aiLL 8/02/23 - I'm Back! The Wild West of AI Explodes While I Was Gone

Transcript

1. 0:00foreign
2. 0:25I'm back I was I was in the doghouse for seven days [Laughter] I'm back [ __ ] hey Robert what's happening David gold
3. 0:42thanks for the world I am back andre44 Andre Andre yeah Andrea Andrea Andrea Andrea thank you I was afraid
4. 0:57you'd been abducted by aliens yeah exactly I got hacked by AI I'm out of jail hey Troy thank you missed you guys too welcome back thank you
5. 1:12I don't even know what to say I haven't done this in so long I'm gonna have to help me through this welcome welcome everybody my name is
6. 1:19Kyle Shannon this is the AI learning lab thanks for doing this stream much appreciated you're welcome I'll try not
7. 1:24to get my ass kicked off this time I was doing pretty good I had like a two-month run going and then yeah I made
8. 1:30a stupid joke and no no I quoted someone making a funny joke and but I think what they said hmm not good so anyway they booted me
9. 1:41whatever can you explain what happened yeah I said the heart r word I I quoted a comedian who said the hard word and I
10. 1:48think me quoting it was classified as hate speech whatever doesn't matter we're back we're back we're live uh if you're new here my name
11. 2:00is Kyle Shannon this is the AI learning lab we learn about Labs here oh and AI uh since I've gone away we've got room temperature uh
12. 2:11superconductivity very controversial might be faked some of the scientists that did it were apparently part of some
13. 2:18faking that happened 20 years ago oh might not happen but you know if we're going to transform the world with AI we might as well have
14. 2:26superconductivity followed by free energy and cures for cancer and all that [ __ ] right Tobias was shaken told you
15. 2:33I'd be back later tonight I don't know how late I can stay up it's a little late already but I came back as soon as I could it was 21.52
16. 2:41mountain time today I wasn't allowed to come back before then uh anyway uh let me get a sip of water here and we'll get jump in
17. 2:56all right let's do this so if you're new to AI it's a generative AI if you're like what is this hey ask stuff everyone's
18. 3:08talking about I don't know these things here are places to get started and you're like but Kyle which one's best
19. 3:19the top one the top one is best in my opinion just because that's the official chat GPT website that's gpt4 it's uh
20. 3:30it's got all sorts of cool [ __ ] on it like custom instructions and code interpreter and plugins and it's just it's just the best now
21. 3:40I don't think it's gonna be the best for long quite honestly because there's a lot of competitors coming so Claude over
22. 3:46there that's ex open AI employees that started a competitor and that thing has a hundred thousand token context limit
23. 3:54so you can put 75 000 words into claude.ai that's a big deal uh Bing is gpt4 connected to the internet it can
24. 4:04also generate images it can also read images um bard.google.com used to suck really bad it's better now it can read images
25. 4:14labs.perplexity.ai that's um perplexity has given us access to the The Meta llama 2 models for free without
26. 4:24having to install [ __ ] and then Pi dot AI is a real chatty chatbot and poe.com lets you do Bots and
27. 4:32things like that so anyway that's where to get started um other things to to pay attention to if you're super brand new are these
28. 4:40things so prompts.chat will teach you about prompting and there are many places you can learn about prompting hey
29. 4:46Silver Fox thank you for the rose that's so sweet um promise.chat will teach you about it it'll teach you this idea of Mr cgo yeah
30. 4:57baby Chief generative officer first in the world uh what the [ __ ] is the chief generative officer exactly uh
31. 5:10openart.ai prompt book that'll teach you about prompting things like mid-journey futurepedia.io that's just a pile of
32. 5:17tools like if you want to go geek out go there just know that 90 of them are pieces of [ __ ] and then you got
33. 5:22mid-journey Adobe Firefly and Leonardo which is built on top of stable diffusion those are kind of the three biggies on the image front
34. 5:30um if you're a geek go play with all the stable diffusion [ __ ] if you're not play with leonardo.ai then you got Runway ml which is videos
35. 5:37and blockade Labs which is just kind of fun thank you for the donut I love me some donuts this is just like one of those uh uh NPC
36. 5:47channels where if if you give me toys I react to them because I'm a trained monkey all right if you have questions
37. 5:55about AI pop those in the comments below uh if you haven't been here before and it's very possible because I've been gone for a week and uh
38. 6:04and uh what's the new 100K Nvidia perfusion model that was just announced I don't know let's go look I saw something cool
39. 6:15I saw something cool from Nvidia today that was not the the profusion model is that the image thing Nvidia it might be
40. 6:25and if it is it's super cool per Fusion and we present perfusion it's lack confusion only Persia yeah this is this looks really cool okay
41. 6:40so so what this is so one of the grand frustrations of large language models is um of the image generation models is that
42. 6:55it's really hard to do predictable stuff and what this perfusion model is from Nvidia is like it's it's actually pretty similar to
43. 7:07um Google announced that thing called the hell is it called not snap something I don't know Google had this image thing
44. 7:16where you could upload one image and it would it would mimic that style this is similar where you upload a character and
45. 7:24then you know you can make different versions of that exact character that's really hard to do right now
46. 7:33um and so like you know you upload that teapot and then you can make a gold teapot or a you know Japanese rose
47. 7:39teapot in a whatever crocheted Garden or teapot with the bear you know in the lake so uh this looks really cool it's
48. 7:49not out yet I just saw this today um this looks like a big deal this was accepted at siggraph um but yeah but anyway there's a white
49. 7:59paper on it and it's called key locked ranked rank one editing for text to image personalization it doesn't matter
50. 8:06nvidia's got some cool [ __ ] coming out so when that comes out image generation is going to get a lot better uh
51. 8:14uh Amelia's wife we need to know have you missed us of course I've missed you who the hell else do I talk to
52. 8:19who well the dog of course you talk to the dog you have to talk to the dog um but yeah I missed you a lot and and I
53. 8:27appreciate a lot of people reached out some were concerned where you been everything okay yeah I was just in Tick
54. 8:33Tock jail just Tick-Tock jail puppies puppies back he's happy he's laying here right beside me he loves hanging out on the lives
55. 8:43all right so I am gonna go find uh the first question and let's get started [Music]
56. 8:59[Music] all right Robert Rossi got the number one gifter
57. 9:14badge wow Robert that's exciting that's exciting I don't know what the hell that means but that's a deal
58. 9:21hello Triumph truth what's happening naughty boy I was a naughty boy I was put in jail can I explain what happened
59. 9:28I said a bad word apparently classified as hate speech if you've been on this channel you know I'm not a hater I'm a lover
60. 9:35I'm not a pushover don't [ __ ] with me but you know I'm a lover all right but I am a little Saucy so if you get a little saucy
61. 9:48there's a chance you're gonna you're gonna get your you're gonna get your little fingers slapped through the ruler so that's what happened
62. 9:59let's get this Armageddon started already oh yeah AI might be an evil thing it might be it is equally likely in my opinion that it will do
63. 10:10uh dramatic good it could also do dramatic bad the reality like always is it's probably somewhere in between there
64. 10:18will be some dramatically good stuff and some really shitty stuff and this channel is about getting curious enough to not be at huh
65. 10:33not be at the mercy of it that's not really it it's just to understand it enough to not be blindsided by it that's
66. 10:40that's really the purpose of the channel um get people to try it get you excited about it if you try it and you hate it
67. 10:47then great at least you tried it but a lot of people right now are afraid of it without trying it or judging it without
68. 10:51trying it or making assumptions about how it works or what it does and that the robots are gonna [ __ ] come get us
69. 10:58no no we're still in charge these are tools these are tools they're tools just like photoshop's a tool right and people
70. 11:05freaked out when Photoshop came out and these are tools just like the calculator was a tool University professors when
71. 11:12calculators came out mathematics professors were on campus with little placards ban the calculator these kids must learn longhand Edition um
72. 11:25same [ __ ] different technology but this technology is kind of you know in the neighborhood of fire printing press steam engine this
73. 11:40you could maybe throw the internet and the World Wide Web in there but I don't think so I I think that's more of a
74. 11:45minor blip this is a major one this is Industrial Revolution scale [ __ ] and what's exciting is we're at the beginning of it even though
75. 11:55machine learning and ai's been around for a long long time this generative AI stuff um has been around since last November
76. 12:03so like nine months into this stuff and the competition is heating up I only put [ __ ] on here that I think is worth
77. 12:10checking out the fact that there's seven things on there that I actually think are worth your time enough to say it's
78. 12:15worth checking them all out is amazing to me um I'd like to get that down to three um but anyway yeah interesting you Americans are obsessed
79. 12:27about drinking water well I live in Denver and it's a high desert so I'm a mile in the air and it's dry all the
80. 12:33time except for this summer when it thunderstorms every afternoon but drinking water in Denver is like a full-time job it's if and if if you
81. 12:44don't do it uh you wake up in the morning or I wake up in the morning with just a raging [ __ ] dehydration headache
82. 12:53and it was and it took me because because I'm a little dim it took me like a year and a half to put together like if I wasn't
83. 13:01[ __ ] pounding water um that uh that that was what was causing the headaches I'm like why do I have all these headaches that's what it
84. 13:09was and so I bought I went on Amazon I got myself 32 ounce tumblers I just fill one of those bad boys up before I go to
85. 13:17bed every night and try to drink it before morning it's crazy and I don't have to pee all night because it's just my body needs it
86. 13:26interesting yeah with iced tea coffee all the time yeah I like ice in it uh welcome back so MP's MPC streaming
87. 13:35didn't work for you that's my other Channel That's where I've been I'm making fifteen hundred dollars a day going what do they say
88. 13:43um thanks for the rose thanks for the rose oh cowboy hat oh cowboy hat I could do that if you want to turn this channel
89. 13:49into that just start gifting me [ __ ] and then we don't we don't have to talk substance at all I don't I don't think I'd last
90. 13:59two days doing that what's this new 100K Nvidia perfusion model that's what I just talked about uh can't take a screenshot due to
91. 14:07privacy I don't know what that means to bias you can't or I can't oh can't take a screenshot of that over
92. 14:14there that might just be a local setting on your on your phone Kyle We need oh Tick Tock just jumped me
93. 14:23to the bottom oh the pet peeve so in the seven days they kicked me off they couldn't have fixed the scroll to the bottom bug no no tick tock
94. 14:31you missed champ too he's right down there aren't you Champion you good boy a good boy he's a good boy all right
95. 14:45welcome back to the AI party are free ways to clone voices without paying the monthly subscription well
96. 14:55yes but so so there's gonna be plenty of them right now the best one is 11 labs and you need to pay money to clone your voice there
97. 15:05um there's probably another one I don't know about 11 Labs is the biggie but in iOS 17 which I'm running right now inaccessibility
98. 15:14if you go to accessibility and speech there's a thing called personal voice and within the IOS operating system you can clone your voice um
99. 15:27they're sneaking it into under accessibility because there's some cool things you can do with that from an accessibility standpoint but
100. 15:34I think that's the foundation for some other stuff now I'm on an iPhone 11. this is a shitty old four-year-old phone or however old it is
101. 15:43and uh I don't have access to that particular feature so I can't tell you if it's any good but I'm pretty sure how it works is
102. 15:50they give you 50 phrases to say so they're basically trying to get you to say all the what are they called phenomes phonemes
103. 15:57the the parts of words they get you to say enough words that you say all the parts of words so they have how you say all your words um
104. 16:06and you record all that [ __ ] and then it makes a voice and then you can send messages with it and things like that so
105. 16:13um it's it's coming it's it's gonna come like like these tools are going to start to be incorporated and all the other
106. 16:19tools we use things like synthesizing your voice synthesizing your face all that's going to come um but I don't know other than the iOS
107. 16:28beta right now I don't know of anything you can do it for free but there are probably some out there in fact there are probably some open
108. 16:36source ones so but I don't know I don't know how to go to sleep without Kyle thank you see okay now that's a
109. 16:44that's a solid that's a solid AI learning lab troll where is it a compliment is it a put down like like he
110. 16:51puts us to sleep huh you see you see this is solid I like it it's good it's comedy it's comedy comedy all right who's a comedian that says
111. 17:02that Mark Norman all right too much twerking we knew this would happen I know I know I tried I resisted the twerking for months and you
112. 17:12all insisted and I said fine I'll twerk and then what happens I Twerk and by the time I turn around to the camera my live has been shut down
113. 17:23oh it wasn't twerking I was just trying to get my hair to do something normal yeah failed all right I tried Gen 2
114. 17:31Runway not impressed at all um try if you haven't tried image to video It's better it's better than so uh so what's your
115. 17:43name Matthias Windsor yeah the the text to video for Gen 2 is wonkier than [ __ ] now if you if you sort of look a year
116. 17:54before images got good they were laughably bad I think the same thing is going to happen with video but in Gen 2 right now
117. 18:03definitely on their website and I think they've added it to their uh their phone app yeah they have you can upload an image and turn that
118. 18:12image into um a video and in fact I'll show you one that I did of a cloud photograph I took it's pretty [ __ ] good dude
119. 18:24Matthias you know deed dude let's go check it out why is he talking like that that's just what he does that's just what he does
120. 18:37he's you know he's okay he's for one of them computer Smart Guys he's all right you know he's all right he's he's a little
121. 18:47he's a little bit of a dum-dum and uh
122. 19:01barberheimer that was a pretty cool clip oh there was there was me with Barbie I thought that was that was good for the
123. 19:07uh for the Barbie movie I I put that I was her uh her uh rebound relationship after Ken Gotta Laugh or two ah here we go check this out so
124. 19:24um so that photograph is a panorama I took with my own camera and did some editing in Lightroom so that's
125. 19:35if you want to see my photography you can go to Kyle shannon.com that's a bunch of my photography so
126. 19:40anyway so I started with that I uploaded that to Gen 2 with no prompt and it turned it into this I mean come on
127. 19:54now it does lose a lot of detail right so it's like it's definitely it's definitely early technology and it's only four seconds
128. 20:03but I mean if you cut that into you know other footage like as as part of an edit that's that's totally believable totally believable
129. 20:18so you know for what it's worth for what it's worth you know you know you put it out there you stick an image up boom
130. 20:26makes a little movie not gonna win any Oscars but you know
131. 20:40did you miss my singing I'm a singer I'm not as good as singers I am a hairdresser what the [ __ ] is happening up there tonight
132. 20:50at least that's I'm sure you've missed the hair all right I think the sound effects really helped I do thank you back again Griswold
133. 21:03Mueller Bueller eleven Labs cost about 20 a month any free Alternatives no that's what I was just saying the uh the
134. 21:12one in apple it's in you can if you download the iOS 17 beta it's in there if you've got a more modern phone than I have
135. 21:20but I just have this janky ass crack screen piece of [ __ ] are you stuck for New York City uh how much will AI be on the agenda well
136. 21:35a lot so we're meeting with one of our Pharma clients and we're talking about three pilot projects they want to do that are
137. 21:44all AI related and then I'm potentially knock on wood fingers crossed say little prayers for me uh I think we're gonna
138. 21:54get a chance to meet with the CMO of a major Pharma company and tell her about what we do and what we're going to do
139. 22:02with AI and why she should be super excited to work with us and I'll go man technology here you want to see it it's pretty cool
140. 22:13she'll be like Ah that's awesome dude but yeah can I buy you lunch she'll be like sure dude pretty sure she's from California you
141. 22:22know got to learn to talk to your audience no she's not she's from New York uh Skynet it's pretty pictures
142. 22:31that's true that was actually that's that's a that was a good example of skagnet painting pretty pictures I tried
143. 22:38to flip my car in the air fast and furious like the only thing it did I had it it added some smoke oh yeah yeah so
144. 22:48so like the images the video stuff like like video stuff's way more complicated because you know what out of whatever
145. 22:56they call it continuity or or it's not it's not continuity or consistency it's coherence whatever it doing [ __ ] that looks good
146. 23:09um it's really hard with images right you see hands that look like this those have gotten better now most people have
147. 23:15five-ish fingers and you know sometimes you'll get hands that are this small or fingers that are like this long but
148. 23:21there's normally five of them now so that's that's getting better but with video it's you know 30 frames a second more complicated
149. 23:31um so I think it I think it'll get better and and honestly the the with the image stuff my experience has been about one in ten
150. 23:41that I do are good the cloud one probably because there's there's so much reference video of slow motion or time
151. 23:49lapse clouds panning across the sky like that one probably has a pretty clear reference that it could find
152. 23:57um and some of the other stuff just just there's probably not enough in the training set or something it'll get better
153. 24:04uh what time is it it's 905 here not a time check and I'm timing to see how far behind you are it is 10 17 right now
154. 24:13so I am uh sorry math 12 minutes behind uh but I assume you're an hour I think I assume you're West Coast
155. 24:23Thelonious so 15 minutes or 12 minutes behind and and if you're new here welcome my name is Kyle Shannon we talk
156. 24:30about AI stuff here and answer questions and we have fun we have fun we fun sweaty balls you know um and uh
157. 24:41I try to read all the comments so if you have questions or thoughts about AI pop them in the comments I try to read them
158. 24:47all I do get behind uh people that are the regulars here they will defend me they will tell you I do answer my
159. 24:54questions if you come in here all truly like like yeah I don't even think he's live He's not what's he reading those
160. 25:01aren't questions that I wrote why isn't answering my questions I've asked them 17 times if you're one of those people man is
161. 25:09this gonna be frustrating oh it's gonna suck oh do I have to sit here and watch the whole thing to wait for him to answer my
162. 25:18stupid [ __ ] question yes yes you do I've been off for seven days so I'm a little wound up all right
163. 25:29what's the process of writing a scripted wait of AI writing scripted or doing a video is that anime person doing lip sync
164. 25:42I don't quite know what the question is well so I'll go I'll go show you how we can make one of those it's pretty fun
165. 25:56um it ain't exactly Hollywood quality yet it's a little all of these Technologies right now are a little as we call in the business janky
166. 26:08so let me show you how we do some scripted to animated voice stuff all right so let's see go over here and where will we go first you know
167. 26:23where we'll go first we'll go to the classic we're going to chat GPT you can make money with chat gbt all right so here in chat GPT and I'll
168. 26:35do GPT 3.5 that's the free one so this just this is just your vanilla flavored chat GPT and I'm gonna say um foreign how
169. 26:52to write good dialogue for a funny Holly would movie and you're like but Kyle why would you ask Chachi PT if it's good at that
170. 27:09because one of the things that happens with good old Chachi BT is the more context you give it the better it gets
171. 27:15so if you just say write me dialogue it'll write you shitty dialogue if you ask it something like you know what's
172. 27:21good dialogue for movies or do you understand that like it said what does it say here um understand your characters timing is
173. 27:30everything use word play and puns use exaggerated give me a bunch of examples so I'll say okay great and here I'm gonna do this with voice
174. 27:41instead of typing because it's annoying to all of you I know hmm [Music] write me a funny opening line of a simpleton who's memorized all of
175. 27:57the books in all of the libraries of all of the world in a world where wait in a world where brains were bites and wisdom was Wi-Fi
176. 28:10I'm the all-knowing simpleton who downloads the entire universe into my Noggin but uh where do I plug in the toaster that's actually not bad
177. 28:20that's not bad all right there's our line pretty funny now it's not going to be delivered with comedy because it's going to be a
178. 28:28[ __ ] robot saying it but okay so you go you if if you can write the words you can write your own words but
179. 28:37we wrote those words and now there's two things I'll show you one's 11 labs and at 11 Labs I have uh trained it to
180. 28:51be able to talk like me sort of you know in in whatever the state of art talking like me sounds like and let's see Rachel KS I'll do chaos
181. 29:04chaos bright so I recorded I recorded two different versions of me I recorded one where I'm talking like a serious
182. 29:10person like this kind of like a newscaster and then I I recorded one where I was I was like I'm in a Pixar
183. 29:15movie isn't this exciting so we'll use that one we'll use the peppy one because it's comedy and then I'm gonna pop that text in
184. 29:23there that I just wrote and we'll say generate in a world where brains were bites and wisdom was Wi-Fi I'm the all-knowing
185. 29:35simpleton who's downloaded the entire universe into my Noggin but uh where do I plug in the toaster not bad not bad so that's that so that's
186. 29:45eleven labs.io is the website that one does cost you money if you want to clone your own voice but you can also go to
187. 29:55dash ID and there's there's many of these there's Synthesia there's a bunch of these these do cost money I think
188. 30:03these things are ridiculously overpriced but They Don't Really require much knowledge so yeah no credit login sign up here we go
189. 30:25personalize your experience okay so here's a bunch of videos that I've made so like this is a a South Asian woman that I made for a
190. 30:36work project the streets of Mumbai are adorned with vibrant colors bustling with excitement so the way it works if
191. 30:43you say I want to create a video you can use one of their avatars or you can use um one of yours so this dude is like a Pixar looking dude
192. 30:55and this I I sort of did this dude to look yeah to look like that bright voice that I just recorded so in this case I'll just pop it over
193. 31:07here oh I can upload oh you know what I can do I can actually upload my voice from 11 Labs so let's grab this I'm going to download this
194. 31:19and then we'll go over to did and then red so so what you can do here is you can pick from um all sorts of different uh languages
195. 31:33and um accents so so in English let's see where's English in English you know you can do Netherlands Australia Canada Hong Kong
196. 31:46India so that that that South Asian one I just did was English comma India or you know parentheses um so so you can just pick a voice and
197. 31:55and generate it based on that or you can upload your own audio so I'm gonna upload the thing that I just created in 11 labs
198. 32:09and then I'm gonna say generate video so it's going to take that thing I made in 11 labs and it's going to put it on that
199. 32:14voice and it's going to animate that picture to look pretty darn swell and so it's off doing that here's another dude I did
200. 32:27hey there it's Kyle Shannon with the AI learning lab and today we're going to transfer that one's bad here's one I did
201. 32:34I think with my voice hello Tick Tock live no how are you that was not my voice one of these has my voice in it this one And So It Begins
202. 32:49there was a time until very recently where you had to live life to experience it that time that was pretty good
203. 32:56all right so here's the one we just did I'll hit the little play button in a world where brains were bites and
204. 33:04wisdom was Wi-Fi I'm the all-knowing simpleton who's downloaded the entire universe into my Noggin but uh where do I plug in the toaster
205. 33:13[Laughter] comedy comedy so anyway that's how you do it did that help did that make things worse
206. 33:28or better did it make things worse or better I don't know he seemed to talk for a long long time did he say anything I don't know
207. 33:35Industrial Revolution 3.0 here we come it's coming it is coming it's absolutely on the way it's absolutely on the way without the
208. 33:44industry all the industry will be replaced with robots but anyway yes the industrial Rebel it's the it's the
209. 33:51knowledge it's the it's the artif it's the robot thanks for the what is that it's a boom box is that a boom box in in
210. 33:56my day back in the olden time of days in the in the 80s and the 90s we had a Boombox you'd walk down the street with
211. 34:03a big thing on your shoulders speakers blasting in your head and you play annoying music and annoy the people around you
212. 34:13now we just get on Tick Tock and talk into our phone I I can annoy you in many different ways I don't need a Boombox I got a
213. 34:22teeny little funk all right do you have an affinity for any reptiles ah I think they're impressive I just saw a
214. 34:38tick tock there's like a snake sort of dangling out of the ceiling and then the entire [ __ ] ceiling collapses and it
215. 34:45was a massive like the tail was just like this teeny part that is massive like 40-foot snake um I like lizards I learned how to catch
216. 34:58lizards in the uh in the desert with a little piece of fish in line with a with a loop on it and you just dangle it in
217. 35:04front of them they kind of freeze like that's interesting and then you just slide it over their head and go a little
218. 35:10slip thing and you catch them so desert lizards are pretty cool they can theorize and ask scenarios based on existing scientific research
219. 35:21it's amazing wait what was that for you can theorize I don't know what that was about who said that glocal Global girl local oh as in
220. 35:34globally local that's a that's a marketing term from like the late 90s glocal that's really funny are you in marketing
221. 35:45you got to be in marketing uh I don't know what you're talking about though it sounds cool um what you're in Denver too I build AI
222. 35:55That's cool you should come to okay so if you're in Denver even if you're not one of the other things that I do
223. 36:02is called the AI salon so if you go to the salon.ai that website it's it's a link tree and we have a Meetup we just
224. 36:10had our meeting last night we meet every two weeks in person in Denver and we do it we have an after party afterwards so we go hang out and
225. 36:19then uh we bring on speakers we do this thing called AI confessionals where people just talk about here's something I did that blew my mind
226. 36:27do something or here's these videos and they sucked right um and we also have a Discord server so there's a the invite to the Discord
227. 36:37server is there as well and uh then we have a LinkedIn group so it's a cool group so yeah if you're in Denver sign up for the salon
228. 36:47so long we're fancy it's French [ __ ] was that that was a Dave Chappelle thing wasn't it um that's silly it's amazingly
229. 36:59smart to drink water yeah it's like you know all of our cells run on it and when you don't drink enough of it
230. 37:05bad things happen so flush that [ __ ] out and I don't drink enough on these lives so by generally I'll do these for like
231. 37:15two hours my ass gets numb and my voice just gets like oh I can't talk people in the live are like drink some water buddy time to go to bed
232. 37:30my fans to take care of them all right and if ultimately wait it will ultimately be the end of human dominance
233. 37:39on Earth rabbits in the Headlights um maybe maybe not maybe maybe not you know it it might be the end of humans
234. 37:54exploiting humans to the extent that they do now to be prosperous which would allow us to be more empathetic and more self-expressed
235. 38:04and less [ __ ] is a possibility um but I suppose there's that possibility uh yeah I think what what we all individually do
236. 38:19right now like in the next year is is more important than I think any of us realize each individual you know if I think
237. 38:29about just one person that I've talked to that has you know was skeptical of this or afraid of it or whatever and they kind of had
238. 38:39that Epiphany they had their their uh Uh Kevin McAllister moment there their home alone moment um and and now like like I just know so
239. 38:52many people who are like their lives are transformed in a positive way um or or in a completely in flux kind of way right
240. 39:02like they're excited but they're terrified because things are changing but at least they're in it and they're
241. 39:07they're active and so I think the more of us that get curious about it and get exploratory with it and if and not evangelize it's not the
242. 39:19right word just share what we're learning good or bad um I think it's going to have a dish disproportionately positive impact on
243. 39:28the trajectory of of how we use these tools as Society so um I I don't at all believe that we are at the mercy of the [ __ ] tools that
244. 39:38we built yet we still have a chance we still have a fighting chance we gotta fight for it we gotta fight for the
245. 39:50right to not party but more like uh responsibly use artificial intelligence yes yes yes yes yes yes yes yes yes yes yes yes yes okay
246. 40:05sorry Tick Tock jump into the bottom again I'm scrolling up to figure out where I was and the girls ding questions
247. 40:26I can screenshot as well okay you're helping one another I like that you're helping whoever it was that couldn't
248. 40:32screenshot troubleshoot think it was local settings all right AI learning lab insert caption is half off screen oh that's because look
249. 40:44it's just I just have a phone on a stick there you go is that better sorry oh he's helpful but he doesn't seem seem bright he's
250. 40:59he's slow on the uptake all right that's Kylie's special like that thank you that's incredible um all right Andrea's helping felonious
251. 41:12it's good it's good Tech Support Tech Support is right around the corner here on the AI learning lab
252. 41:22all right Robert Rossi killing it in the number one gifter badge Arena sent Donuts thanks for the donuts what phone
253. 41:29do you have I have an iPhone 11 Pro Max I don't know the sort of big one the four inch one is gonna oh how is AI going to help non-profit
254. 41:44organizations holy crap let me count the ways what are non-profit organizations perpetually faced with not enough resources
255. 42:01because not only do they have to go beg for money all the time they don't pay well oh David gold I'm getting many worlds thank you for the
256. 42:12worlds fight a World on a String I I good Lord thank you I don't I need to come up with the what do I do for a world
257. 42:24it's the world it's the world no I'm not gonna do it I promise I can't I can't resist it but thank you so much for all of that
258. 42:33goodness goodness um so so how AI is going to help non-profit organizations is lots and lots of ways it can it can make you thank you very
259. 42:45very much for that that does very nice um it can make you much more productive personally right so it can help you
260. 42:54write emails it could help you write personalized emails to donors it could help you write personalized stories
261. 43:01about you know impact that you've had much more quickly it can you know generate lots of organic content
262. 43:09um it can help you write grants it can help you read you know Grant requests or whatever their Grant offers or whatever
263. 43:17or if they're not rfps rfps in in the business World um it can help with program development it can help with um
264. 43:28you know creating training materials like it it it's in a lot in a lot of ways you know I think about AIS for non-profit is is it feels like
265. 43:42one of the greatest areas of impact because they're so under resourced um and and as the as the open source large
266. 43:51language models get increasingly optimized where you're going to be able to install like a GPT 3.5 like
267. 44:01um large language model natively on your phone without having to connect to the internet think about that anywhere where
268. 44:07you've got third world countries or rural um you know thanks for the donut you know rural uh communities that don't
269. 44:16have access to the internet you're going to be able to have these tools there locally and and so you'll be able to
270. 44:21build all sorts of amazing tools um the way I would think about it is this start to get literate if you're not
271. 44:31literate the reason I have these on screen for my entire lives here is like these are things to just go play with go
272. 44:37start getting used to playing with them like Pi dot AI is one that you can talk to and it talks back to you so there
273. 44:44there might be things with accessibility that could be really powerful with that one it's very very conversational so it
274. 44:50might be good for kids in some way um and and just start getting used to these things and and the way to approach them
275. 44:57is rather than think about how can I use AI for non-profits think about your non-profit and say what's a problem that I'm trying
276. 45:05to solve and then start to figure out how can I use these tools to solve that problem and it it totally it changed my
277. 45:14relationship with these tools where if if you just sit down with um like chat GPT you know it's this little bar
278. 45:26and it's like it's kind of like a Google search it's like well what do I do with it and it because it does so much like that
279. 45:34is such a deep well to dive in if you just sit in front of it and go what can I do with this it's almost impossible
280. 45:39but if you don't start with the technology start with your problem start with you know what's one problem
281. 45:46that you've got within your non-profit that if you can solve it would make the biggest difference start with that one and say okay
282. 45:55how like what do I need to solve this and and just go start interacting with these tools and and and start having a
283. 46:02conversation with these tools almost like you just hired an intern who's um an idiot savant right very very simple you've got to tell them
284. 46:11everything but they know everything it's almost as if they've memorized the entire internet that's what it's kind of like working
285. 46:20with these things you need to give them really good direction because they're dumb as a stone but they know everything and if you give
286. 46:28them good instructions they're your best friend they're your best assistant so start with a problem and try to solve it
287. 46:33that should help but I like staggeringly powerful impact on non-profits I think for ones that embrace it like if if I'm
288. 46:42in a non-profit and I'm not like leaning heavy into this world idiot do you know how to wait ubicate a property on Google Maps with
289. 46:56only a couple of picks of the property I don't know what ubicate means do you know how to lubricate is that a typo
290. 47:05well hell now subtitles right across your eyes on tick tock's Revenge oh you should be able to turn off subtitles too
291. 47:17there is not is-17 at least not on Apple's website ah yeah iOS 17 right now is in beta so you have to go into your settings
292. 47:29go into your settings and search for beta and I think you have to sign up for beta access so it's probably in like software
293. 47:37updates and then you turn on beta access and if you turn that on iOS 17 is now a public beta I just got
294. 47:44version two but I still don't have the stupid personal voice thing so I hope that helps I hope that helps it's
295. 47:566 a.m here here in Majorca wow welcome welcome Matthias I uh hope you're gonna join a fine coffee or tea uh I don't
296. 48:06know what you what what what's the beverage of choice at six in the morning in Majorca 12 minutes not bad it's early
297. 48:12your mountain time I am Mountain Time Kyle answers questions I do I do darn it I don't always answer them well but I do answer them
298. 48:21you well you can be guaranteed is that I will answer your question what's up for grabs is you know all right but what are his qualifications
299. 48:37sir do you have any qualifications hmm rather well I was trained at the Academy of the Arts in the fine thespian art of acting
300. 48:48no no I meant to to be on an AI Channel what are your qualifications to talk about machine learning in deep mathematics rather
301. 48:59seemed to be a bit caught off guard with that one a little a little hot in here hmm grab my pizza uh so I went to World coin oh Thelonious
302. 49:14that's right and I I gave away my Iris data to the to to the orb and all I got was this sticker well we're in America
303. 49:22so that I think they just didn't the SEC just tell coinbase you can't trade any crypto at all other than Bitcoin so yeah
304. 49:29being in America if you want to do anything with crypto sucks but with what Thelonious is talking about there last week
305. 49:37Samuel Altman's new company which is called World coin launched and what worldcoin is as creepy as this sounds I
306. 49:44actually think it or something like it's really important so what what they're trying to do they've they've created three components
307. 49:51they've created um a a transaction service so like an app like PayPal where you can you know transact
308. 50:03um they've created a crypto coin called World coin and then they've created this thing called World ID and World ID is you get
309. 50:11verified as a human so you go to one of these centers that have these orbs and they're like these Chrome balls do I
310. 50:17have anything yeah I did yeah they're like this size but they're but they're Chrome and they got a hole in them and you go up to it
311. 50:25when you stick your eye on it it scans your iris verifies that you're human and then um then you get a world ID and that's
312. 50:34blockchain verified so what they're creating is a network of verified humans and I think as we will see what the 2024 elections are going to
313. 50:45be like but I have a sneaking suspicion that all this AI [ __ ] is going to make lots and lots of misinformation it's going to make 2016
314. 50:54and 2020 look like [ __ ] child's play um so I think I think after that there's going to be a real demand for I want to
315. 51:01know that someone's a human I want to know that that video that I'm watching hasn't been modified I want to know that
316. 51:07that article that I'm reading you know is the original and hasn't been modified and so I think this is a step in that
317. 51:13direction is it creepy as [ __ ] yeah now apparently they're not keeping the data they're not keeping the biometric data I
318. 51:20think you can explicitly give them permission to use it for other [ __ ] which I don't know why you do that but
319. 51:27whatever we've already given away all of our [ __ ] to Google um but the idea of of having a verified network of actual human beings I think
320. 51:37is a really interesting idea and tying it to cryptocurrencies um fascinating the long-term vision of world coin is to
321. 51:45have it be a platform for Universal basic income funded by the prosperity that AI generates and if you listen to Sam Altman I mean
322. 51:55he's very clear he thinks that this is going to have a massive disruptive effect on employment and there's going to be a need for there's
323. 52:04going to be an increase in prosperity but a decrease in the number of people needed to create it and so if you can tax AI companies that
324. 52:14are benefiting from that Delta and turn that into Universal basic income then that could be kind of cool
325. 52:20so that's that's about anyway glad to see you're still here and they didn't zap you with a laser and Fry the brain but you know
326. 52:29maybe they did you just don't know it yet does AI really work on resumes oh what's your name Knight Naya Naya
327. 52:42it works incredibly well on resumes so um so depending on what you want to do you can copy and paste your LinkedIn
328. 52:52profile and paste it into chat GPT and then copy and paste in a job description and even copy and paste in an old
329. 53:01version of your resume and then say you know update my resume with my current work experience and follow resume best practices and um
330. 53:13you know Focus my experience you know to apply for that specific job and it'll do it and then you can just tell it to
331. 53:19write you a cover letter for that job as well and it'll do that it's amazing at that stuff chat GPT like all of these large
332. 53:26language models are particularly good if you feed them a document and tell them to analyze it in in any way they're
333. 53:34really good at that like amazingly good at that yes keep watching you'll learn well well you wait to have your question answered
334. 53:44that is kind of the sad reality of this channel it's like I just want to know about Lang chain why doesn't he talk
335. 53:53about land chain I want to know that Lang chain is he ever going to get to my oh that was interesting oh I didn't know
336. 54:01I could do that yeah that's kind of what this channel is all about is frustrating the [ __ ] out of everyone here
337. 54:07I hope I I hope I'm succeeding all right that dude is you that dude is you did you know that they used AI to synthesize
338. 54:20Val Kilmer's voice and Top Gun oh he did I heard that yeah and did the did the new Beatles song came out yet came out yet what did the
339. 54:31did the English language get good yet let's see new Beatles song I thought it came out end of July
340. 54:52Beatles release final song when are they gonna release it maybe it's August
341. 55:17well that was a useless article tell me when it's coming out when is it coming out I don't know when it's coming out but they um
342. 55:29this was something so this was like a cassette recording that that Lennon recorded and it was so noisy and bad
343. 55:35they've tried to release this thing like two or three different times and they failed every time because the technology
344. 55:40just sucked and then all of a sudden with this generative AI stuff they isolated his vocals got it clean enough
345. 55:47to put in a track and recorded the finals Beatles song so I don't know if it's out yet but it's coming
346. 55:54oh wait uh two release when NPR you'll have a date please tell me you've done some reporting here when is it coming I don't know
347. 56:14yeah but yeah all that all of that synthesis stuff I mean this is one of the things if if you're following the SAG and and wga strike
348. 56:26strikes um there's a lot of a lot of what they're talking about is you know actors are going to be synthesized and and I
349. 56:34think in the future there was an interesting piece in The Hollywood Reporter By Justine Bateman um Jason Bateman's sister and she's
350. 56:43she's a brainiac she's really smart but she was talking about you know she said to the SAG folks this is your last
351. 56:49strike where you have any leverage because you know in the future you're going to be competing against
352. 56:56young digitized Val Kilmer and you know perpetually young Tom Cruise and they'll just be able to pop people
353. 57:04into whatever movie and then you know the future after that is that's all going to happen dynamically I want to
354. 57:10watch Top Gun with Jim Gaffigan in the lead okay there you go yeah so it's just crazy I just ate my cat all right hope it was tender
355. 57:25um what site was that to animate animate faces it was called d-i-d.com as in like digital ID I think Dash id.com
356. 57:41do you work on AI well I don't know maybe do I I spent two hours yesterday no three hours yesterday writing two prompts
357. 57:56just trying to get them right and then I did which is cool it's really gratifying when you get the prompts right what's what's funny about
358. 58:05prompting with these things is that you can get to good really really quickly and then to get it to great or or like
359. 58:15consistently good every time that that takes a lot longer so I did that um today I was trying to pull some data
360. 58:25from our database and I had to pull some crap out of postgres on Heroku using sequel queries I don't know SQL queries but
361. 58:37there was one in there and it needed to be updated so I just copied and pasted that crap right into chat GPT and I said
362. 58:44update that sequel It's query like tell me tell me what needs to be updated in this from a time and date perspective and it did that
363. 58:54and then at one point it was giving me inaccurate data and I said I pasted it back in and I said why might be this be
364. 59:00giving me inaccurate data I'm expecting this and I'm getting this and it said oh well it looks like there's a discrepancy
365. 59:07between your dates and I said well can you analyze this and is there is there a way that you could fix it so that the dates are better
366. 59:15and it said well in the current SQL query you've got all your dates hard-coded and we can change those to be
367. 59:22dynamically generated and I said yeah let's go ahead and do that so it completely rewrote the query and
368. 59:28inserted Dynamic date generation and it fixed the data problem now it took me like an hour to do that but
369. 59:37the alternative would have been because I don't know SQL and well I could have figured it out it was a really long
370. 59:43query I didn't feel like figuring it out I didn't have the time to like learn SQL and learn this query
371. 59:50I would have just given that to my tech team and they're in Prague so they would have charged us you know the 80 or 100
372. 59:57or whatever the [ __ ] they charge us they charge us a lot of money and it probably would have taken him two hours to like
373. 1:00:02open it up figure it out analyze it all that sort of stuff so in an hour I did it myself fixed the query got the thing working
374. 1:00:10um and solved the problem so so that's the kind of thing these things can do where like that's one of those
375. 1:00:17just sort of things in business that you accept that if you come against something that's a little too
376. 1:00:22complicated for you for your skill level you're like oh just hand it off to an expert well with this stuff you've got
377. 1:00:30kind of an expert in your pocket and if you're willing to sort of go through the pain in the ass of learning how to talk
378. 1:00:36to these things they can fill in where you might otherwise have to hire people so so it's the kind of thing that allows
379. 1:00:44a single person like a single proprietor or a two or three person startup to be able to act like a five to ten person
380. 1:00:53startup or 20 person startup if they're really clever with it these are ability amplifiers like dramatic ability
381. 1:01:00amplifiers it's not just words it's amazing uh so yeah and then I build content generators and I'm re-architecting my
382. 1:01:08entire product vision of my company my company's called storyvine and it's a automated video production
383. 1:01:15platform we've been around since 2012. so we're an 11 and a half year old company and have not used AI for the
384. 1:01:23automation of our video editing we found another way to do it um but I'm completely rethinking what the
385. 1:01:31product is and what the value is and how we sell it and how we deliver it and so yeah so I work on it a lot
386. 1:01:38but I am not a machine learning guy so I I love jumping in early on technology at the moment at which it switches from only
387. 1:01:50the Geeks can play with it to the rest of us get to play with it and for me that moment was November 30th 2022. chat
388. 1:01:58gbt absolutely represents this single you know Hatchet moment in history where before that machine learning was for the
389. 1:02:06researchers and the scientists and the mathematicians and after that it's the rest of us get to play with it very
390. 1:02:12similar to the World Wide Web how it it sort of landed on top of the internet right the internet previously was
391. 1:02:20command line and it was researchers and Librarians and all that kind of crap scientists and mathematicians and then
392. 1:02:27the World Wide Web comes along and all of a sudden we can make magazines and we can make websites and we can make blogs
393. 1:02:33and everyone can be self-expressed and now we can build social media now someone can invent the iPhone and all of
394. 1:02:39that leads up to this [ __ ] where everything's gonna change all right I'm having trouble get AI to generate graphics with text yes that's a
395. 1:02:49problem that will be fixed but you just kind of have to deal with it right now so you have to do it dust off your Photoshop skills do your
396. 1:02:57text in Photoshop or go play with if you've got access to photoshop go play with Adobe Firefly and they've got a
397. 1:03:04text generating if you want to do headlines and [ __ ] like that they have a cool text generating thing where you can
398. 1:03:09do Standalone text with with patterns and things in it um but yeah the it wasn't sdxl it was either a new hang on
399. 1:03:25there's a new version of stable diffusion or it was a new version of no no no you know I think I think I've lost since I don't
400. 1:03:38have a track sense of time anymore since I haven't been doing my lives the days have all Blended together um
401. 1:03:45sometime last week I think this is a there's a new model coming out from open AI thank you for the rose I appreciate
402. 1:03:52that there's a new model coming up from open AI um they're the ones that have Dolly too which is the image tool sitting inside Bing
403. 1:04:02um it sucks it's it's bad like if you compare it to mid-journey and stable diffusion it's just not on the same
404. 1:04:08level but there's apparently a model coming out from them that does text really well so text is coming so
405. 1:04:16what's your name Johnny's 24 honey okay 9.5 out of 10 on the name love the name thumbs up if if you're gonna be in here you should
406. 1:04:26you should work on your name game because we got good ones in here I know AI learning lab really straightforward
407. 1:04:31kind of boring you know I would give AI learning lab a 6.8 out of 10. not all that creative it's clear but not all
408. 1:04:40that's out thank you Johnny Johnny's 24. honey bringing in the Boombox and the world appreciate it Saturday Saturday come to the buck
409. 1:04:53tractor pulls anyone else from redneck Pennsylvania I grew up in York PA I've been to tractor pulls what's a
410. 1:05:05tractor pull you ask go look it up on YouTube Johnny's 24 honey coming in with the lightning and the whatever the hell else that was unicorn
411. 1:05:17Saturday Saturday Saturday all right I found a site with lots of famous voices you can type a phrase and download it oh that's cool felonies
412. 1:05:25remember what it's called I got you on London wishing my brother a happy birthday super cool yeah you know what's that what's that uh
413. 1:05:35oh Christ what's that site that um famous people you can pay them to like do a recorded message for your loved ones that that
414. 1:05:45site's over right you're gonna be able to just do that Vinnie barbarino hair oh yeah I promises what was his uh tagline
415. 1:05:59I remember the fonzes what was Vinnie barbarino's ah welcome back your dreams are your ticket out welcome back
416. 1:06:14welcome back Connor that was a great show damn was that a good show thank you for the memory that was really cool it's
417. 1:06:20free of course and no sign up wait what was the site Thelonious oh 101sounds.com see this yeah I see how to
418. 1:06:28see it hang on hold on hold your pants dancing on this we're doing the best we can good Lord man one 101
419. 1:06:42soundboards.com uh who's a good one uh uh Samuel
420. 1:06:57Jackson Samuel Jackson play what again say what again I dare you I double get you [ __ ] say what one more goddamn challenge Black on
421. 1:07:11just take one again Samuel Jackson isn't black [Laughter] oh it's comedy it's comedy where's the one how do you make him say it where's
422. 1:07:25the thing help me Thelonious I don't I don't I don't know where to find the thing where's the thing to make him say something
423. 1:07:34Will Ferrell sounds Jimmy Hopkins sounds Christina ritzy Pulp Fiction oh TTS computer voice
424. 1:07:50oh enter your text to be spoken um paste that was a thing we did before generate it we're making your sound
425. 1:08:05your phrase is being generated text to speech technology the process of adding into the sound board usually takes a few
426. 1:08:12seconds to a few minutes okay when are they gonna tell me it's done I'm whiny I don't know how to use this technology it makes me sad
427. 1:08:27anyone else get like this where you like you go to something new and if you can't figure it out within 30 seconds you're like yeah you hurt
428. 1:08:34I don't like being incompetent I think it's an 80d thing patience is not my virtue uh Jenner okay generous oh tssq
429. 1:08:52the waiting list of phrases that are in the queue to be generated is one um in a world where brains were bites and
430. 1:09:04wisdom was Wi-Fi I'm the all-knowing simply don't lose downloaded the entire universe into my obnoxious of where do I
431. 1:09:11close oh yeah that was it that was the one I just did and they chose me yeah and they chose all right so that was 101 soundboards.com
432. 1:09:26not great but again with with all of these AI tools so as you play with these some of your response might be that's it is it me or
433. 1:09:36does that suck yes a lot of this stuff sucks right now but how it is right now is as bad as it's ever going to be so the fact that we can
434. 1:09:50do this at all is insane and so it's all going to get better and better and better so the investment you
435. 1:09:57make now to just get up to speed on what this stuff is how it works and how all of this stuff is related how the tools
436. 1:10:03that we use to do language stuff are the same tools that are behind image generation and audio generation and
437. 1:10:09music generation there was a site um a buddy sent me to today hang on let me uh is it slides
438. 1:10:43this is a workshop I'm doing for a big company with the dude named Kenny Friedman who's awesome and it's called suno.ai okay
439. 1:10:55so check out suno suno dot a i uh oh try the alpha on Discord that's right this is on Discord so this is like mid
440. 1:11:07Journey you have to do it in Discord or as soon oh there it is um so go to suno.ai and then you go into these chirp
441. 1:11:23channels and you do come on I'll get to the bottom you do slash chirp is it slash chirp yeah slash chirp
442. 1:11:38and then we're gonna do we'll do like uh uh what was my thing well uh we'll do we'll do bouncy um pop and then you can put in your lyrics so
443. 1:11:53I'm just gonna put in those lyrics that we generated before in a world where brains were bites and wisdom was Wi-Fi
444. 1:11:58I'm the all-knowing simpleton that one and so now I hit submit and now that's gonna just like mid Journey will make
445. 1:12:05you an image this is going to make us a song where it actually sings those lyrics on the style that that we created I like
446. 1:12:13I just found out about this today so I don't know how to do this well and I might not be doing it right like I I'm
447. 1:12:21not sure if you type when you swipe type slash chirp if you type the thing after it I don't know if it's actually paying
448. 1:12:27attention to that but I think it is but we'll it's generating I guess we'll come back to that what else can I show you here while it's
449. 1:12:36doing that I'll go show you so mid Journey looks like my internet is slower than crap [Music] um if you haven't played with mid during oh
450. 1:12:48yeah there's it's slow I got I got slowness going on if you haven't played with mid-journey um the way you do it you come into
451. 1:12:57Discord if you haven't played with Discord before here's all your channels once you're in a channel these are all
452. 1:13:02the different or these are servers so those are your different communities these are channels within a server and
453. 1:13:08then so I'm in the salon Discord right now we have a mid-journey channel and so you type slash imagine so just like I
454. 1:13:15typed in slash chirp for that suno.ai type in slash imagine and then you um you know paste in or you type in a prompt so um God rays
455. 1:13:31bursting through a thunder head Storm of the Century Bang and now that'll go off and make that image for me
456. 1:13:46but let's go let's see my song's done is Our Song done yes all right so you heard it here first [Music]
457. 1:14:01downloaded the entire universe it's like Noggin by the way I plug in the toaster so that was crap but in a horrible Angel fights and wisdom
458. 1:14:13with Wi-Fi I mean all knowing something he's downloaded the entire universe into like Noggin where do I plug in the toaster actually
459. 1:14:20you know not bad not bad so this is all coming and and again let's see do we have our images here oh those are cool check those out
460. 1:14:32let's see uh this is pretty groovy I like the one with the lightning although this one's the most god-ray let's do we'll do upscale one
461. 1:14:51so there's that beautiful thing and I'm like oh crap I wish I wouldn't have done that as a square image I wish that had done that as as uh
462. 1:15:00what's it called it 16x9 well you could do that now so I'm going to do aspect ratio of um 16x9
463. 1:15:14and all the other stuff's good I'll do Zoom of 1.5 make it a little bit zoomed out from that thing and now it's gonna make me
464. 1:15:29four versions of this that are wider it'll extend that image out so crazy huh isn't that crazy crazy all right then we're moving on up Tick
465. 1:15:43Tock just jump me to the bottom because it's awful and it hates me it hurts my it hurts me it hits me hated it internet
466. 1:15:54hopefully some people are falling asleep to this because they said they do they I can't fall asleep without you I'm like
467. 1:16:02that's that's my favorite sort of compliment ever I think [Laughter] ah it's comedy we're doing comedy
468. 1:16:17tonight okay I found out where it was do you think anything about the scientific wait do you know anything about the
469. 1:16:24scientific background regarding AI optimization modeling Etc I are you just talking about being a user I know a
470. 1:16:33little bit about it I know about how it works I can explain how the large language models do what they do I can
471. 1:16:38explain how diffusion models do what they do um I know marginally about how embedding works and things like that I know about
472. 1:16:48Technologies like Lang chain but if you if you're getting into like you know weights in the mathematics of it I'm useless
473. 1:16:56but if you have basic questions about how this [ __ ] works yeah I'm pretty good on that I think it's actually important to know that stuff
474. 1:17:04um so yeah these are kind of cool I like the bottom right one all right let's make the bottom right one so I'm gonna upscale that one
475. 1:17:17oh you know what Let's do let's do it so we got this beautiful image right okay thank you Johnny's 24.
476. 1:17:28beautiful right so let's take this image and we're going to turn this into a video save image all right I'm gonna go to downloads
477. 1:17:43that's good so now I'm gonna go sweet it's very sweet of you uh we're gonna go we're gonna go over where are we going where are we going
478. 1:17:54we're making pretty pictures oh Runway we're gonna go to Runway ml .com and then we're gonna go to the login
479. 1:18:06I think we're gonna log in with the Google I think that's how I did this one we're gonna go to the Google and then we're gonna go
480. 1:18:17Gen 2 text to video let's see yeah okay so when you go to gm2 why is the screen so shiny is it because of this thing maybe
481. 1:18:31it's dirty good Lord what a mess all right so you come to Runway uh Runway ml you can type in a prompt but you can also upload an image
482. 1:18:46so I'm going to click on upload an image and there's a bunch that I've done before I'm just like a bunch of my cloud photos
483. 1:18:54and [ __ ] like that and so I'm going to say upload here and I'm going to go to downloads and I'm gonna go get my little God raise pitcher
484. 1:19:06and so there's my God Ray's picture right there and now I'm going to say generate and we're going to turn what didn't exist
485. 1:19:16before which is now this beautiful Sun bursting through the clouds moment that was generated out of nothing
486. 1:19:27and now we're going to turn it into a video and we're going to see pretty clouds with a pretty sunset wait what was that
487. 1:19:36try unlimited that's an upsell that's what you call marketing folks someone was sitting there going like hell you know what while someone's
488. 1:19:44sitting there waiting for their little Cloud picture to turn into a video why don't we do a little pop-up that says spend more money
489. 1:19:52yeah that's a good idea we'll get more subscriptions that way Let's see we do the math if we can get another five dollars out of ten percent
490. 1:20:01of the people that do that we can have as much as 25 30 every month I like it I like let's do that let's
491. 1:20:09let's do that pop-up is that what they call it a pop-up Pop Tart pop up we do Pop Tart pop up with the with the give us more money and
492. 1:20:18then they'll give us more money and then maybe I can get that Toyota Super I was looking at oh here's our video wait ah
493. 1:20:25okay let's see does it suck oh that's really cool the clouds are kind of parting and disappearing like this is it's very religious it's a
494. 1:20:43lot of a lot of uh that's cool they don't really look like clouds moving they kind of look like set pieces moving out of a play but
495. 1:20:54but anyway pretty groovy huh so that's Runway ml Gen 2 upload an image so that the workflow there was go to Mid Journey
496. 1:21:04make an image download it or expand it however you want make it whatever the hell you want it to be then come over to
497. 1:21:11runwayml.com go to the app go to Gen 2 and then instead of typing a prompt upload that image and just start
498. 1:21:18experimenting with it and like that one's weird it's not quite right so you know I would probably dick around with
499. 1:21:24that one a bit and and get it to do something get it to do something cool but you know um Troy Decker I'm only here for the
500. 1:21:38entertainment factor LOL I don't know what's entertaining about this Mr Decca this isn't funny this is artificial intelligence
501. 1:21:54and there's nothing intelligent about it all right carrying on could you explain how machine Learning Works thanks greetings from Brett
502. 1:22:06Buenos Aires Argentina Brenda Buenos Aires uh yes I can um yes let's go
503. 1:22:22ah there's the slide
504. 1:22:39okay so we'll talk about a couple of things here um so so
505. 1:22:54let's let's start with everyone talks about this thing called GPT chat GPT let's chat EBT or or as the the people that are like you give me
506. 1:23:04money when I chat jbt those guys have you ever noticed that those guys also always switch around the letters they call it jet chat GTP
507. 1:23:11I every time they say it I want to smash my face into the phone I'm like the letters are right there I know they
508. 1:23:17sound alike but like learn what they are and they'll do it like 10 times in a video Once is okay but like on screen it says chat GPT and
509. 1:23:26they're going chat GTP no anyway I'm not telling you how this [ __ ] works but one of those stupid three [ __ ] letters mean okay they mean
510. 1:23:38G is for generative meaning it's generating content so one of the misconceptions about these large language models is that They're copying
511. 1:23:47and pasting from other people's work right now while the lar the core base large language model was trained on most of the internet
512. 1:23:57um it's not copying and painting pasting it's actually generating so that's the G the p in GPT is pre-trained and and what
513. 1:24:07that means is that historically large language models you kind of have to bring bring your own data so you'd bring
514. 1:24:12your Radiology data and and you could do these sort of small relatively small data sets and you kind of have to supply your own
515. 1:24:21um in 2017 Google invents this thing called the Transformer which is the T in GPT um Transformer generative pre-trained Transformer and
516. 1:24:33the Transformer allowed these machine learning models to be trained on massive data sets so open AI in particular said
517. 1:24:41hey why don't we if it's all about the data and throwing as much data and and high quality data as we can why don't we
518. 1:24:49go get as much of the public internet as we can and ingest that into this machine and we need lots of gpus to do that so
519. 1:24:56that's why they partnered with Microsoft there's all sorts of you know that that's why this is going on so
520. 1:25:03it's a model that's been trained on all of the internet using this technology called the Transformer you'll you'll
521. 1:25:09hear them talk uh talk about llms sometimes that just means large language model so chat GPT is is built on top of
522. 1:25:17a large language model called gpt4 that's the underlying um I think they're calling what are they calling them now they're calling them
523. 1:25:25Frontier models so if you're doing the sort of expensive base model creation that's called Frontier models and then
524. 1:25:33chat CPT is an application layer sitting on top of it then you've got this thing called the latent space and and this is this is
525. 1:25:41where we get a little bit technical but this is where it also gets fascinating as fun okay so what you're looking at there
526. 1:25:48is a 2d representation of just like an x y plane with a bunch of um what are called vectors so when you when you
527. 1:26:00train these models you you basically go pointed at a lot of documents and it takes all the words of those documents
528. 1:26:07and it groups them in semantic clusters right and so the black ones and the blue ones and the purple ones are all
529. 1:26:13different semantic clusters and you might have and this is dramatically simplified right but you might have the
530. 1:26:19word dog and the the word dog can live in one of those clusters as the the cute little chubby animal that sleeps on your
531. 1:26:26carpet and the other one is you know that guy was a dog and he didn't treat his girl right right and so so you can
532. 1:26:32have the same word that's got multiple meanings so you have these clusters this is a very simplified 2D representation
533. 1:26:38these Vector spaces are like 500 or a thousand Dimensions like these are these are massively complicated mathematical spaces Okay so
534. 1:26:51that's that's sort of all the math you need to know but but you'll you'll hear things like the latent space and when
535. 1:26:58they talk about the latent space they're talking about these mathematical clusters of words that are semantically grouped
536. 1:27:05now how the how the models work is based on the The Prompt that you type in it basically takes your prompt and turns
537. 1:27:16that into a mathematical formula and says okay based on all of the things that you've done what is the next most
538. 1:27:22probable word and so it kind of works like this and then it does that over and over and over again so if our prompt is
539. 1:27:29the quick brown fox jumps over the lazy blank we know as humans if you know that phrase from typing class
540. 1:27:37um that word is dog right so it it sort of takes your prompt and it does this mathematical thing and then effectively
541. 1:27:44what happens in the latent space is that if it essentially deselects semantic regions that are not of interest to it
542. 1:27:52but you'll see there's these light blue dots here um are are the are the ones that it looks at and within those light blue
543. 1:27:59dots one of those dots is mathematically the highest probability of being the thing that comes next and if you look at
544. 1:28:07what's actually in the database it's a long string of numbers it's wait let me zoom in on it because it's just crazy
545. 1:28:16so so what these Vector databases are is is there databases of just these long numeric strings which is basically just
546. 1:28:24the point in the vector space that this word is and that number is associated with the word dog and so then it inserts
547. 1:28:32the word dog into your sentence it's the it's the one that's got the highest probability and so when it writes you a
548. 1:28:39full paragraph of stuff it's doing it word by word by word with this mathematical calculation that's also why when they talk about
549. 1:28:47large language models hallucinate what what they're basically mean is that they're full of [ __ ] and they lie they
550. 1:28:53will confidently give you an answer even if it's wrong and why that's happening is it's just saying hey out of all of the words that I've
551. 1:29:02got in this you know out of the trillions of words that I've got in this thing um this combination of them is sort of
552. 1:29:10the highest probability of what you asked for but and it's just doing that mathematically one word at a time it
553. 1:29:16actually has no comprehension at all of what it just wrote so sometimes you'll say hey you got that wrong could you do it again it'll say oh
554. 1:29:25I'm sorry I got that wrong how about this and it'll get it wrong again and sometimes you'll do that four or five
555. 1:29:31times and you're like how can it possibly be getting this wrong especially with things like counting or
556. 1:29:36mathematics or like give me a list of 17 things it'll give you 14 and then it'll give you 18 and then it'll give you 12
557. 1:29:43and you're like why can't you just give me 17 because it doesn't know so future versions may know but but this
558. 1:29:51is basically how they work and and how the images are generated is is is basically just a layer that sits on top of this model
559. 1:29:59um it's it's fascinating how it works like whoever figured this out deserves all sorts of medals and prizes
560. 1:30:06um it's probably been lots of people over the years but that's how they work so I hope it helps I hope it helps you know
561. 1:30:14I'm just out here trying to heal and and sometimes it doesn't heal and sometimes it helps sometimes it helps sometimes it
562. 1:30:23doesn't heal other times it heals why does he talk like that I don't know I don't know he thinks it helps please don't
563. 1:30:35is that about twerking Lou Grant oh Lou Grant was good that was a good show Gen X is like I got a pile of gen xers
564. 1:30:44in here yeah that hair is something oh geez it's 11 24. I got work tomorrow you people so demanding all right hope that was a good
565. 1:30:56explanation of that um how did you get AI on your blackberry that's good that's solid that's a solid burn could you please explain how
566. 1:31:06machine Learning Works in all caps thanks greetings from Buenos Aires we just did that but you're welcome nice
567. 1:31:12shirt too thank you yeah I like this shirt UB Cafe is Spanish slang command wait oh oh Yuba Kate oh I see yubakate is a
568. 1:31:21Spanish slang command that means get your bearings oh that's cool that's cool felonious touch and hold the subtitle and you can
569. 1:31:29move it on screen you're welcome touch and hold the subtitle oh oh for the person thank you Dave gold I'm
570. 1:31:37getting the worlds He Got the Whole World in His Hand he got the whole wide world in his hand wow David gold
571. 1:31:50knocking the marbles out thank you very much I appreciate that that's very cool um what do you think of adobe's photo AI they are building
572. 1:32:02um I think it's remarkable so here's what's happening with Adobe um as of January or February last year they
573. 1:32:11weren't even sure if they were going to get into this space because they knew it was kind of problematic and potentially
574. 1:32:16cannibalistic on their business and then uh they said some version of [ __ ] it let's try and they did it really smart and when
575. 1:32:26you when you have a company like Adobe that's got good resources and smart talented people when they decide to move
576. 1:32:31they can move fast so they did a couple of things really right they they started out with this thing called Adobe Firefly
577. 1:32:37which if you've got a an adobe subscription a cloud subscription you can play with Adobe Firefly so that's
578. 1:32:45basically their playground and they've got all these modules they got a text module they've got a text to image
579. 1:32:51module and they've got what's the other one uh I don't know they got a couple of modules there and they're and they're adding a bunch more
580. 1:32:59um [Applause] so so they built a little playground off to the side and what what's smart about that is all those different modules they're
581. 1:33:09basically testing those right so if people use them and they get good feedback on them those are going to get incorporated into
582. 1:33:16their tools if they suck they just sort of flush them down the toilet right so that it's smart on that front the next
583. 1:33:22smart thing they did was they trained the the image models on public domain and Adobe stock images that they owned all the rights to
584. 1:33:32so none of their work has been trained on commercially produced you know stock photography or illustrations or you know people's [ __ ]
585. 1:33:42they're going to be allowing artists to opt out of having their images trained on or opt into having their images
586. 1:33:49trained on because some artists is it would benefit them to be in a model where they can be discoverable right so
587. 1:33:55some artists are going to want to do that at some point they're going to figure out compensation so how they've
588. 1:34:01built their model is in a very artist friendly way it's also very business friendly because mid-journey are a bunch of [ __ ]
589. 1:34:09Cowboys and they're like [ __ ] it we trained it on everything go ahead and sue us well everyone's going to go ahead
590. 1:34:13and sue them and they're probably they're probably not going to survive long term if they keep it up with that attitude
591. 1:34:21um stable diffusion's a little craftier because they're open sourcing all of their stuff so they're like [ __ ] it we
592. 1:34:26trained on everything but here you can have the models so if they get sued out of existence at least the technology lives on
593. 1:34:34um but adobe's making that kind of image generation safe for corporations which is brilliant and then the final thing is as they've
594. 1:34:44rolled those things in to photoshop like generative fill within Photoshop is just [ __ ] amazing like how they
595. 1:34:51Incorporated it and how you call it and how you do it if you haven't done if you haven't seen it I'll show you how to do
596. 1:34:57it let's go play with it because it's it's cool oh yeah this is this is an image that I did for uh for the uh for a workshop
597. 1:35:06that I'm working on we just hang on we had this Square image and we needed
598. 1:35:22it to be the whole width of a of a PowerPoint slide so I just did generative fill and expanded it there's a new version of the Photoshop
599. 1:35:33beta available woohoo all right so let's open something here oh that's kind of cool yeah that's kind of cool we'll do that we could do that
600. 1:35:45one what else do we have that was uh if you haven't played with this remedy thing you can do virtual headshots
601. 1:35:58that's the that's the cheesy the cheesy me in the in the jean jacket headshot all right let's let's open this thing up
602. 1:36:13all right so there's there's our our little photo and let's say we want it wider well I can just grab this thing and kind of extend it out
603. 1:36:29like that and then I can grab my Marquee tool and I can just highlight that white section overlap it a little bit and then there's this
604. 1:36:40context sensitive little menu bar down here and this button to the left says generative fill and so I just click on
605. 1:36:48it I can put in a prompt so so like what am I what I might put in here is oh let's do this we'll do um scary green monster with scales
606. 1:37:02and sweat I don't know and so it's now going to extend this image and if it's if it's a cool monster then this would make a cool ad for that
607. 1:37:16muscle car that's not bad and it gives you three options that's pretty good [Laughter] I think that that one's probably a bit
608. 1:37:31more realistic but you know it sort of continued the thing on so so you can just keep doing that and you know I
609. 1:37:36could I could you know highlight the eyes and say generative fill and say glowing eyes and so now it should
610. 1:37:50make those eyes a little more scary here we come that's bad here we come we'll save the date that's pretty good
611. 1:38:02there you go there's some glowing eyes and so anyway so that's generated Phil so I like it I I mean you know it's it's
612. 1:38:11the kind of thing that um I I think they're doing it smart I think I think adobe's doing it quite smart right now um
613. 1:38:26beverage without sugar he's having water he likes the water glad refreshing thanks still unsure about screenshots uh
614. 1:38:40can it pretty much wait can do pretty much can do it pretty much all others oh oh sorry to hijack you Kyle I don't know
615. 1:38:51you're fine you're good you want a screenshot you got a screenshot you got issues screenshotting you got to talk about it
616. 1:38:58you gotta you gotta bring it up you got issues get it well hey a bunch of people just joined what are y'all doing it's late
617. 1:39:06you damn insomniacs are you on the other side of the world anyway welcome if you're here my name is Kyle Shannon this the AI learning lab we
618. 1:39:13talk about stuff here we don't know what we talk about I'm completely unqualified but that's okay all right
619. 1:39:21uh it can also mean get real or get serious to indicate someone is being unrealistic oh that's cool are you sure you're not psyops I am
620. 1:39:30absolutely not sure I'm not psyops I might be psyops I might be gpt5 did you see the gpt5 trademark uh application anyone anyone no
621. 1:39:43let's go to the Twitter let's go to the Twitter you people got to keep up with this stuff come on can't be expected to do all of this
622. 1:40:02GPD five so open AI files a U.S trademark here's what's included artificial production of human speech and text
623. 1:40:16conversation of audio a conversion of audio data files into text voice and speech recognition so already in gpt4 is image recognition
624. 1:40:28they haven't rolled it out yet but you can use it on Bing um and it does all its wordy wordy stuff and now it's going to do audio audio
625. 1:40:37stuff and video can't be too far behind so gpd5 is likely going to be a pretty slick multimodal crazy ass
626. 1:40:48the the thing about all these tools right now so I showed earlier that that singing Generation tool and we're doing
627. 1:40:56mid-journey things here we're going to Runway ml for there and all these things right now are these disparate pieces of crap little
628. 1:41:05standalone and I don't mean to diminish the incredibly powerful pieces of crap that they are but they're just they're they're all raw right
629. 1:41:18and within a year we're going to start to see these things do this and these tools are going to get so sophisticated and so slick
630. 1:41:31so cool what's coming they're cool they're cool that's a cool cat that's really cool Kyle that's cool I'm learning AI technology awesome ah
631. 1:41:44the bear with my little dangly lights I love the dangly lights the Vanoss Thelonious I'm still here I didn't put
632. 1:41:51my eyeball into the orb just looked at it I don't know if I was lasered might have been lasered uh now and then as the Beatles song
633. 1:41:59release date 2023 when 2023 I know 2023 when is it is it out is it coming [Laughter] Hey guys does bars has an app
634. 1:42:19I don't know what you saying it won't ever be good enough to replace actual writers um are you sure about that you may wanna look into that
635. 1:42:36um I know a lot of writers that have embraced this and one who's one of the organizers of the AI salon
636. 1:42:48is a copywriter I met her at a creative mornings and I said what do you do and she said I'm a copywriter and I said this is like
637. 1:42:56back in January I said if you played with chat gbt she goes I heard people talking about that she came to came to
638. 1:43:03the first salon and started playing with it and like the second Salon she came to she was like and now she's like you know she basically
639. 1:43:14uses the tool to do most of her writing for her and and you know to be fair you know to your point you know is it
640. 1:43:25going to write the Great American novel or a fantastic screenplay on its own no but if you're churning out [ __ ] social media video Scripts
641. 1:43:36it can absolutely do that and it can do those things [ __ ] 30 times faster than you then you could get draft one you know you
642. 1:43:46could have 30 versions of those so so for certain kinds of writing it's there now and then if you start to look at more
643. 1:43:54sophisticated kinds of writing like screenwriting and novel writing and things like that um go check out a project called GPT author
644. 1:44:06which you have to be a little geeky to play with it you have to run it in Google collab but it's a series it's what it what it
645. 1:44:13effectively is is a series of prompts strung together to write a complete novel so what you give it is here's the kind
646. 1:44:21of story that I want here's the writing style that I want here's the number of chapters that I want here's my open AI
647. 1:44:29key and my stable diffusion key go and it writes an outline for a book with the number of chapters that you're
648. 1:44:37talking about with the style that you're asking for and then it writes one chapter at a time so it writes a chapter
649. 1:44:46then it takes the outline in chapter one and it writes chapter two then it takes the outline in chapter one and two and
650. 1:44:51it writes chapter three and so and you're just watching this and it's just going dude and then it gets to the end and it's
651. 1:45:00written this whole book and then it writes a prompt for stable diffusion to create a cover for the book and then it saves it as uh an ebook
652. 1:45:11and it does it in like I don't know three minutes now is the book brilliant not yet but go look at that like but is is it a
653. 1:45:25coherent well-structured story in three minutes that's a book yes it is so again my contention with these tools
654. 1:45:38is they're the worst they're ever going to be right now so making a comment like they won't ever be good enough to replace actual writers
655. 1:45:49they're already good enough to replace actual writers they're not good enough to replace all the writers but what's coming
656. 1:45:59is significantly Beyond I think what you're giving this technology credit for and listen if you're a writer and I'm a writer I've
657. 1:46:12written seven screenplays and a one-man show and all I do for my business is write you know story structures so I get it
658. 1:46:21it's [ __ ] existentially terrifying so I get why you want to be in that place but I wouldn't be so confident there if
659. 1:46:34I were you I would I would dig deeper and learn more um just learn learn what these things can do and in fact if you're a writer
660. 1:46:42and if you've got that position I would go play with all of these the reason I have all these up there is they're all
661. 1:46:47good at different things gpt4 up here in my opinion is the best writer of all of these but they've all got different
662. 1:46:54skills so depending on the kind of writing you do there might be certain of these tools that do things better so
663. 1:47:00write yourself an interesting prompt as a test and go test GPT 3.5 gpt4 test Claude test gpt4 at bing.com testbard
664. 1:47:12chat test this is meta this is the the Llama 2 model test Pi which is super conversational like all of those
665. 1:47:22like that it would be a good education for you just to see where these different models are which ones are good at what
666. 1:47:30um but I I would not sort of I I wouldn't I would not sit with much confidence in the statement that you made
667. 1:47:38because the [ __ ] that's coming is gonna melt some [ __ ] faces it's gonna be like the faces just fall off
668. 1:47:46what happened to my face I don't know I said that it couldn't write as good as a writer right before I got laid off from a writing job
669. 1:47:53[Laughter] all right hey buddy hello Stevo what's happening what's happening happening machine
670. 1:48:08learning tight language hmm yeah you got tight language that language is I like turtles Chris that's what got me
671. 1:48:19in trouble in my last life I showed the I like turtles thing and then that reminded me of the other guy and I
672. 1:48:24showed that video and then I said what the guy said in the video and then my live web so we're not doing Turtles joke anymore
673. 1:48:33but thank you good joke like it go look up I like turtles on YouTube if you don't know what we're talking about exciting times
674. 1:48:41sdxl sdxl looks pretty cool I would love a tutorial of mid Journey if you get a chance I did that hopefully you were here
675. 1:48:49I did kind of a kind of a minor one but I did did the slash imagine let's imagine what we could do here let's turn it into a video shall we
676. 1:49:02I'm a mid-journey mod if you need help oh that's cool uh uh I don't need help right now but it's cool that you're here um
677. 1:49:16I here's something that you could answer for me replicant which by the way 9.5 out of 10 on the name I like it um
678. 1:49:25I think with the qualifier it could be better what kind of replicant you know lazy replicant busted replicant I don't know anyway um
679. 1:49:38what would be fascinating for me to know if if you're connected to mid-journey in any way one is I heard there's only like
680. 1:49:4612 employees is that true that sounds um ungodly small stuff to me for something that powerful but if it is [ __ ] cool uh and then my other
681. 1:49:58question is uh when are we gonna not have to go and [ __ ] Discord to make mid-journey images please tell me
682. 1:50:04there's a website coming please please please please or an app or something anything Bueller um but thanks for coming and thanks for
683. 1:50:17uh offering up your uh thoughts oh replica grew up in Grand Cranberry Township PA I grew up in York Pennsylvania actually Wrightsville PA
684. 1:50:28just across the river from Lancaster or as they call it not from Pennsylvania Lancaster Pennsylvania Dutch Country
685. 1:50:37hey you going down to the mall today what are you hey what are you doing are you you want to go fishing later
686. 1:50:44that's what I grew up with you go down the mall today Kyle why would you why would you uh why would you go to college why would you
687. 1:50:52want to leave here yeah that's what I grew up with four four people out of my graduating class in high school left town four
688. 1:51:04they were like why would you want to go I was like why would you want to [ __ ] stay I don't know I just always I always had
689. 1:51:14the itch to be somewhere else then I got to New York City and I was like ah here's my people here's home now I'm living in [ __ ] Denver
690. 1:51:24I gotta get back to the East Coast because them East Coasters they just got a different Edge to them they're so aggressive yeah exactly I
691. 1:51:33like it [Laughter] oh Cameo Cameo that's the one God I'm I'm significantly behind on my questions Cameo was like an hour ago when I was
692. 1:51:44asking about cameo oh well up your nose with a rubber hose that's a good Gen X line what are the methods that will be used
693. 1:51:53for AI regulation and how will alignment become more efficient well Tick Tock astronaut I'm going 10 out of 10 on the name
694. 1:52:05Tick Tock astronaut I like it um I don't know what methods will be used for AI regulation I think that
695. 1:52:22I I like what's where Sam Altman's head on this is feels right to me that if you're building Frontier models which is
696. 1:52:29what they're calling you know like the people that make gpt4 gpt5 Bard you know so Google open AI Microsoft
697. 1:52:38if you've got the resources to make Foundation models that should be regulated and then what he's saying is if you're doing anything
698. 1:52:49the next level more powerful than that like the AGI stuff the stuff that's like as intelligent as any human
699. 1:52:57you know ever or as intelligent of all the humans combined if you're doing stuff at that level of power or capability then you need to um
700. 1:53:10get licensed like you get a license to build a nuclear power plant and then what he's saying is so so if
701. 1:53:18you're doing the super powerful [ __ ] you've got to license it if you're doing Frontier Model stuff like Foundation
702. 1:53:24model stuff that's got to be regulated and then anything below that he's saying don't regulate it which I think that
703. 1:53:31that sounds very right to me um and then how will alignment become more efficient so again back to open AI
704. 1:53:37I'm you know I'm a bit of a show for them just because I I you know what they're actually doing in the world like
705. 1:53:44a lot of people are talking right now like Elon with his you know x dot AI is going to understand the universe great
706. 1:53:50when are you gonna have this technology um openai they they just committed to dedicate 20 of their Computing um power
707. 1:54:03to what they're calling super alignment so so doing the alignment and alignment with these tools basically mean do the
708. 1:54:11imperatives of these tools align with the imperative of these beings that's alignment if these tools go hey these beings are a
709. 1:54:25bit of a pain in the ass and they're getting in the way of my imperatives why don't I just get rid of them
710. 1:54:31that's misalignment that's bad that's so super alignment says let's keep these things going in the same direction
711. 1:54:42that this is about prosperity and making these things better and not competing with them because the soon as soon as these things start
712. 1:54:52competing with these things autonomously then we're in trouble and that's that's the fear right so um so that's going to help and then and
713. 1:55:06then there was a big Consortium of the big players just got together in Washington and agreed to I don't know we're gonna do non-shitty
714. 1:55:15stuff together which that in a bus pass will get you downtown in the comments are easier to read against the darker shirt I know that
715. 1:55:26I know that I wear dark shirts all the time but I could actually read a little bit against this shirt for some reason but you know [Music]
716. 1:55:37you should astroguide.ai is this gonna put [ __ ] malware on my computer I'll be cranky if it does Astro guide dot a I I get
717. 1:55:52I get viruses so you don't have to um oh ask the Oracle oh this is funny dream meanings oh this is cool my co-founder
718. 1:56:05and I are looking at this [ __ ] all the time hahaha we talk about dreams all the time hello I'm the dream Oracle um okay let's see
719. 1:56:15um I dreamed I was let me see I dreamed I was a peanut butter and jelly sandwich that tripped and fell down a hill
720. 1:56:32and spent the afternoon picking small rocks out of my bread thank you for sharing your dream with me dreams can be quite Whimsical the
721. 1:56:45presence of peanut butter and jelly sandwich could symbolize the blending of contrasting elements or qualities in your life very nice
722. 1:56:55oh this is cool so this is this is what's commonly called a wrapper app right so they're taking this is some pre-prompted
723. 1:57:03pre-prompted stuff sitting on top of chechi PT 3.5 or GPD 3.5 the the base model this would be better if they used
724. 1:57:13gpt4 but it's a lot more expensive so I understand why they're doing 3.5 um that's pretty cool I like it
725. 1:57:20I'll send that to my business partner she will get a kick out of that type of prompt uh I did what where oh barbarino's tagline yeah
726. 1:57:33what where oh yeah that's right that's right Vinnie barbarino up in those with the rubber hose I'm so confused what where
727. 1:57:41that's right there's a cube your download is at the top of the list are you a machine learning engineer no but I've met some
728. 1:57:51what are your qualifications sir so what do you refer are you a machine learning engineer oh no what are your qualifications with regards to
729. 1:58:07it's called the AI learning lab I assume by lab you're a scientist correct no that's just marketing that's what we call marketing
730. 1:58:22a little thirsty carry on up your nose with a rubber hole all right man John Travolta was so good in that
731. 1:58:38you know it's amazing if if you watch Welcome Back Cotter the fact that John Travolta escaped that character like
732. 1:58:46there's some actors that when you do a character that identifiable like Henry Winkler it took Henry Winkler decades to escape the funds
733. 1:58:58and uh and Travolta Travolta escaped that character that was pretty pretty remarkable uh suno dot AI it's not slash chirp so so
734. 1:59:11if you go to suno dot AI It'll point you to their Discord server so you have to join their Discord server
735. 1:59:18and then you go into one of the chirp channels then you type slash chirp type in the style of music and then
736. 1:59:25it'll pop up a dialog box to either paste your music in or have it AI generate your your lyrics for you and
737. 1:59:32then it'll make the music it's pretty cool all right I gotta go to bed it's almost midnight here people what are you doing
738. 1:59:37with my life what are you doing what are you doing weirdos we're going down to the 405. all right let's see what do we got here
739. 1:59:49I'll answer I'm I'm it just jumped me to the bottom I grew up in a town of 12 000 people I was scared I'd be stuck there
740. 1:59:56tell me about it man I couldn't believe no one wanted to leave it's just a different thing like I and you know it's weird like it was as
741. 2:00:05foreign to me to stay there as it was for them to leave and I mean partial partial part of that was I was
742. 2:00:12an Air Force brat so I moved around a lot so I'd see me seeing some other places but I just I don't know
743. 2:00:20there was a lack of curiosity in that small town that just made me nuts uh replica whoa the message was a while ago and hello Nexus 9.
744. 2:00:38uh there's a small there's a small amount of employed team there's a site already in testing that's cool replicant thank you for sharing
745. 2:00:47that oh they announced us they they announced it oh cool yeah like listen I love mid Journey I've got it on my little I got it on my list
746. 2:00:59of of [ __ ] you should [ __ ] you should use I've got mid-journey Adobe Firefly and Leonardo and mid Journeys first because it's the best
747. 2:01:10it's the best although you know sdxl is is looking sweet but I assume the [ __ ] that you're playing with that we can't
748. 2:01:17play with is probably pretty mind-blowing uh I'm just learning to use AI very useful any advice for teachers with their students um yeah so
749. 2:01:30um don't hide from it at all um understand that these tools are going to be everywhere I know um
750. 2:01:45so so understand that these tools are going to be everywhere um you know the the thing that the teachers that
751. 2:01:55that I've heard talk about this stuff that I that I think is really interesting what they're doing are
752. 2:02:00teachers that are bringing it into the classroom and increasing the the critical thinking around these tools so
753. 2:02:09rather than saying don't use these tools to write a paper bring it into the classroom write the paper have it write
754. 2:02:15the paper and then have everyone you know dissect it and discuss what it did wrong where did it hallucinate
755. 2:02:25um did it use any weak arguments what would they have done different um you know get into that critical thinking because I I
756. 2:02:33if I sort of crystal ball the future you look at tools like these and you say all of the Tactical execution from
757. 2:02:43drawing to writing to music the Tactical execution of it the grammar the the writing the composition is going to be automated
758. 2:02:56what's not going to be automated or or will take longer to automate is the critical thinking of like why am
759. 2:03:03I writing this who am I writing it for what's my goal like how do I want this to impact people um what does success look like what's
760. 2:03:12quality look like critical thinking critical thinking critical thinking so I feel like the the value of critical
761. 2:03:18thinking goes up dramatically in a world where tactical execution is automated um so I would do everything in your
762. 2:03:26power to increase using the Noggin uh actively all right is there a way to check students writings to see if it
763. 2:03:34came from AI there are there are tools to check it but they have a lot of false positives a lot of false negatives and
764. 2:03:40smart students can get around all of those Checkers they're useless give up on the checkers these tools are going to be everywhere
765. 2:03:46embrace it bring it in instead of trying to push it out because you can't we can't uninvent it and it ain't going
766. 2:03:53away don't use the Checkers in my opinion now if your school district says you have to great um New York City Schools banchat GPT at
767. 2:04:03the beginning of the year and six months later they unbanned it because they get it city of Boston wrote
768. 2:04:09um generative AI guidelines go look at the city of Boston's AI guidelines generative AI guidelines use that as a model to talk to your
769. 2:04:18um your school district about it your school district should probably have a policy on generative AI and not one that's like [ __ ] you
770. 2:04:27generative AI but one that's like we it's here it's coming here are the challenges with it here are the dangers
771. 2:04:34of it here are the good ways to use it city of Boston generative AI guidelines really powerful all right it is midnight
772. 2:04:42I gotta go I gotta leave you people it's really good to be back it's fun hanging out sorry I had to start so late but you
773. 2:04:51know that was what the tick tock said so really good hanging out with you all peace out peace peace peace peace pace
774. 2:04:59uh I appreciate all of you hanging out take care