:56 natural language our native languages
51:58 English in my case
52:00 as a programming language
52:03 just changes everything it changes
52:06 [ __ ] everything
52:09 right
52:10 because right now
52:12 right now it doesn't seem like that
52:14 maybe maybe sort of how profound it is
52:17 that you can say you know write me a
52:20 poem or write me a python application or
52:22 write me this or write me that it
52:24 doesn't seem that profound because
52:26 Chachi BT is kind of like this little
52:28 this little you know hermetically sealed
52:31 tool off to the side right we're still
52:34 doing our business and we still think
52:36 about writing papers and writing emails
52:38 and thinking about problem solving in
52:41 our business oh and then maybe I'll go
52:42 over here and play with chat GPT right
52:44 it's this thing off to the side
52:46 as these things get more integrated into
52:48 what we do
52:53 it will increasingly be that you just
52:56 ask for what you want and the machines
52:58 will go do it for you
53:00 and sort of bring you back the results
53:02 and then you just go no
53:04 Bad Bad Robot flick
53:07 or brilliant robot great
53:11 what does that mean
53:14 it's crazy ask ER yeah Sherry Banks I uh
53:17 my first uh computer was a trash 80.
53:21 hang on look I've got a do I have it
53:23 here
53:24 is it that hang on
53:27 let me get the the presentation up
53:32 um
53:33 [Music]
53:44 so this is this is the the beginning of
53:47 uh of another one of these two weeks
53:50 from now
53:51 um
53:52 this is my first day of math class in
53:55 1977. I walked into my classroom and it
53:59 kind of looked like this this is a
54:01 mid-journey representation of a 1970s
54:04 classroom with a computer on every desk
54:06 and on the desks were trash 80s that
54:10 that model right there with the Dual
54:12 floppy drives and the orange the the
54:14 orange on button
54:16 and the big giant clackety keys
54:19 and the first thing that we learned was
54:22 L print which you could print a line of
54:24 text to screen and my buddy Scott
54:26 buchart and I we made an animation of
54:29 ASCII art of a Rocket taking off back to
54:32 back to Rockets right mine looks like a
54:34 rocket
54:35 um we made an animation of an ASCII
54:38 rocket it was like a Saturn V like you
54:40 know scrolling up on the screen and then
54:41 the plumes of fire came out at the
54:43 bottom we were so proud of that and it
54:45 was one command we learned one command
54:47 and we were making cool animations with
54:50 it
54:51 um
54:52 and yeah and this was I I mean the the
54:55 thing that I was acutely aware of when I
54:57 was in seventh grade
55:00 was that
55:02 that machine as I learned with that
55:04 machine did I like what hit me was oh
55:07 [ __ ] this is like the printing press
55:10 and
55:12 I'm just too young to have been in on
55:15 the ground floor of that because
55:18 you know Steve Jobs and Wozniak were on
55:21 the cover of you know Newsweek and Time
55:24 Magazine and Bill Gates and the team at
55:27 Microsoft was was you know doing their
55:30 thing and I was in seventh grade and I
55:32 was like oh man I just missed like I
55:34 could have I could have been somebody I
55:36 could have been a player
55:38 and and that it like it broke my heart
55:40 and then and then I didn't think I'd see
55:42 another one of those in my lifetime
55:44 because I knew that
55:46 that these things don't come along very
55:50 often right the printing press was
55:52 whatever it was 500 years before more
55:55 than that
55:57 um
55:59 and it made me really sad and then in
56:02 1994 I stumbled on this thing called the
56:04 world wide web and I published
56:06 an online magazine
56:09 in my apartment in Brooklyn
56:11 and three weeks later
56:14 there was a full page article written
56:16 about it in Paris in Liberation on the
56:19 Parisian News Daily
56:21 and
56:23 I had I had I just I had an epiphany of
56:26 holy [ __ ] the world just changed and
56:29 holy [ __ ] the World Wide Web is like
56:33 this like this PC moment
56:36 and so I went all in on the world wide
56:39 web stuff and I co-founded agency.com
56:41 and I it was a crazy crazy time and I
56:45 didn't think I'd see
56:47 it again in my lifetime and then I
56:49 played with chat GPT and I was like holy
56:51 [ __ ] here it is again
56:54 there's like three of these in my
56:55 lifetime like I I if if you if you are
56:59 not acutely aware how lucky you are to
57:02 be
57:05 able to play with this stuff in this
57:07 moment it's it's an amazingly rare thing
57:09 and this one in particular
57:12 this is going to make the invention of
57:14 the personal computer and the world wide
57:17 web and the iPhone all of them seem
57:19 quaint right because this is just it's
57:22 the way this stuff is using the computer
57:24 is just a fundamentally different thing
57:26 so it's crazy
57:28 so anyway listen listen it's just not
57:32 that many of us in here tonight don't
57:34 tell anyone I get this cuckoo with this
57:36 stuff
57:37 he seems very obsessed with it did you
57:39 notice that I think he's a little
57:41 touched do you think so too do you think
57:44 so I don't think it's healthy I don't
57:45 think it's healthy I think you think
57:47 this is religion do you think he thinks
57:49 it's God oh I don't know he might
57:52 thank you
57:55 I'm sharing some deep gems some deep
57:58 something
58:00 lucky discs I had a TRS-80 with cassette
58:03 oh wow yeah that that looks good you
58:05 remember these
58:07 side quest drives
58:09 huh
58:11 oh yeah Urban desires that was my online
58:14 magazine
58:15 I think this is a 40 megabyte
58:20 it's a classic
58:26 [Music]
58:29 ours were IBM 5150s well you know what's
58:32 funny Clint so I had TRS 80s in seventh
58:36 grade and then my buddy Scott buchart
58:39 his dad was in real estate so he had a
58:40 lot of money so he had a he had an apple
58:42 II and then he had a Lisa he bought his
58:46 dad a Lisa for the real estate business
58:48 it was a ten thousand dollar computer
58:50 that was like a really giant shitty
58:53 Macintosh
58:55 so he had a Lisa and then when the first
58:57 Mac came out you know he and I were
58:59 playing on the Mac and we had modems and
59:01 we were calling up you know
59:03 um
59:05 it was cool you could you could call up
59:07 computer systems that didn't have any
59:09 Security on them you could just download
59:10 lists of phone numbers so I remember my
59:13 my only actual hacking experience we got
59:16 one of those lists and and basically you
59:18 would you would get one of those lists
59:20 and then you would just have the
59:21 computer just randomly call phone
59:24 numbers until it hit a phone number
59:26 where it logged into a computer and so
59:29 we did that and we logged into the
59:31 control system of a mall in North
59:34 Carolina
59:37 and we you know dug through the menu
59:39 systems and we turned the heat up to 99
59:41 and we reverse the escalators we
59:44 reversed the direction of the escalators
59:46 so ah those were good times all right I
59:51 had chat wait I tell chat GPT to make my
59:54 email casual professional and I've been
59:57 100 satisfied with the outcome yeah it's
59:59 great at that stuff it's also good if
1:00:01 you haven't tried it silver fox
1:00:03 um if you've got an email that you've
1:00:06 written or anything that you've written
1:00:07 that you're really happy with you know
1:00:09 paste paste it in there as an example
1:00:10 and just say follow the tone and
1:00:12 structure of that other thing I wrote
1:00:14 it's really good at that as well we
1:00:17 still use Fortran 77 in the HPC world
1:00:19 yeah yeah yeah I know well and that's
1:00:22 another place where these where the the
1:00:24 large language models you know because
1:00:26 you've got all these computer languages
1:00:27 that are that are fading where you where
1:00:29 you have you don't have expertise for
1:00:31 people that can do Fortran and and
1:00:34 Cobalt and things like that in these old
1:00:36 Mainframe systems that are still still
1:00:38 out there
1:00:39 um you know I think the the the ability
1:00:42 to
1:00:44 um refactor code or just to even
1:00:46 understand and document code
1:00:49 um that may be of a previous era could
1:00:52 be really really powerful
1:00:55 I hope my persistence comes off as
1:00:57 nothing short of professional absolutely
1:01:01 schedule a call yeah drop me a line
1:01:03 happy to do it happy to chat
1:01:05 um compact portable the size of a sewing
1:01:07 machine I don't think I owned
1:01:11 did I own the the Mac
1:01:14 they called it the luggable
1:01:16 I don't think I owned one but I think
1:01:18 some place I worked had one so I carried
1:01:20 one around a couple of times those those
1:01:22 [ __ ] were heavy
1:01:23 they were really heavy and the screens
1:01:26 were really dim they weren't back
1:01:27 backlit I think they were if they were
1:01:30 backlit it was like with a with like a
1:01:32 Bic lighter this is pretty bad
1:01:37 um we had apples in fifth grade very
1:01:39 cool yeah that I mean the Apple II was
1:01:41 was they did really good stuff in
1:01:43 education I guess Apple's still doing
1:01:44 good stuff in education but they kind of
1:01:47 owned that market for a good while
1:01:51 um Lan parties oh yeah
1:01:54 all right that's what I that's what I do
1:01:56 I'm the email Queen oh that's great
1:01:57 that's awesome toe dance where can I
1:01:59 find better prompts to get my
1:02:01 mother-in-law to act more like a human
1:02:05 actually you know what you can probably
1:02:07 do is is you could say here's the issues
1:02:09 I'm having with my mother-in-law give me
1:02:11 give me some strategies to uh to uh to
1:02:15 win her over
1:02:17 or be able to tolerate her that's really
1:02:20 funny uh
1:02:25 um
1:02:26 Dr J it passed the Kyle test and he's a
1:02:28 cgo so that that's a that's a funny
1:02:31 thing so
1:02:33 one of my titles right now is I'm a
1:02:35 chief generative officer which sounds
1:02:37 very [ __ ] impressive doesn't it
1:02:39 um I like it it's a it's a title that
1:02:41 makes me happy it's for a uh
1:02:44 an organization called content Evolution
1:02:46 which is a sort of a loose Federation of
1:02:48 marketing companies that are just kind
1:02:50 of there to co-market with one another
1:02:52 but I built I built something for
1:02:54 Content Evolution if you go to
1:02:56 contentevolution.net on the home page of
1:02:59 contentevolution.net is is this thing
1:03:01 called the uh challenge engine so you
1:03:04 give it a business Challenge and then
1:03:05 it'll generate a series of strategic
1:03:08 questions that you should be asking
1:03:10 yourself to solve that problem
1:03:12 and so along with some other folks in
1:03:15 content Evolution we built that tool and
1:03:17 so when we delivered that tool the guy
1:03:20 that runs content Evolution goes I'm
1:03:22 going to promote you to Chief generative
1:03:23 officer I'm like I'm in so
1:03:27 so I think I'm the first Chief
1:03:28 generative officer on the planet
1:03:32 you didn't think I had qualifications
1:03:34 cgo baby
1:03:38 amateurs amateurs you think I'm just an
1:03:41 ex-actor
1:03:42 it's I am
1:03:44 [Laughter]
1:03:49 cheers
1:03:53 all right let's see let's say say let's
1:03:57 get some more questions going here
1:03:58 tandy1000 HX love it the Osborne one I
1:04:01 remember that thing
1:04:03 swear it was a sewing machine case it
1:04:05 probably was you know what my my my lust
1:04:08 machine
1:04:10 when I was in seventh or eighth grade I
1:04:12 wanted the Timex Sinclair I think it was
1:04:15 a thousand CX it was a computer that was
1:04:18 about this big and it had a membrane
1:04:20 keyboard on it
1:04:21 it was here I can probably go find one
1:04:23 we have the internet
1:04:26 time x
1:04:28 Sinclair 1000 yeah man
1:04:32 and it was like it was like 129 dollars
1:04:35 and it was like that was way out of my
1:04:37 [ __ ] price range
1:04:40 [Music]
1:04:42 look at that bad boy look at how cool
1:04:44 that thing was
1:04:47 wait
1:04:50 see it had it it had all these shapes on
1:04:53 the key so you had the letters but you
1:04:54 had all these shapes that you could do
1:04:56 ASCII art with and then it had all this
1:04:58 other [ __ ] like all this programming
1:05:00 [ __ ]
1:05:01 and it was all just in this tiny little
1:05:03 box come on
1:05:07 that was a [ __ ] future man
1:05:10 never got one I should buy one now
1:05:12 except with my 80d I'll probably spend
1:05:15 like three months like just hyper
1:05:17 focusing on it and
1:05:19 just waste my life away so it's good I
1:05:22 don't have one
1:05:24 Timex Sinclair from the drugstores 4K
1:05:27 extra external RAM module exactly
1:05:30 jockster this is the link you're looking
1:05:33 for what is it
1:05:36 oh yeah
1:05:42 when you're ready to travel Macintosh
1:05:45 can easily go along
1:05:46 the durable carrying case Has Fallen
1:05:50 your Macintosh keyboard mouse manuals
1:05:53 and so on
1:05:57 even fully loaded the carrying case is
1:06:01 light enough to go anywhere you go hey
1:06:03 dang it
1:06:05 holes
1:06:08 um
1:06:13 come on
1:06:20 oh come on
1:06:22 it ain't working now
1:06:25 is it
1:06:27 and when you're ready to travel
1:06:29 Macintosh can easily go along the
1:06:32 durable carrying case is strong and
1:06:34 waterproof
1:06:35 your Macintosh keyboard mouse manuals
1:06:38 and software all fit neatly inside it's
1:06:41 like a cooler
1:06:46 [Laughter]
1:06:49 the carrying case is light enough to go
1:06:51 anywhere
1:06:52 [Laughter]
1:06:57 oh man
1:06:59 that's awesome
1:07:03 uh
1:07:05 you had Clint had the Timex Sinclair
1:07:07 okay I hate you Clint I'm jealous of you
1:07:10 in retrospect
1:07:12 did you ever see the Microsoft Excel
1:07:14 elevator commercial now I didn't
1:07:17 um optional audio cassette interface oh
1:07:19 man you were you were you were [ __ ]
1:07:21 loaded man you had all the toys
1:07:24 could fit that on a plane with you for
1:07:28 AI assisted encoding like copilot and
1:07:30 others how well do they protect company
1:07:32 code and IP ah so joaquim I think it's
1:07:35 Joachim or Joachim Joachim or Joachim
1:07:40 um
1:07:46 offend your audience in one easy step
1:07:48 with Kyle Shannon
1:07:50 um so listen Joachim
1:07:52 um
1:07:55 with these tools right now unless they
1:07:59 explicitly say it in writing like a
1:08:02 legal document
1:08:05 assume that all of these things have no
1:08:08 privacy and security
1:08:10 if you've got to send your data to Bing
1:08:13 or to open AI or to
1:08:16 Google bard or wherever just assume that
1:08:19 it's not secure
1:08:21 um
1:08:23 chat EPT has a new thing called chat GPT
1:08:25 Enterprise where they're starting with
1:08:27 people that can spend 10 or 15 grand a
1:08:29 month and then it'll come down to to us
1:08:31 plebes at some point
1:08:34 um
1:08:35 or if you do a standalone instance on
1:08:37 Microsoft Azure or something on AWS
1:08:41 where you know you've got a hermetically
1:08:43 sealed environment where you're you're
1:08:45 running these large language models you
1:08:47 know in some sort of thing
1:08:48 assume that assume that the sort of
1:08:52 consumer versions of these things are
1:08:54 not secure and so does it mean you can't
1:08:56 use them no it doesn't but but what it
1:08:58 might mean is you go through and
1:09:01 de-personalize a document so maybe
1:09:03 you've got a legal document and instead
1:09:05 of the person's name you put in Mickey
1:09:07 Mouse right and you know you you take
1:09:10 out real addresses and then you go
1:09:12 process the document like that so so
1:09:15 that's anything that I'm doing that
1:09:16 involves client data I'm depersonalizing
1:09:19 it if I'm dealing with anything that is
1:09:21 not in a standalone hermetically sealed
1:09:23 thing so I would I would just assume
1:09:27 um
1:09:28 just assume it's not safe assume it's
1:09:31 not secure because otherwise you'll end
1:09:33 up like that group out of Japan and
1:09:35 Samsung where they they put a bunch of
1:09:37 stuff up on chat CPT and it got it got
1:09:41 snacked up by the community who
1:09:44 hackers gonna hack
1:09:47 and so yeah
1:09:48 so
1:09:52 [Music]
1:09:55 I am the oldest before color TV there
1:09:58 was a show to put plastic over the
1:10:01 screen and color over black and white
1:10:03 yeah I remember that it was a Magnavox I
1:10:05 had
1:10:06 my stepdad got that system the Magnavox
1:10:10 Odyssey
1:10:11 and we had like a 27 inch like one of
1:10:14 those big console TVs it was like a 27
1:10:17 inch TV which was Giant at the time
1:10:19 because it was a tube
1:10:21 right and it had all the the election
1:10:23 like it would it would literally shock
1:10:25 you if you walked up to it because it
1:10:27 was just a giant like Tesla coil just
1:10:29 basically ready to electrocute you at
1:10:32 any moment and then yeah the the
1:10:34 Magnavox
1:10:35 Odyssey video game for like the duck
1:10:38 shooting game
1:10:39 you'd put the big piece of plastic it
1:10:42 would static electricity to this Tesla
1:10:44 coil TV and then the sky would be blue
1:10:48 and the water would be you know the the
1:10:50 grass would be green on the bottom it's
1:10:52 [ __ ] awful
1:10:56 and like the controllers would be like
1:10:57 clack clack clack
1:11:03 oh my God
1:11:05 remember so all there's there's enough
1:11:07 gen xers in here one of my one of my
1:11:10 great
1:11:11 um
1:11:12 one of my Great disillusionment Moments
1:11:15 you know you there's those moments where
1:11:17 you grow up right so so this was in the
1:11:20 kind of late 70s early 80s you know you
1:11:23 had arcades you know like physical
1:11:26 arcades where you'd go to a building
1:11:27 with machines in it and you put quarters
1:11:29 in it and you play video games
1:11:31 and then the Atari 2600 came out
1:11:35 and they're like oh it's so exciting and
1:11:37 you can play video games at home and
1:11:38 it's like the arcade experience at home
1:11:40 and then they made the boxes
1:11:43 the artwork on the boxes looked like the
1:11:46 [ __ ] arcade games and then you'd get
1:11:48 the game home and it was like six blocks
1:11:50 bouncing around scream and you're like
1:11:53 this isn't this isn't an arcade game and
1:11:57 it was just like I kept I kept getting
1:11:59 suckered into buying
1:12:02 Atari 2600 games because someone would
1:12:05 come out with clever box art and I'd go
1:12:07 maybe this one's different and I look at
1:12:09 the back and I'm like yeah this one's
1:12:11 gonna be better when I go play it no
1:12:13 it's still just shitty arcade games
1:12:19 oh man
1:12:21 my first Mac came with one megabyte of
1:12:24 ram I uploaded upgraded to two megabytes
1:12:26 for five grand so my first Mac it was a
1:12:30 512 ke
1:12:32 and I think it was my buddy Scott
1:12:33 buchart the guy that that
1:12:35 um you know he and I did computer stuff
1:12:37 as a kid
1:12:39 um
1:12:40 he had a five megabyte hard drive and it
1:12:44 was actually it's funny I've got a box
1:12:45 right here that my lights on it was a
1:12:48 five megabyte hard drive this was about
1:12:50 the size of it and it was this steel
1:12:52 enclosure right it was a multi-planner
1:12:55 thing it's a five megabyte hard drive I
1:12:58 think he sold it to me for 200 bucks I
1:13:01 think he paid like 1500 for it and I
1:13:04 think he upgraded to like a 50 megabyte
1:13:06 hard drive and so I had I had a Mac 512
1:13:10 ke with an external hard drive I was
1:13:12 like baller man like yeah I got the
1:13:15 system
1:13:17 then I upgraded to an se30 with the
1:13:19 color monitor I don't [ __ ] around
1:13:25 oh man it's crazy and I like for AI
1:13:27 assistant coding how well do they
1:13:29 protect yep we talked about that
1:13:32 Microsoft can close your account and
1:13:35 your documents are lost
1:13:38 wait what
1:13:40 what is that in reference to
1:13:44 oh oh if they close your GitHub if
1:13:46 you're talking about GitHub co-pilot
1:13:47 yeah I assume so if if they should if
1:13:50 they if you do something that violates
1:13:52 their terms of service and they shut it
1:13:54 down yeah
1:13:55 so all these CL all these cloud services
1:13:58 are like that right if they if they go
1:14:00 down like when what was it the the flick
1:14:04 flicker was it flicker the the photo one
1:14:06 that went out of business and everyone's
1:14:09 [ __ ] photo libraries were all just
1:14:11 [ __ ] to bed you could easily get into
1:14:14 trouble with your company if you upload
1:14:15 any code yeah exactly yeah just with all
1:14:19 this stuff assume it's assume it's
1:14:20 dangerous until it's not
1:14:22 um
1:14:23 especially if you work for a big company
1:14:26 um you know wait for your it Department
1:14:28 to have to either create a Sandbox for
1:14:30 you
1:14:31 or do
1:14:33 work
1:14:35 tangential to what you do officially on
1:14:38 a computer that's not a work computer
1:14:40 just so you can learn this [ __ ] if you
1:14:42 don't have you know official places to
1:14:44 play I did a workshop for a big
1:14:47 um uh Energy company not an energy
1:14:50 company battery company
1:14:53 um
1:14:55 a couple of months ago and it was for
1:14:58 their Marketing Group and the IT the IT
1:15:00 department basically installed you know
1:15:02 a special secure version of chat GPT and
1:15:07 Bing and like all the like the Microsoft
1:15:10 version of chat GPT
1:15:12 um just for the marketing part
1:15:14 department to do this Workshop but they
1:15:16 had access to it they you know it was
1:15:18 blessed by it and things like that so if
1:15:20 you've got that cool but just assume
1:15:22 it's a mess
1:15:23 let's see that's interesting I asked
1:15:25 about cleaning up my CRM a few nights
1:15:27 back using GPT oh yeah oh yeah Paulie z
1:15:32 um
1:15:32 hang on
1:15:33 to organize my CSV file so
1:15:37 come on
1:15:38 oh wait someone could someone re-pin
1:15:40 that because I I couldn't read the
1:15:42 second half of the thing because of tick
1:15:44 tock's wonderful interface they've
1:15:46 they've got icons that are literally
1:15:48 like three pixels wide and like you have
1:15:51 to wait there it is organize my CS5 so I
1:15:54 probably got
1:15:56 probably shouldn't do this a
1:16:00 um oh
1:16:10 I don't know enough about how
1:16:13 um code interpreter works
1:16:17 um
1:16:18 it's now called Advanced data analytics
1:16:20 but I'm pretty sure
1:16:22 that
1:16:26 when you fire up code interpreter in
1:16:29 chatpt it's it's it's gen it's uh it's
1:16:33 creating a container
1:16:34 so it's basically creating a little
1:16:37 hermetically sealed
1:16:39 um operating environment where it's
1:16:42 running all that python code so I don't
1:16:44 know that any of that's getting sent
1:16:46 back out but I don't know pay Pate might
1:16:48 know that more than I do
1:16:51 um
1:16:55 yeah I don't know I don't know if
1:16:59 if it's CRM data
1:17:05 I don't know I don't know that that's
1:17:07 that's a
1:17:12 you know
1:17:13 here's the deal policy I would go to how
1:17:17 big is your company what's your
1:17:18 tolerance for risk what's their
1:17:20 tolerance for risk if it's just if it's
1:17:22 a small organization and it's and it's
1:17:24 your company and you're like well if you
1:17:27 know someone gets this information like
1:17:29 is that gonna be
1:17:31 catastrophic or is it going to be like
1:17:33 well it would suck but it's actually
1:17:34 kind of fine then like it's a call you
1:17:37 got to make
1:17:38 um
1:17:39 but I would I would I would
1:17:42 research how how code interpreter
1:17:46 actually works because my understanding
1:17:48 of it is it's running inside a container
1:17:51 in in sort of a like a temporary little
1:17:53 Run State
1:17:55 um that's doing all of that processing
1:17:57 so I don't think it's sending a bunch of
1:17:59 stuff out but I could be very wrong so I
1:18:01 I don't don't listen to me on that do
1:18:04 some research on that and then you need
1:18:06 to figure out where your risk tolerance
1:18:07 is so
1:18:09 yeah yeah
1:18:10 all right all right all right all right
1:18:14 all right
1:18:15 20 megabyte hard drive four megabyte Ram
1:18:18 no modem yeah
1:18:22 they have to wait they have new terms of
1:18:24 service valid from November who do open
1:18:27 AI does mantred
1:18:31 um vintage Visionary good I love that
1:18:34 name it's 10 out of 10 name
1:18:36 I bought an early 512 added a monster
1:18:38 Mac conversion yep nice that's solid
1:18:44 um
1:18:47 [Music]
1:18:55 bite magazine just for the free game
1:18:57 code each month yep I remember that
1:19:03 um
1:19:05 [Music]
1:19:05 um
1:19:13 oh it's Joe
1:19:15 Joe Kim ah okay cool but I'm not
1:19:18 offended I'm the same age as you gen Z
1:19:21 cool
1:19:22 um Joachim cool
1:19:25 um oh it is so so I said Joe a Chim it's
1:19:27 go
1:19:28 Joe a Kim cool that's a cool name
1:19:33 um
1:19:33 but yeah depending on where you're from
1:19:35 could easily be Joaquin but Joe Kim
1:19:37 that's cool Microsoft can punish you for
1:19:39 the content of your documents oh that's
1:19:41 interesting
1:19:43 I suppose I suppose anyone can right I
1:19:46 mean Tick Tock does they boot me off
1:19:47 this thing I've been booted off this
1:19:49 thing three times
1:19:50 because I because I said some code word
1:19:52 if there's certain code words where
1:19:54 you're just like I'm talking merrily
1:19:56 along and it's like zap you're gone for
1:19:59 a week
1:20:06 oh Gen X I mean yeah I figured I figured
1:20:10 that all right let's see let's see if
1:20:13 you wrap something they do not like grow
1:20:15 up when exactly Joker My Pal owns a
1:20:19 ginormous arcade with all the old games
1:20:21 it's so rad oh that's so cool Bruno yeah
1:20:23 it's those old arcade games are just the
1:20:26 best I mean they they just had such good
1:20:28 playability and
1:20:30 like like as much as I respect the new
1:20:34 you know 3D environments and all that
1:20:36 exciting stuff there's something about
1:20:37 just
1:20:39 like the speed and quality of like
1:20:41 centipede
1:20:42 and then millipede and
1:20:45 you know missile defense and asteroids
1:20:48 like asteroids the gameplay of asteroids
1:20:51 was so pure and it was like you just
1:20:53 became one with that machine
1:20:55 those were those were cool days
1:20:58 and it was like and it was it you know
1:21:00 that's it it's funny why arcades died is
1:21:04 because the home consoles got as good as
1:21:06 them right you didn't need in like in my
1:21:09 day in my day you had to go to the
1:21:12 arcade to have a good video game
1:21:14 experience you know the one at home was
1:21:16 like this sad depressing Wonka
1:21:19 bad thing and then you go to the arcade
1:21:22 and it's like
1:21:24 amazing
1:21:27 I remember typing in games and basic
1:21:29 from magazines into my Atari 400 that's
1:21:31 awesome
1:21:33 hahaha
1:21:36 um
1:21:42 [Music]
1:21:44 that's interesting I asked about oh yeah
1:21:46 that's right okay
1:21:50 um
1:21:52 [Music]
1:21:55 Missile Command and Dragon's Lair
1:21:57 Dragon's Lair Dragon's Lair was one of
1:21:59 the most frustrating games in the
1:22:00 history of video gaming
1:22:03 because all the other video games had
1:22:06 you know zero latency or like very
1:22:09 little latency and Dragon's Lair
1:22:12 you you had to hit the moves right but
1:22:15 there was like that half second delay so
1:22:17 you'd sort of like go up up left right
1:22:20 up and then and then it would go
1:22:23 and then and then it would play whatever
1:22:26 part of the laser disc it played
1:22:28 I mean it was really cool to look at but
1:22:30 the gameplay on it was so [ __ ]
1:22:32 frustrating because of that delay well
1:22:34 it's you know what's funny the um
1:22:37 all of the talking chat Bots right now
1:22:40 the the ones that are you know like like
1:22:42 pie and and whatever the chat CPT One's
1:22:45 Gonna Be that half second delay in them
1:22:47 is is is almost that identical weird
1:22:50 delay that Dragon's Lair had
1:22:53 and like that'll go away like that that
1:22:54 latency is going to go away probably
1:22:56 pretty quick but
1:22:59 but it's it's weird right now
1:23:02 you know what's funny I'll tell you what
1:23:04 here's where we are it's funny we're
1:23:06 talking about all this video game [ __ ]
1:23:09 we are in
1:23:12 let's see not that
1:23:19 this is kind of the the Apple II you
1:23:23 know PC Junior
1:23:26 kind of you know State like we're we're
1:23:29 at such an early stage of this like this
1:23:32 isn't even the Macintosh hit like this
1:23:34 is like the Apple II and maybe the the
1:23:37 multimodal version that can see and
1:23:39 generate images with Dolly three is
1:23:41 gonna be like the Macintosh right but
1:23:44 maybe not like maybe it's like it goes
1:23:47 from like the Apple II to to I don't
1:23:49 know
1:23:50 it's it's still kind of command lining
1:23:52 and weird right now
1:23:54 and that's that's the stage we're at
1:23:56 with this stuff so if we you know you
1:23:58 think back on how far computers have
1:23:59 come
1:24:01 since those really nostalgic early days
1:24:04 of you know your computer had to call
1:24:06 another computer I remember when when my
1:24:09 buddy Scott had a 300 baud modem you
1:24:11 could actually you know sort of watch it
1:24:13 you know type [ __ ] across the screen and
1:24:15 then we got a 1200 baud modem and it was
1:24:18 like you couldn't keep up with it and
1:24:19 then when you got you know 56k it was
1:24:22 just like
1:24:24 and it's like that's kind of the stage
1:24:25 we are with these things right now
1:24:27 they're kind of like 300 baud modem text
1:24:29 interfaces
1:24:32 and like think about you know a PS5
1:24:34 today versus that [ __ ] well what's going
1:24:38 to be the PS5 version of that
1:24:42 [ __ ]
1:24:46 I mean the PS5 version of that is going
1:24:49 to be make me a movie that's exactly
1:24:52 what I want it to be make me a video
1:24:54 game make me an immersive
1:24:57 anti-anxiety environment that I can
1:24:59 enter right now
1:25:01 I've got 12 minutes to to get my
1:25:04 amygdala unhijacked go
1:25:11 it's like being in the Dos days of AI it
1:25:13 is it's this is yeah we're in that we're
1:25:17 in Das
1:25:19 it's like what so look it's a experience
1:25:22 but but let's let's go look at something
1:25:26 um the best image generation tool right
1:25:30 now is mid-journey you have to you have
1:25:33 to make images in this thing called
1:25:35 Discord which has got like all these
1:25:37 like server these are like bulletin
1:25:39 board systems and then within each of
1:25:41 your little bulletin board systems you
1:25:43 have your stupid little text channels
1:25:45 and all these little text channels have
1:25:47 their little talky talk and people talk
1:25:49 about their crap
1:25:50 and then in here you have a mid-journey
1:25:53 channel
1:25:54 and and what do you do to make an image
1:25:56 slash imagine right slash your magic so
1:26:00 you have to know the
1:26:03 command line to make an image and then
1:26:05 prompt what do you say here I don't know
1:26:07 I'm gonna say uh uh let's see uh a
1:26:13 photo
1:26:14 of
1:26:16 let's see a
1:26:19 um prehistoric
1:26:22 Hedgehog
1:26:25 on a plateau
1:26:31 um
1:26:32 Ansel atoms
1:26:35 and then I get I'm going to do dash dash
1:26:38 AR
1:26:40 check this out dash dash
1:26:43 ar16 colon 9.
1:26:48 what's that stand for
1:26:50 aspect ratio 16 9 what's that that's the
1:26:54 size of a of a QuickTime movie how do I
1:26:57 know that because I'm a geek that like
1:27:00 the fact that we're having to make
1:27:01 [ __ ] images with command line
1:27:03 interfaces we are absolutely in in the
1:27:06 Dos era of AI right now this is just
1:27:09 [ __ ] stupid
1:27:11 laughs
1:27:13 what's what's the uh what's Ye Old AI
1:27:17 learning lab uh um term janky pieces of
1:27:23 [ __ ]
1:27:23 now we're gonna don't get me wrong
1:27:26 they're still remarkable we're gonna
1:27:27 look back fondly on these days and like
1:27:29 remember when you had to type in command
1:27:30 lines to make a picture
1:27:32 but there you go
1:27:35 it did it and they're fun they're good
1:27:38 look at that one I like that one let's
1:27:40 do the let's do the one in the little
1:27:42 frame
1:27:44 we'll upscale him
1:27:48 I won't bow
1:27:51 look at that
1:27:54 boom with a little command line we made
1:27:56 magic this is like the Commodore 64
1:27:59 being able to do video
1:28:02 was it the Commodore 64 yeah
1:28:06 wait what was it the Amiga what was the
1:28:09 big video machine that was the video
1:28:11 machine forever anyway
1:28:14 yeah this is how early we are this is
1:28:17 how early we are
1:28:18 someone's coming in here though we're so
1:28:20 early that it hedgehogs Ruled the Earth
1:28:22 no
1:28:24 [Laughter]
1:28:29 edlin prompts exactly
1:28:32 oh man
1:28:35 crazy that's a good one
1:28:39 they are learning lab
1:28:42 we're learning tonight actually we're
1:28:44 just kind of reminiscing tonight I like
1:28:45 it
1:28:48 um
1:28:51 it started talking nonsense about the
1:28:54 multiplicative identity property oh
1:28:58 well just copy a greater than b oh God
1:29:03 I asked that what the heck it was
1:29:05 thinking about and it started
1:29:07 apologizing or talking about Holodeck
1:29:10 anyone I got one in my garage five years
1:29:13 from now yeah
1:29:15 that's what's coming
1:29:18 have you seen I mean I've showed this on
1:29:20 the channel before but have you seen
1:29:22 um oh what the [ __ ] is it called
1:29:25 blockade labs
1:29:28 this is stupid and and this is the the
1:29:31 the actual technology here is relatively
1:29:34 simple but what it's doing is super cool
1:29:38 so if you go to what is it
1:29:42 something.lockadelabs.com
1:29:45 um Skybox Skybox
1:29:49 .blockidelabs.com so if you go to
1:29:50 skybox.blockaidlabs.com
1:29:55 beta072 this time we're introducing more
1:29:58 control
1:30:00 so we got an API
1:30:05 all right whatever
1:30:07 so
1:30:08 so this is a cool 3D World right neat
1:30:12 you can look up in the sky you look down
1:30:14 in the valley swell but I can also
1:30:18 it says dream up your world so I can go
1:30:29 space station deep in the cosmos
1:30:34 with an exploding planet in the distance
1:30:38 and Fighters flying through the skies
1:30:43 shooting at one another
1:30:47 and then let's see what style we'll go
1:30:55 [Music]
1:31:00 I guess we'll go
1:31:02 techno Noir cyberpunk
1:31:05 some of the better models
1:31:08 oh I've run out of generations oh this
1:31:11 used to be free
1:31:13 wait let me log in maybe
1:31:17 I get some credits if I log in
1:31:20 mm-hmm
1:31:22 come on
1:31:23 come on
1:31:28 I am so done with
1:31:33 17 Factor Authentication
1:31:40 uh
1:31:54 [Music]
1:31:59 why does he sing like that I don't know
1:32:02 Martha I'm pretty sure he's been
1:32:03 drinking it looks like coffee oh that's
1:32:05 what he says it is I'm pretty sure
1:32:07 that's not coffee boom boom confirm
1:32:17 do I have any available
1:32:19 um let's see what did I say
1:32:24 space station deep in the cosmos
1:32:27 exploding nebula in the distance
1:32:33 is it going
1:32:35 I think it's going yay
1:32:49 so there's their little default world
1:32:53 but then
1:32:56 boom
1:32:58 there's a little space world
1:33:00 that we just invented and it's sort of
1:33:03 3D this is actually not 3D this is just
1:33:05 a flat image that they're generating in
1:33:08 a way that it can be warped onto a
1:33:11 sphere
1:33:12 but there are people now that are taking
1:33:15 these 3DS or these you know mapped
1:33:19 spheres into Unreal Engine 5 and they're
1:33:23 adding objects into the thing and making
1:33:26 them making them so you can interact
1:33:27 with them
1:33:28 so
1:33:30 so right now these are kind of still
1:33:32 janky pieces of [ __ ] where it's this
1:33:34 isn't a real 3D world but we are within
1:33:38 spitting distance of being able to
1:33:41 describe a 3D World it'll generate it
1:33:43 describe the kind of gameplay that you
1:33:46 want it'll generate that describe the
1:33:48 kind of objects the kind of weapons the
1:33:50 kind of vehicles the kind of players the
1:33:52 kind of
1:33:53 um Network ability is it just your
1:33:55 friends is it strangers the kind of
1:33:58 scoring system you just describe it and
1:34:01 you'll be playing this hyper
1:34:04 personalized game that you want to play
1:34:06 and you can do that Solo or you can do
1:34:08 that with your friends right and so
1:34:12 and this is going to be everything this
1:34:14 is going to be movies there's going to
1:34:15 be music Imagine a Spotify that instead
1:34:19 of just being a playlist of songs that
1:34:22 you like that other people wrote It's a
1:34:25 playlist of those songs mixed with
1:34:28 original songs written for you based on
1:34:32 your emotion in the moment
1:34:36 okay
1:34:38 three years from now
1:34:40 that's gonna be Spotify or whatever you
1:34:43 know whatever is the new Spotify right
1:34:44 because Spotify will be like it's all
1:34:46 about real music and real musicians and
1:34:48 then some punks will come up with one
1:34:50 where it's like hey how about if you had
1:34:52 something that just composed music for
1:34:53 you on the fly in real time based on the
1:34:55 emotional
1:34:57 you know senses that we get from your
1:34:59 iris from your smart glasses
1:35:05 and then it's just you know every day is
1:35:08 just an endorphin injection we're all
1:35:10 [ __ ] you know hooked up so ready
1:35:13 player one we should all what we should
1:35:15 do here one night since we're on here
1:35:17 two and a half hours anyway we should
1:35:19 just start watching you know
1:35:21 Ready Player one and her start watching
1:35:25 movies together see if we can get the
1:35:27 copyright police to hunt us down
1:35:30 laughs
1:35:32 janky piece of [ __ ] oh yeah
1:35:37 oh man
1:35:39 very funny
1:35:43 I might upgrade to chat gpd4 actually
1:35:47 pay for it
1:35:48 I need to step up my game
1:35:50 listen it's I think it's probably worth
1:35:54 it Becky just because like if you want
1:35:56 to start playing with this stuff the
1:35:57 fact that it's now multimodal the fact
1:35:59 that you could just show it you know
1:36:01 take a picture of
1:36:03 some location somewhere or take a
1:36:05 picture of like it's I I just have no
1:36:10 idea how we're going to use it I mean I
1:36:11 I think a lot of what this channel is
1:36:13 going to be in the next month or so is
1:36:16 just kind of discovering in real time
1:36:17 together like what is this stuff how do
1:36:20 we play with it like where's the value
1:36:21 here
1:36:23 because I don't know
1:36:26 our grandkids will laugh oh oh yes
1:36:30 so cool
1:36:33 who what I don't I didn't see who sent
1:36:35 that I got I got uh
1:36:38 mesmerized by the blue whale jumping
1:36:40 through the screen that was very cool
1:36:42 thank you whoever that was that was
1:36:44 really awesome I appreciate that wow we
1:36:48 had sound effects and everything it
1:36:49 totally caught me off guard it was like
1:36:51 the other night if you weren't in here
1:36:53 like a week ago
1:36:55 um one of the one of the people who's a
1:36:58 regular in here
1:36:59 um decided to do some
1:37:02 um some Tick-Tock view purchasing on on
1:37:06 our behalf so we're like cruising along
1:37:08 like tonight we were like at 50 people
1:37:09 for like half an hour and then all of a
1:37:12 sudden I look up and we're like four
1:37:13 thousand I was like wait what and then
1:37:17 it went it went all the way up the
1:37:18 highest we hit was like 17 700 and it
1:37:22 was just a bunch of people coming in
1:37:24 that with you know I think it was just
1:37:25 an engagement farm so the the engagement
1:37:27 was really low but there was like you
1:37:30 know
1:37:30 ten thousand plus people in here for
1:37:32 like an hour and a half totally freaked
1:37:34 me out
1:37:35 so the whale didn't freak me out that
1:37:37 much but it was pretty cool all right
1:37:39 awesome
1:37:42 met Gates and Allen at a Sting concert
1:37:44 in the early 80s My Heroes at the time
1:37:46 yeah I I met um I met guy Kawasaki
1:37:50 fairly
1:37:52 fairly early in apples
1:37:55 in apples in in his tenure at Apple as
1:37:57 the as the lead evangelist for Apple guy
1:38:01 Kawasaki was a hero of mine Steve's jobs
1:38:03 was a hero of mine
1:38:05 um
1:38:06 yeah
1:38:07 yeah amazing amazing I mean just
1:38:11 like the those those those early days of
1:38:14 of all that stuff was just
1:38:18 you know you you listen to the was it
1:38:21 Noah Bushnell who started Atari and you
1:38:24 know you listened to Wozniak and you
1:38:26 listen to let's just all those early
1:38:28 guys what it was like it was it's a very
1:38:31 small
1:38:32 community and it you know when we were
1:38:34 doing the the web stuff in New York City
1:38:36 in the mid 90s it was it was a
1:38:38 relatively small community right it was
1:38:40 a community of I don't know
1:38:43 300 to 500 people that really had their
1:38:47 fingers on the pulse of what was going
1:38:49 on
1:38:50 and you know it quickly grew but it was
1:38:54 it was it was a really amazing time
1:38:56 because nobody knew this [ __ ] was coming
1:38:58 and we were all like holy [ __ ] if people
1:39:02 knew what was coming
1:39:05 they'd be acting different
1:39:08 it's just like what we're doing here
1:39:11 if people only knew pay attention people
1:39:14 pay attention
1:39:17 reported today for bullying an AI video
1:39:20 post not the poster
1:39:22 the bot the post was about
1:39:26 reported today for bullying an AI video
1:39:29 post that's really funny quite sure the
1:39:33 bot isn't gonna come alive and go read
1:39:35 comments oh man
1:39:39 um how did you learn mid Journey so you
1:39:41 have to find it inside Discord yeah so
1:39:44 listen they call me tater tot listen
1:39:47 tater tot here's the deal tater tot okay
1:39:50 for one thing can we talk about Tater
1:39:52 tot's name they call me tater tot
1:39:56 she uses camel case so you can read the
1:39:58 words she separates them with the period
1:40:00 and her name is they call me tater tot
1:40:03 it's a 10.5 out of 10. it's my first
1:40:06 north of 10 rating on a name it's the
1:40:09 best name in the history of names in
1:40:11 this channel I love it
1:40:13 I'll never forget it they call me tater
1:40:15 tot
1:40:17 um here's the deal
1:40:21 you don't have to learn mid-journey
1:40:24 because Dolly three's coming out and I
1:40:27 think Dolly three is a mid-journey
1:40:29 killer and it's gonna live inside chat
1:40:32 gpt4 and it's going to be included in
1:40:34 your 20 bucks I think I assume
1:40:37 um
1:40:39 and it's probably going to be out in the
1:40:40 next two weeks so yes you could go learn
1:40:43 mid journey and listen quite frankly
1:40:45 mid-journey is probably still going to
1:40:48 be a significant player for some amount
1:40:50 of time but
1:40:51 Dolly three can do text Dolly 3 can do
1:40:55 incredibly coherent prompt to image uh
1:41:01 matching so so like if you put a bunch
1:41:03 of details in your prompt they actually
1:41:05 all show up in the image that's one of
1:41:07 the things that mid Journey struggles
1:41:09 with like if you put in
1:41:11 um a bunny and a turtle in a in a pond
1:41:16 setting with a beaver what you end up
1:41:19 with very often is like a bunny Turtle
1:41:22 where they've melded into the same kind
1:41:24 of animal and there's like a giant
1:41:26 beaver like smacking him with his tail
1:41:28 like the the sort of predictability of
1:41:31 mid-journey is challenging at best but
1:41:34 the quality is really good
1:41:36 but I would say just wait for two weeks
1:41:39 for Dolly three and start playing with
1:41:41 it because I think it's gonna be better
1:41:43 so
1:41:45 um I the the other thing about mid
1:41:47 Journey there's a couple of things about
1:41:48 it the CEO is a bit of a like an nft bro
1:41:51 he's a bit of a bit of a douche any and
1:41:54 mid-journey is very much about his whims
1:41:57 from what I can gather like the cus you
1:41:59 know the the customer base wants these
1:42:01 things he's like I don't like those
1:42:03 things and he's so he's just like he's
1:42:05 just like a flipping you know sort of
1:42:07 douchey little CEO
1:42:09 um and I don't think he's taking what
1:42:11 he's got on his hands all that seriously
1:42:12 that said they've got like 200 million
1:42:15 dollars in Revenue so so he's doing
1:42:18 something right like you know I don't
1:42:19 fault the guy he's got the the best tool
1:42:22 out there right now but I think just
1:42:24 wait for Dolly three and it's gonna be a
1:42:26 much more professional tight
1:42:29 um tool and and
1:42:32 the way they've coded it chat gpt4 will
1:42:36 write the prompts for you so you can
1:42:38 just say to chatgpt4 help me write a
1:42:41 prompt to make this kind of image really
1:42:43 good it will write its own prompt and
1:42:45 then generate the image so I think it's
1:42:47 going to be here
1:42:50 so but if you want to learn mid-journey
1:42:52 um I'm happy to do a little walk through
1:42:54 about how you get there it's it's a pain
1:42:56 in the ass but it's it's fine
1:42:59 it's the only reason I know how to use
1:43:01 Discord is I did like a year and a half
1:43:03 sort of deep diving into blockchain and
1:43:05 nft land with all the crypto Bros
1:43:08 um so I got I got decent at Discord then
1:43:12 so anyway
1:43:14 all right how to keep from breaking AI
1:43:17 from breaking into everything social
1:43:18 media bank accounts power grids
1:43:21 um
1:43:24 generic
1:43:26 um
1:43:27 I don't know
1:43:29 I don't know I mean this is this is one
1:43:31 of the risks right so if you wanna if if
1:43:34 you want trouble sleeping tonight go
1:43:36 look at a site called worm GPT
1:43:40 it is a it's based I I assume it's based
1:43:44 on one of the Llama models from meta is
1:43:47 my guess but it could be something else
1:43:48 but it's an open source it's based on
1:43:51 open source large language models it is
1:43:54 designed for hackers for a hundred bucks
1:43:56 a month to be an autonomous agent that
1:43:59 just goes and does evil [ __ ]
1:44:01 foreign
1:44:02 that's out there
1:44:06 um why am I laughing I don't know
1:44:07 because I'm nervous
1:44:10 um
1:44:12 they're going to be lots of tools like
1:44:14 that there's going to be lots of Bad
1:44:15 actors the good news is
1:44:17 the white hat hackers the the good guys
1:44:20 also have access to these tools so if
1:44:23 someone does something pretty egregious
1:44:25 you know the good guys are going to have
1:44:27 access to tools that can either stop it
1:44:29 prevent it fix it or go find the bad
1:44:32 guys and
1:44:33 do evil [ __ ] to them too so I don't know
1:44:36 um
1:44:39 you know I I think this is one of the
1:44:41 reasons it seems to me that that the the
1:44:44 Primitive nature you know we're talking
1:44:47 about these tools being in in DOS mode
1:44:49 right they're kind of in this weird
1:44:51 primitive state
1:44:53 and as powerful as chat gpt4 is
1:44:57 it's not AGI it's not doing reasoning
1:45:00 it's not it's not doing autonomous
1:45:03 actions right it's it's not you know
1:45:05 performing things on our behalf there
1:45:07 there are things
1:45:08 like
1:45:09 um you know Lang chain and baby AGI and
1:45:12 auto GPT that are autonomous agents but
1:45:16 they're also janky and [ __ ] up they
1:45:17 just don't work most of the time
1:45:20 it's the next round of these things it's
1:45:24 it's the GPT fives and and you know if
1:45:27 Gemini isn't so if let's say GPT three
1:45:31 three and three point five are down here
1:45:32 four is here
1:45:34 Gemini this new Google thing is either
1:45:36 going to be you know at gpt4 maybe it's
1:45:39 going to be up here
1:45:41 when when we get to kind of gpt5 when we
1:45:44 get to this level this this sort of next
1:45:46 round or gpt6
1:45:48 that's when these tools start getting
1:45:50 really really powerful and what what the
1:45:54 if if you want to feel a little bit
1:45:57 better
1:45:58 um a group of the big companies that
1:46:01 that have have access to all these GPT
1:46:03 used to do what What's called the
1:46:05 frontier models they put together this
1:46:07 Consortium called the Frontier Model I
1:46:10 forget what they're called it's not the
1:46:11 Consortium what is it
1:46:14 um AI Frontier Point
1:46:18 um
1:46:23 forum
1:46:24 the Frontier Model forum
1:46:27 so you can go read on open ai's blog
1:46:30 about the Frontier Model Forum so a
1:46:32 bunch of the big companies basically got
1:46:34 together and said when we start getting
1:46:36 to the next generation of these things
1:46:39 we need to coordinate and figure out
1:46:41 what if that needs to be actually not
1:46:44 just regulated by government but but
1:46:46 licensed
1:46:47 by government
1:46:49 similar to how you like you give a
1:46:51 license for someone to operate a nuclear
1:46:53 plant or you give someone a license to
1:46:55 operate you know a power plant
1:46:58 some of these models are going to be
1:47:00 powerful enough that the companies
1:47:03 making them are going to the government
1:47:05 to say you should license these things
1:47:07 now there's a cynical side of me that
1:47:10 says
1:47:11 you know if you're early enough you get
1:47:13 the government to you know regulate
1:47:16 things in your direction right so so
1:47:19 they might be doing that but
1:47:21 um they're they're at least talking
1:47:22 about it they're at least saying when
1:47:24 these things get more powerful the
1:47:25 danger is very real and they should be
1:47:27 treated like that just like we treat you
1:47:30 know licensing pilots and airplanes and
1:47:33 airplane manufacturers and all that sort
1:47:35 of stuff when when a lot of lives are at
1:47:38 risk so
1:47:39 I think the current tools are primitive
1:47:42 enough that maybe they're not you know
1:47:44 super a super threat but you know
1:47:48 a bad actor with not much knowledge
1:47:50 could probably do some damage so
1:47:53 sleep well sleep tight
1:48:00 give that garbage which garbage The
1:48:03 Prompt that I gave to make it give that
1:48:05 garbage I don't know what garbage you're
1:48:06 talking about
1:48:15 Becky Roo okay you've tucked me into it
1:48:16 I'm upgrading Bang
1:48:19 another satisfied customer
1:48:22 go to Chachi PT right now sign up at
1:48:25 1995 YouTube can have 1995
1:48:29 1995. yours only buy one get one free
1:48:36 I don't make money from that by the way
1:48:41 oh
1:48:43 but that's cool
1:48:45 okay uh let's see not C I meant C it
1:48:50 keeps popping C for x no it was a step
1:48:53 of an explanation what you you all are
1:48:55 you all are doing wait are you all
1:48:58 helping each other with math equations
1:48:59 in in the comments here
1:49:02 that seems anathema to someone who can't
1:49:05 do math
1:49:07 feels like you're going behind my back
1:49:09 here and solving actual math problems
1:49:11 I mean we got paid in here paint can do
1:49:14 actual math problems so
1:49:17 um I just don't understand how they are
1:49:19 so afraid of these things
1:49:21 and they fail to regulate guns well
1:49:24 listen I
1:49:27 have you watched any of the 48 PM shut
1:49:30 up
1:49:32 have you watched any of the the
1:49:34 Congressional hearings they don't know
1:49:36 what [ __ ] Wi-Fi is
1:49:38 I mean they at least know what a gun is
1:49:40 but they don't know what Wi-Fi is how
1:49:42 are they gonna how are they gonna
1:49:43 regulate something they don't have any
1:49:47 concept they don't understand how social
1:49:48 media works
1:49:50 much less the nuances of a large
1:49:52 language model
1:49:54 and how
1:49:56 how it's generating content and not
1:49:59 copying and pasting it like I don't know
1:50:02 I don't know I mean what I'm hoping what
1:50:04 I'm hoping they do is that out of the
1:50:06 out of the Chuck Schumer meeting you
1:50:08 know where they all met I you know I'm
1:50:09 hoping that they create the AI
1:50:11 equivalent of like the FTC or the FAA
1:50:15 where they put someone like Jeffrey
1:50:17 Hinton in charge of it that actually
1:50:19 knows what the [ __ ] he's doing
1:50:20 understands what the risks are
1:50:23 and and they have an organization that
1:50:25 they can that the Congress can you know
1:50:28 look to and say provide us guidance
1:50:31 based on reality not you know politics
1:50:36 so
1:50:38 I don't know I don't have I don't I I
1:50:41 have I am in a in it
1:50:44 I am in a dark and cynical place about
1:50:46 our state of politics over the past
1:50:48 decade so
1:50:50 I don't have a lot of faith in them and
1:50:52 at the same time
1:50:54 um this shit's really powerful like this
1:50:56 [ __ ] is not a it's not going away it's
1:50:58 not getting uninvented
1:51:00 and
1:51:01 um it needs to be dealt with it needs to
1:51:03 be dealt with at a governmental level
1:51:05 and it quite frankly needs to be dealt
1:51:07 with at a global level right this this
1:51:09 isn't even country by country at this
1:51:11 point
1:51:12 training AI to discern between
1:51:14 fraudulent data and factual data is the
1:51:16 challenge now yeah I think that's right
1:51:19 and I think that um I I think AI will
1:51:23 likely drive us toward back toward you
1:51:26 know tokenization blockchain technology
1:51:28 maybe nft technology maybe some other
1:51:30 flavor of that
1:51:33 but you know can you verify that that
1:51:37 reporter wrote that article and that
1:51:39 that article hasn't been [ __ ] with
1:51:42 how do you track all that well Large
1:51:44 Marge Marge you got to see this picture
1:51:47 you got to see this picture
1:51:49 they caught Biden French kissing a
1:51:51 gorilla at the Cleveland Zoo you know he
1:51:53 was there giving that speech in
1:51:54 Cleveland well apparently he's got a
1:51:56 thing for gorillas he went down to the
1:51:58 zoo and they got a picture of him French
1:51:59 kissing Gabriela can you believe that
1:52:06 what are we gonna do with that
1:52:09 are you using a mobile phone or a laptop
1:52:11 camera for this live curious since frame
1:52:15 size is good
1:52:17 um
1:52:19 phone but it's an iPhone 14 pro
1:52:23 so the back camera is a 45 megapixel
1:52:27 camera if it's using the the 1X camera
1:52:31 and then I think it's a decent
1:52:32 front-facing camera
1:52:34 the it's better my my previous phone I
1:52:37 think I had an 11 Pro and the focus this
1:52:40 [ __ ] was really bad with the 11. this
1:52:43 thing if I have my hand in the thing the
1:52:45 focus is bad but if it's just on its own
1:52:47 it's actually really good
1:52:48 so yeah yeah yeah
1:52:52 and another thing yeah
1:52:55 hey Peter yeah we're gonna need you to
1:52:58 go on ahead and coming in on Saturday
1:53:01 yeah oh and another thing Peter yeah
1:53:06 um we're gonna need you to go on ahead
1:53:08 and come on in on Sunday too yeah
1:53:13 yeah okay
1:53:15 I just got an update did you guys
1:53:18 Sherry Banks if I just got an update
1:53:21 I'll
1:53:22 jump through the screen and give you a
1:53:24 smooch on the forehead come on chat be
1:53:27 there baby
1:53:31 get started let's get this party started
1:53:35 get this party started let's get this
1:53:38 bodies started
1:53:40 I like turtles
1:53:44 welcome to chat adg
1:53:46 [Laughter]
1:53:49 anything new no no see I should have a
1:53:53 little picture I should have a little
1:53:55 camera icon in my in my thing let me go
1:53:58 to hang on
1:53:59 I don't think I have it yet I got
1:54:02 nothing get no see them
1:54:07 all right Sherry they call me tater tot
1:54:11 I'll wait until chat gbt has it thanks
1:54:14 they call me tater tot as a Ron White
1:54:16 reference c c yeah exactly I know that
1:54:19 that Ron White reference that's really
1:54:21 good the commentator that's great they
1:54:24 call me Tater that's awesome
1:54:27 um I just don't understand how they're
1:54:29 so afraid of these things oh yeah that
1:54:31 okay uh planes gas plants gas station
1:54:34 sushi in large quantities everything but
1:54:38 guns huh well listen
1:54:41 the
1:54:42 citizens united
1:54:45 there's your answer right like
1:54:47 undue Citizens United and maybe things
1:54:49 will get better everything's bought and
1:54:51 paid for right now so it's none of this
1:54:54 none of this is about the actual
1:54:56 politics this is all about who's bought
1:54:58 and paid for right so that's where we
1:55:01 are right now so it you know that's why
1:55:04 I'm so cynical this is you can this is a
1:55:07 the great mistake of the of the the
1:55:11 non-conservative side of this argument
1:55:14 is that they keep making the mistake
1:55:18 life's tough stuff tougher when you're
1:55:19 stupid
1:55:20 they keep making the mistake of thinking
1:55:22 that logic
1:55:24 will make a difference
1:55:26 it isn't about logic
1:55:30 it's a game it's a game button paid for
1:55:34 independent of good logic and good
1:55:37 intention
1:55:39 that's why I'm so cynical and bitter
1:55:41 bitter you betcha
1:55:47 um training AI to discernment yep we
1:55:50 talked about that don't trust AI in math
1:55:52 although
1:55:54 um code interpreters not bad because
1:55:56 it's writing and executing python code
1:55:58 and and it's better it's still not great
1:56:01 it's still not great
1:56:04 like I'm I've got a
1:56:06 I've got a report that I run for some
1:56:09 client stuff
1:56:11 oh my God it's like
1:56:14 I can get it
1:56:16 to get the numbers right after about
1:56:20 somewhere between 10 and 20 iterations
1:56:24 it's just it's just bad right now
1:56:27 well so so actually you know Mantra you
1:56:31 know your Point's a good one so so if
1:56:32 especially if you're new to this stuff
1:56:35 an important thing to to
1:56:38 understand is use these tools enough so
1:56:42 that you understand what they're good at
1:56:44 and what they're bad at
1:56:45 and where they're bad
1:56:48 use your [ __ ] brain use your critical
1:56:51 thinking
1:56:53 right if you know that it might not be
1:56:55 good at math if you know it hallucinates
1:56:57 if it spits out a bunch of citations for
1:57:00 the article it just wrote for you where
1:57:03 you're like hey this article is good I
1:57:04 wonder you know could you point me to
1:57:07 some some magazine articles that you
1:57:09 wrote that from it will give you the
1:57:11 magazine articles and most of them
1:57:13 probably won't exist
1:57:16 and unless you play with this stuff you
1:57:19 know you run the very real risk of
1:57:21 embarrassing yourself or Worse right
1:57:23 there was a lawyer that wrote a legal
1:57:26 brief handed it to a [ __ ] judge
1:57:30 and the judge says none of this case law
1:57:32 exists what are you doing you can't make
1:57:35 this [ __ ] up
1:57:37 and the lawyer was like uh sir I use
1:57:41 chat jpt uh sorry about that oh yeah
1:57:44 I'll go back and proofread it next time
1:57:47 foreign
1:57:51 Ty that's on shitty user Behavior
1:57:54 understand your tools these are just
1:57:57 tools
1:57:58 uh diag
1:58:01 o at goatori diagulatory diagulatory
1:58:05 diagulatory diagote I couldn't see the
1:58:09 rest of your name so I don't know what
1:58:10 it says but it looked like a cool name I
1:58:12 think this is one I couldn't pronounce
1:58:16 Congressman said can my wife get on my
1:58:19 Wi-Fi exactly exactly
1:58:24 these guys are gonna regulate our AI oh
1:58:28 my God was he a senator it doesn't
1:58:30 matter it's exactly Silver Fox same
1:58:34 exactly
1:58:36 oh they just bought and paid for shills
1:58:38 ugh it's so gross it's so gross you know
1:58:41 what I mean it's so gross
1:58:44 um are you using a mobile phone we did
1:58:46 that one
1:58:47 and some of those hearings they were
1:58:49 saying to regulate the deployment which
1:58:51 I was thinking
1:58:53 well yeah they could regulate here's the
1:58:55 problem so wait oh Diego Torre attorneys
1:59:01 um
1:59:01 so here's the here's the problem
1:59:06 it's already rolled out
1:59:08 [Laughter]
1:59:11 the the
1:59:13 what's Sam
1:59:15 the reason Sam Altman took open AI from
1:59:18 a non-frop profit to a for-profit was
1:59:21 first he realized he needed more money
1:59:23 than he could raise in a non-profit to
1:59:25 train these things
1:59:27 and then once they trained them he was
1:59:29 like oh [ __ ] we can't just let this out
1:59:31 in the world like was our intent right
1:59:33 so his idea was to slow roll the
1:59:36 deployment
1:59:37 of the base models
1:59:39 and then Chachi PT did what it did
1:59:42 and Google got all excited and they
1:59:44 started doing their thing but they're
1:59:45 all doing it within these closed you
1:59:48 know environments and then meta is like
1:59:50 Hey we're irrelevant over here we've
1:59:53 been talking about the metaverse for
1:59:54 three years no one seems to give a [ __ ]
1:59:56 they're like hey look over here we've
1:59:58 got AI too and they released it
2:00:02 they put it out in the world
2:00:04 so
2:00:06 so it's out there so like that ship has
2:00:09 already sailed now the next more
2:00:13 powerful versions of these maybe there's
2:00:16 some sort of government regulation that
2:00:18 says meta can't do that anymore but now
2:00:22 that the stuff's out there
2:00:24 I'm sure the developer Community is
2:00:26 trying to figure out how they can you
2:00:28 know do virtual GPU server Farms to
2:00:32 you know do their version of gpt5 so I
2:00:35 don't know we'll see
2:00:37 we'll see but this stuff's moving way
2:00:38 too fast and I I think the uh
2:00:41 I think the deployment ship has has done
2:00:45 deployed
2:00:47 government government missed that one
2:00:50 and they're gonna you know they're gonna
2:00:51 try their damnedest to put it back in
2:00:53 the box but they're not going to be able
2:00:54 to
2:00:56 go to chat GPT now please I did and it
2:01:00 didn't like me let me go try to reload
2:01:02 oh no I logged in from scratch it makes
2:01:05 me sad but what did I do Sam Altman
2:01:10 ham
2:01:11 Sam give me my toys I want my chat gbt
2:01:16 to be able to see
2:01:24 it doesn't work
2:01:26 I got nothing I got nothing nothing I
2:01:30 got nothing honey nothing honey
2:01:33 will require us to have some
2:01:35 permit to buy gpus they might I mean
2:01:38 above a certain level I mean if you buy
2:01:41 200 gpus I don't think they give a [ __ ]
2:01:44 if you buy 10 000 yeah
2:01:47 I mean inflection
2:01:49 the ones that do PI AI they just raise
2:01:51 1.3 billion dollars I think 950 million
2:01:54 of that is earmarked for 22 000 servers
2:01:57 or 22 000 gpus
2:02:00 so
2:02:02 that might be the kind of thing that a
2:02:04 regulatory body might say okay if you're
2:02:06 going to buy more than x number of GPU
2:02:08 or X number of compute you know X number
2:02:10 of cycles of compute then you have to be
2:02:14 regulated and if you're going to buy you
2:02:16 know 10x that then you have to be
2:02:17 licensed I mean I I don't know like I
2:02:20 don't know enough about that world at
2:02:21 all
2:02:22 it's the wild west and the laws playing
2:02:24 catch up it's totally that
2:02:27 um
2:02:28 I the the thing that's the thing that's
2:02:31 a little concerning though is I think
2:02:33 that like Corporate America
2:02:36 government America thinks they have more
2:02:39 time than they have I think that they
2:02:41 think they have well in three to five
2:02:43 years they like like a similar sort of
2:02:46 ramp up of the World Wide Web right the
2:02:48 world wide web from from when it was
2:02:50 kicked off to when it hit 100 million
2:02:52 users was six years so that ramp up was
2:02:55 six years chat GPT went from zero to 100
2:02:59 million users in six weeks
2:03:02 six weeks and this the the technology
2:03:06 here is not just the World Wide Web like
2:03:09 this simple point-and-click interface on
2:03:11 the internet no no the output here is it
2:03:14 writes code it writes words it makes
2:03:16 pictures it makes music
2:03:19 and knowledge workers are
2:03:22 about to be
2:03:25 shaken up
2:03:28 hard
2:03:29 and not like three years from now like
2:03:32 2024 is going to be insane
2:03:36 and so I think any company right now and
2:03:40 any government entity that thinks that
2:03:42 they have three years to prepare for
2:03:45 thinking about when to start
2:03:48 regulation is going to be smacked upside
2:03:53 the head
2:03:54 look at my hair
2:03:57 that's what the Senators and congressmen
2:03:59 are going to look like a year from now
2:04:01 when they realize uh we [ __ ] up what's
2:04:05 up there's that one Tick Tock sound and
2:04:07 that's what he knew he [ __ ] up
2:04:10 that's gonna be us
2:04:12 when is it going to be September
2:04:15 except this time next year
2:04:19 this time next year I think we're going
2:04:21 to realize
2:04:23 we need to get some sort of tokenization
2:04:25 and
2:04:27 um media authentication in place
2:04:31 personal verification and identity in
2:04:33 place
2:04:34 and we got to start getting our [ __ ]
2:04:37 together because
2:04:39 I think the 2024 election is going to be
2:04:42 is going to be
2:04:44 a special mess
2:04:46 so we'll see maybe not I don't know
2:04:49 how cautious wait oh sorry what was that
2:04:51 whoever pinned that could you pin it
2:04:52 again
2:04:56 my hair is gone it's gone all sorts of
2:04:59 [ __ ] wacky hasn't it
2:05:03 this is the best toupee I've had
2:05:07 it's actually three topaz but uh wait
2:05:09 how cautious about this information
2:05:11 being
2:05:12 farmed and analyzed in China are you I
2:05:15 don't really for me personally I don't
2:05:17 give a [ __ ]
2:05:20 um
2:05:22 I don't think China's our Nemesis I mean
2:05:25 Google and Facebook are
2:05:29 that we don't need it
2:05:32 okay so China's gonna Farm it all right
2:05:36 it's been it's been being farmed here
2:05:39 voluntarily on our parts for 20 years
2:05:41 now it's the ship has sailed our day the
2:05:45 date of privacy thing isn't done like
2:05:48 people talk about this AI [ __ ] like oh
2:05:50 my God the Privacy privacy is going to
2:05:52 be a real situation what do we do
2:05:54 what do we do
2:05:56 how long have you had a Facebook account
2:05:58 oh for 20 years I put everything on
2:06:00 there really
2:06:01 you concerned about AI
2:06:06 but Mark Zuckerberg he seems harmless
2:06:08 enough
2:06:10 Google their original saying was first
2:06:13 do no evil I'm pretty sure they haven't
2:06:14 done evil I'm pretty sure they haven't
2:06:16 I'm pretty sure they have it because
2:06:17 they said it
2:06:18 they said it I saw it somewhere it's not
2:06:21 on their website anymore but I remember
2:06:22 hearing that from someone first do no
2:06:24 evil so they're not evil
2:06:27 so yeah is it a risk sure is it a risk
2:06:30 from Russia sure
2:06:32 the thing that scares me about the AI
2:06:34 stuff is not so much
2:06:36 China's gonna do this Russia is going to
2:06:38 do this whoever's going to do this it's
2:06:40 that
2:06:42 I think we
2:06:44 mostly rightly
2:06:48 assumed that
2:06:51 if someone was really really resourceful
2:06:54 and had a lot of resources
2:06:57 and really wanted to get to our data
2:06:59 they could get to it but for the most
2:07:01 part our data was safe
2:07:03 and the problem with AI is is that it's
2:07:07 democratizing sophisticated
2:07:10 programming
2:07:13 to non-programmers
2:07:16 so the implication of that is Hackers
2:07:19 going to hack but now you've got
2:07:21 clueless hackers are going to start
2:07:23 hacking well
2:07:25 and that's that's where it gets a little
2:07:27 scary because it because a minor player
2:07:30 could do some major damage they can't
2:07:31 even pass a simple budget extension yeah
2:07:34 I just I I'm I'm just done the whole
2:07:37 it's it's it is comically tragic it is
2:07:41 our government right now is
2:07:43 Shakespearean and it's it is idiocy it
2:07:46 is it is
2:07:49 it is
2:07:51 tragic comedy
2:07:54 it would be [ __ ] hilarious if it
2:07:56 wasn't so [ __ ] destructive I made a
2:07:58 pobot that threatened to delete it if it
2:08:01 gives me false information that's the
2:08:04 Dan that's the Dan model Mantra is a
2:08:07 hacker from way back I can tell
2:08:09 Manchester is [ __ ] with PO
2:08:13 is it working is it giving you answers
2:08:15 that you want because have you gotten
2:08:17 around the safety guard rails
2:08:21 I like that uh we had script kitties for
2:08:24 eons
2:08:26 his videos are already being used
2:08:28 without his consent who's that me
2:08:31 the bot was scared to death so it did
2:08:33 not return any information at all
2:08:36 that's [ __ ] crazy
2:08:39 um
2:08:41 I mean these videos right wait any
2:08:44 information you share here can be used
2:08:47 without your consent oh do I care about
2:08:49 that no I've been listening
2:08:53 I've been I've been doxxed
2:08:56 not not fully docs I have been I've been
2:09:00 just kind of openly on the internet for
2:09:03 you know
2:09:04 many many decades
2:09:07 so I don't I don't really have anything
2:09:09 to hide I don't I don't there's there's
2:09:11 not really much
2:09:13 at risk
2:09:15 is it possible that someone takes these
2:09:17 videos and synthesizes me well yeah they
2:09:20 could do that but they only need like
2:09:22 three seconds of my voice and probably
2:09:25 10 seconds of video
2:09:28 and maybe if they wanted to do a
2:09:29 sophisticated version of this they could
2:09:31 take three of my lives I mean these are
2:09:33 two and a half or three hour sessions
2:09:35 I'm putting on YouTube
2:09:37 um
2:09:42 I mean
2:09:46 you know Joe Rogan and Jordan Peterson
2:09:48 have more to worry about than I do right
2:09:50 like those guys are being spoofed a lot
2:09:53 already right but they've got millions
2:09:55 of followers I mean I just have a
2:09:58 relatively small following I mean
2:10:00 there's I like I don't there's no real
2:10:02 risk right
2:10:05 is it possible that someone goes oh
2:10:09 Kyle looks just like a cross between Tom
2:10:12 Arnold and Rosie O'Donnell and he's just
2:10:14 what we're looking for in a movie but we
2:10:16 don't actually want to deal with the
2:10:17 human being so we're just going to
2:10:18 synthesize him and make him start in
2:10:20 this movie is that a possibility I
2:10:23 suppose so but I think the risk is
2:10:25 really really really really really
2:10:26 really low
2:10:28 so so if someone wants to put me in a
2:10:31 movie because I have special attributes
2:10:33 they can just [ __ ] call me or they
2:10:36 can steal it and then you know I don't
2:10:39 know I'll call up my sag buddies and go
2:10:41 hey what do I do here they'll be like
2:10:43 pay your dudes [ __ ]
2:10:46 emotional based hackers
2:10:49 yeah there's gonna be all sorts of crazy
2:10:51 weird you know what's going to get weird
2:10:53 is the social engineering stuff have you
2:10:55 heard some of the some of the
2:10:56 conversations with these sales Bots this
2:10:59 shit's trippy go to
2:11:01 um if you want to see some weird stuff
2:11:03 again these right now are kind of like
2:11:05 the Dos version of these right so these
2:11:07 are the shitty versions of these go to
2:11:09 air.ai
2:11:11 and just listen to the the initial
2:11:14 um there's like an example
2:11:17 demo
2:11:19 hey James
2:11:23 yeah no worries I hate calls out of the
2:11:25 blue too but I'm actually from Apple and
2:11:28 I saw you were checking out Vision Pro
2:11:29 and we're about to pre-order but left
2:11:31 the page I was curious what had you
2:11:33 looking into potential okay I can
2:11:35 respect that can I just ask one question
2:11:37 before you go
2:11:39 uh fine shoot what's up
2:11:43 why did you well why did you decide not
2:11:46 to pre-order The Vision Pro so so those
2:11:49 kind of conversations like like I
2:11:52 when when
2:11:54 the ability to have a synthesized bot
2:11:57 call up and have a natural conversation
2:11:58 when that when that latency goes away
2:12:00 and the robotic nature of that goes away
2:12:02 which again
2:12:04 maybe a year
2:12:06 those Bots are going to be real time
2:12:08 with zero latency and they're going to
2:12:10 sound human and they can just dial the
2:12:12 phone 24 7 to all phone numbers
2:12:17 so the amount of social engineering that
2:12:20 can happen so that the hacking isn't
2:12:22 even necessarily going to happen at the
2:12:25 um computer like breaking into computer
2:12:27 systems and still stealing [ __ ] that's
2:12:29 one risk but security firms are getting
2:12:32 better and better at that the weak Link
2:12:34 in all of this is us it's always been us
2:12:38 it'll always be us
2:12:40 and so you're already hearing stories of
2:12:43 you know the daughter calls the mother
2:12:45 and says Mom I'm in trouble you got to
2:12:47 send me two thousand dollars right now
2:12:48 and it wasn't the daughter
2:12:52 question from Beach Palm do you put all
2:12:54 the live streams on YouTube all of them
2:12:56 question mark yeah I'm I'm up to
2:12:59 September 15th I think right now
2:13:02 so I'll I'll probably get completely
2:13:05 caught up
2:13:06 um Saturday is my hope
2:13:10 so I started doing these live streams
2:13:13 April
2:13:15 so there's May June July August half of
2:13:20 September
2:13:22 and you'll notice if you go to the
2:13:24 YouTube channel so the um
2:13:29 hang on I'll show you the address in a
2:13:31 second here I gotta sign in and then
2:13:34 switch accounts and then you need any
2:13:39 by the way if you want to feel stupid
2:13:44 um and really learn a lot about where
2:13:46 this shit's headed go watch a YouTuber
2:13:49 named David Shapiro
2:13:52 um he's a brilliant dude so let's see
2:13:54 videos so if you go to
2:13:59 um
2:14:04 uh AI learning lab Dash TT on YouTube
2:14:09 so AI learning lab was taken someone
2:14:11 else has that but AI learning lab Dash
2:14:13 TT for tick tock huh clever like a
2:14:16 [ __ ] steel trap hmm
2:14:19 go there you can see all the all the
2:14:22 back
2:14:23 um videos
2:14:24 but you'll notice the first seven seven
2:14:27 or so don't have descriptions but all
2:14:29 the descriptions
2:14:31 were written by Claude
2:14:34 so I took I took the two and a half or
2:14:36 three hour transcripts of each live
2:14:39 pasted it into Claude and I've got this
2:14:41 magical prompt I wrote like
2:14:44 I've probably put in I don't know a
2:14:47 week's worth of work getting this
2:14:48 [ __ ] prompt to work
2:14:50 and it works about 25 percent of the
2:14:52 time but all of the descriptions of the
2:14:55 videos and all of the chapterization is
2:14:58 all done with Claude it's pretty cool
2:15:01 so
2:15:02 yeah
2:15:04 so all of these are there I don't know
2:15:06 why anyone would want to watch them but
2:15:08 knock yourself out I don't even know why
2:15:11 I'm putting you know I'm putting them on
2:15:13 on YouTube because someone said oh could
2:15:15 could you put this [ __ ] on YouTube I was
2:15:17 like all right
2:15:18 I don't my plan for this Channel right
2:15:22 now is I don't have an end game
2:15:25 um I made a commitment to myself that
2:15:27 I'm going to do this so so just to give
2:15:29 you a sense of
2:15:30 when I decide to do [ __ ] like this I do
2:15:33 it like when I decide to start a company
2:15:34 I [ __ ] do it when I decide to
2:15:36 whatever I do it so when we went into
2:15:39 lockdown I've got my company storyline
2:15:42 we got a video storytelling platform
2:15:44 and
2:15:46 we went into lockdown and I'm like I'm
2:15:48 an extrovert I like I get my energy from
2:15:50 people this is going to be [ __ ]
2:15:52 torture so I decided to do a daily video
2:15:55 diary so if you go to YouTube uh for
2:15:59 storyvine you'll see there's a whole
2:16:01 video series there it's it's a video for
2:16:04 me every day for a year talking about
2:16:07 being an extrovert it's called an
2:16:09 extrovert's View From Within
2:16:12 and I did it for a year straight and got
2:16:15 like 65 followers so The Tick Tock
2:16:19 things has been very different like this
2:16:21 is actually interesting and engaging and
2:16:23 there's a really cool Community forming
2:16:25 here
2:16:26 um that didn't happen with with that
2:16:27 other project but but that's my
2:16:30 my intent right now is just to keep
2:16:33 doing these why I don't know
2:16:37 I think it's important I think it's
2:16:39 valuable it seems to be
2:16:41 helping some people
2:16:45 um
2:16:48 and I don't know I think it's I think
2:16:51 it's I think it's going to be something
2:16:52 of value I don't I don't know what I'm
2:16:55 doing with it though so do you have any
2:16:57 ideas let me know
2:16:59 uh you mean someone can download your
2:17:02 live as it's happening or showing I
2:17:04 don't oh they might be able to I don't
2:17:06 think so they can record you can record
2:17:08 your phone
2:17:10 actually I think I think some of the
2:17:12 Irregulars on here say that they that
2:17:14 they do that that they record these or
2:17:16 they record sections of them like when
2:17:17 I'm talking when I'm doing tutorials and
2:17:19 [ __ ]
2:17:21 um people people have done that
2:17:23 Jim Ross I've been playing with air.ai
2:17:25 to make sales calls for me the woman's
2:17:27 voice is much better oh that's cool so
2:17:29 Jim have they been have they been
2:17:30 working I mean like how does that work
2:17:33 do they get through like when Once when
2:17:35 when like one book's a meeting like do
2:17:38 you get to listen to the interaction do
2:17:39 they does it summarize it I've been
2:17:42 curious about that
2:17:45 that's really fascinating
2:17:48 it actually it might be cool to have you
2:17:51 like share that at a salon or something
2:17:53 like that Jim
2:17:55 Poe showed six criteria for trustworthy
2:17:57 news and said no news on the internet
2:18:00 matches all the criteria
2:18:02 that doesn't surprise me
2:18:04 I think I think it would be a fun thing
2:18:06 I think you could use AI to do I'm sure
2:18:08 someone's already done this
2:18:10 you could find the original the original
2:18:14 um event that happened and then you
2:18:17 could compare different news stories
2:18:19 coverage of it and see how far it
2:18:22 deviates from the facts of the event
2:18:25 oh
2:18:37 and offering boatloads of cash to
2:18:39 influencers for a tech that no one asked
2:18:42 for smells real fishy who's offering
2:18:45 boatloads of cash to influencers to me
2:18:47 wait
2:18:49 there's someone offering me buttloads of
2:18:51 cash I'll sell out I'll do it I'm in
2:18:56 I'm in send me the butt loads of cash
2:18:58 Kevin
2:19:01 what the [ __ ] are you talking about
2:19:03 all right
2:19:04 um
2:19:05 what would be a good model for creating
2:19:07 non-playable characters oh
2:19:10 uh Diego Torres Diego Torres you need to
2:19:14 see something listen
2:19:17 uh uh let's see what's it called
2:19:21 what the [ __ ] is it called hang on
2:19:26 um
2:19:27 oh I can't remember the name of the
2:19:29 paper
2:19:30 someone who's got to remember the name
2:19:32 of the paper
2:19:33 wait uh
2:19:36 let's see wait a i NPCs white paper
2:19:45 yeah I got it okay
2:19:53 so what's it called
2:19:59 this thing This Is Amazing by the way
2:20:02 so this white paper I gotta go find the
2:20:05 white paper hang on I'll do it
2:20:08 where's my mouse come on there we go
2:20:22 fantaverse this is it does it
2:20:34 no that's all right
2:21:04 I'm trying to get you the name of the
2:21:06 paper
2:21:21 so this was from back in April
2:21:24 so so basically what it is this paper
2:21:30 the paper oh generative agents that's
2:21:32 what it's called Interactive
2:21:35 simulacra of human behavior that's right
2:21:37 hang on
2:21:39 um
2:21:41 archive
2:21:45 um generative
2:21:48 agents
2:21:54 so if you go to Archive which is
2:21:56 oh wait that's archivevanity.com that's
2:21:59 not ah
2:22:01 [ __ ] Google and Bing the the search
2:22:04 engines are so [ __ ] useless
2:22:09 a r x i v
2:22:13 dot org
2:22:17 um
2:22:19 generative agents
2:22:25 [Music]
2:22:33 alternative agents Sim
2:22:37 you like
2:22:44 here we go
2:22:46 so go to Archive arxiv.org
2:22:51 and then find this paper
2:22:54 called generative agents Interactive
2:22:57 simulacra
2:22:58 of human behavior
2:23:01 and then what they did this is it's so
2:23:04 cool
2:23:05 um and I and I think there's like this
2:23:08 was the original paper I think there's a
2:23:10 there's a
2:23:11 um
2:23:12 a project on hugging face right now or
2:23:16 GitHub I forget which one where you can
2:23:18 actually go play with this thing now
2:23:22 let me load the white paper in here
2:23:30 Sherry Banks your lives keep me grounded
2:23:32 actually oh thank you for that I
2:23:34 appreciate that
2:23:36 I try to keep it real people I try to
2:23:39 keep it you know at least entertaining
2:23:41 semi semi-educational and then yes semi
2:23:45 like like what the [ __ ] are we all doing
2:23:47 here okay so the way this thing works is
2:23:51 there's 25 NPCs so NPCs are non-player
2:23:55 characters for those of you not in the
2:23:57 gaming world and they're those things
2:23:59 that like when you walk into the bar
2:24:00 he's like would you like a drink and
2:24:02 they have like their script of [ __ ] they
2:24:04 can say and they you know you ask them
2:24:06 things and if you try to have an
2:24:07 interaction with them you can't because
2:24:09 they've been just programmed it's kind
2:24:11 of like the voice trees when you call
2:24:14 the insurance company where they've only
2:24:16 been trained to say certain things
2:24:18 so what they did here was they made each
2:24:22 of the NPCs
2:24:24 have its own Persona and its own job and
2:24:27 then it's tied to chat GPT and the NPCs
2:24:31 can interact with each other so they
2:24:32 wake up in the morning like the artist
2:24:34 wakes up in the morning goes to the cafe
2:24:36 gets a coffee comes back to the studio
2:24:39 and starts painting the farmer makes
2:24:42 eggs with the wife and then goes out and
2:24:44 starts farming
2:24:46 um and then when these NPCs interact
2:24:48 with one another they have conversations
2:24:51 and they remember what they talked about
2:24:53 so one of the examples they give in this
2:24:56 paper is that is that um one of the NPCs
2:25:01 said to the other one it was his
2:25:03 birthday next week
2:25:04 and so the other one planned a party
2:25:08 with the other NPCs so he got together
2:25:11 they planned a party for this this other
2:25:14 one
2:25:15 um and it remembers it and and it's and
2:25:17 it's so so these things are like living
2:25:20 in this world
2:25:22 independent of of you or whatever other
2:25:25 players going through it and they're
2:25:27 developing memories and they're
2:25:29 developing you know personalities and
2:25:32 they're getting work done
2:25:35 what's fascinating about this to me is
2:25:37 that
2:25:39 the distance between this being a game
2:25:42 activity and being a personal agent for
2:25:46 for yourself is very small
2:25:49 so so imagine this rather than it being
2:25:51 a game like this imagine that you've got
2:25:54 25 agents that represent you
2:25:58 you've got a little entertainment agent
2:26:00 whose job it is to go out and find
2:26:01 interesting stuff for you to do and
2:26:03 you've got a work agent that's job is to
2:26:05 go out and research things for work so
2:26:08 that you you know rather than reading
2:26:09 the morning paper it knows what you like
2:26:11 and and you know on your way to work it
2:26:14 it calls you and you have a conversation
2:26:17 about the current events of the day and
2:26:19 then you have a vacation agent and then
2:26:21 you have an insurance agent you have all
2:26:23 these independent virtual agents
2:26:26 that are are out interacting with other
2:26:29 agents with other systems
2:26:31 and checking back in with you and
2:26:34 they're remembering all of the stuff
2:26:35 that they do so the vacation agents like
2:26:38 hey Kyle it's been it's been six months
2:26:40 since you you guys have had a vacation
2:26:42 you know are you thinking about doing
2:26:44 that anytime soon you're like oh that's
2:26:46 a good idea I forgot to do that you know
2:26:50 um
2:26:50 what do you what do you recommend and
2:26:52 then the agent might say well do you
2:26:54 want to you know do you want to do
2:26:56 something you've done before or do you
2:26:57 want to do something new oh I want to do
2:26:59 something completely new that I've never
2:27:01 thought of and it will know what you've
2:27:04 done right so this paper to me
2:27:07 I think from a gaming standpoint the
2:27:10 games are going to get really
2:27:11 interesting but then from a if you if
2:27:13 you just look at what's the ideas
2:27:15 sitting underneath this paper
2:27:18 holy [ __ ]
2:27:19 holy [ __ ] so anyway that was that was a
2:27:24 long answer to it to a question like
2:27:26 what about NPCs
2:27:28 there you go
2:27:31 yowza all right
2:27:33 archive is pronounced archive
2:27:36 it's the letter chai nice
2:27:40 cool didn't I call it archive
2:27:44 I think I did
2:27:47 looks like Legends of Zelda yeah again
2:27:49 don't mistake the artwork for how big
2:27:54 that idea is
2:27:56 on a Legend of Zelda game engine yeah
2:27:58 exactly exactly
2:28:01 that's where they'll test all this stuff
2:28:03 but I think I'm pretty sure that that
2:28:06 like like a month ago so that paper was
2:28:09 released in April and I think a month
2:28:12 ago the project came out where you can
2:28:14 actually log in and play with that thing
2:28:16 so go go Google the [ __ ] out of that
2:28:20 thing there's probably some fun stuff to
2:28:22 look at at this point
2:28:25 but now chat GPT wait now in chat EBT
2:28:29 have been playing so far useless oh
2:28:30 interesting
2:28:31 useless on the like the image reading
2:28:34 stuff yeah it's gonna be interesting to
2:28:36 see what's the thing I'm excited about
2:28:38 is the image Generation
2:28:41 Um but
2:28:42 we'll see I I think the the
2:28:46 ability for it to see is I I don't think
2:28:50 we really know what to do with that and
2:28:53 and quite frankly Bing's had that
2:28:55 capability for
2:28:58 four or five months now
2:29:00 it's just that it's in Chachi BT now it
2:29:02 seems like it's a big deal what was that
2:29:03 site again the site for the
2:29:06 um
2:29:07 for that paper is archive a r x i v
2:29:13 dot org and then search for generative
2:29:16 agents interactive
2:29:19 simulacra in fact I'll post this in the
2:29:24 um
2:29:25 I'll post this in the Discord
2:29:29 for
2:29:31 um
2:29:32 for the AI salon
2:29:34 in let's see
2:29:37 General discussion
2:29:40 um
2:29:41 this is a cool paper if you haven't
2:29:48 seen
2:29:50 it Bang
2:29:53 and if you're wondering well what's the
2:29:56 salon Discord if you go to if you're
2:29:59 interested in joining a community that
2:30:01 does cool [ __ ]
2:30:04 um the salon.ai is a group that I
2:30:08 started last December
2:30:11 um for people interested in playing
2:30:13 around with AI
2:30:14 and we have a Discord server and we have
2:30:17 bi-weekly meetups so there you go
2:30:22 um
2:30:25 girl wake up because you need to get
2:30:27 perplexity AI for this kind of stuff all
2:30:30 right you're helping one another that's
2:30:32 cool all right listen people
2:30:35 um let me see if there's any other
2:30:36 things here
2:30:38 dead out down what's the idea behind
2:30:41 this paper in your expert summary way
2:30:45 um the the idea is that the NPCs have
2:30:49 their own lives interact with one
2:30:51 another and have memory
2:30:54 and then if you if you attach that model
2:30:57 where you've got these agents that run
2:31:00 around in that case it's a game world
2:31:02 but you could see them sort of
2:31:04 navigating any world or Internet space
2:31:08 if you attach that behavior
2:31:13 if you attach that behavior to actions
2:31:17 so if you have a little agent that can
2:31:19 run around the world
2:31:21 and has access to your bank account and
2:31:23 your travel accounts and things like
2:31:24 that
2:31:25 and you've given it permission to do
2:31:27 that
2:31:28 it can just run around and do [ __ ]
2:31:30 and it can interact with another agent
2:31:32 so for example
2:31:34 in the game
2:31:36 the one dude said you know it's my
2:31:40 birthday and the other one planned a
2:31:41 party for it
2:31:43 in a real world use case you could say
2:31:45 hey you've got your travel agent is
2:31:48 going out and finding you cool vacation
2:31:50 stuff but the travel agent could
2:31:52 actually check in with the work agent
2:31:54 and say hey does Kyle have any important
2:31:57 meetings coming up that he can't miss
2:31:58 and the work agent would go oh yeah the
2:32:01 second week in October he's really busy
2:32:03 so the third week in October looks good
2:32:06 so book something then and then the
2:32:08 traveling so so the idea that these
2:32:13 these independent entities
2:32:16 can interact with one another
2:32:19 talk in natural language have their own
2:32:22 imperatives one's an artist one's a
2:32:24 farmer one's works at the store
2:32:28 and then have memories of what they
2:32:30 talked about from day to day is wholly
2:32:33 new because NPCs up to this point have
2:32:36 been hand written
2:32:38 logic like there's a writer that writes
2:32:42 the script of here's all the [ __ ] this
2:32:44 MPC can say when someone says this you
2:32:47 respond with this when someone says this
2:32:48 you respond with this and it's it's a
2:32:51 fixed set of instructions
2:32:54 and it can only move in certain places
2:32:56 it can't in operate independently on its
2:32:59 own
2:33:00 so
2:33:01 it's just it's just these things are
2:33:04 like
2:33:05 living organisms basically chat GPT on
2:33:08 iOS has voice conversations feature I
2:33:11 signed out and signed back in okay
2:33:13 and it worked okay
2:33:15 thanks retropunk that's awesome I'm
2:33:17 gonna go play with that
2:33:18 um listen it's 10 30. my voice is going
2:33:20 I'm sneezing whatever's happening shit's
2:33:24 going down oh wow that is beautiful
2:33:26 thank you Zippy Zimbo that was awesome
2:33:29 thank you very much
2:33:32 um
2:33:33 I'm out of here
2:33:34 thank you all you all are really cool so
2:33:37 let's see
2:33:38 um Jox has been in here tonight Sherry
2:33:41 Banks thank you for sharing what you're
2:33:42 doing paint similarly thanks for helping
2:33:45 out and sharing your wisdom that it
2:33:47 really does make a difference
2:33:48 uh let's see retropunk good to see you
2:33:51 uh who else we got in here
2:33:54 um beat from Firefly good night I saw
2:33:57 Silver Fox was in here for a while
2:33:59 cabruno thank you all thanks for all of
2:34:01 you who modded tonight I appreciate that
2:34:03 Clint good to see you how to partner how
2:34:06 you doing how you doing oh you're such a
2:34:09 little cutie
2:34:11 mods and Irregulars present Turtles
2:34:14 y'all are great thanks all for your time
2:34:16 I will see you tomorrow hey go play with
2:34:18 this stuff if you've got access to this
2:34:20 go start figuring it out get curious get
2:34:23 curious get curious
2:34:24 peace