aiLL April 30, 2023 - The Future of AI is Here Faster Than We Realize

Transcript

1. 0:03hello hello hello I guess I am live it's the AI learning lab turn it on the light here let's see does that help um [Music]
2. 0:23hello hello hello it's AI learning lab Kyle Shannon here talking about things Ai and how yeah a little bit how is everybody
3. 0:37if you've got any questions put them in the chat I might even bring if if you're interested in having a conversation I
4. 0:44can bring people up on stage or just learn that so we can try that if you have any questions about AI that's what this
5. 0:51channel is all about I'm happy to answer any questions um I don't know much but I'm learning so that's what we're looking to do here
6. 1:01hope everyone's well happy what day is today Sunday happy Sunday everyone um getting ready to uh to put some work in
7. 1:11on the everyday AI newsletter and and talk about that one of the things I talked about yesterday was uh how I see
8. 1:24um the current state we're in it with AI as having very significant parallels in um in the early days of the web so in the
9. 1:34early days of the web the internet had been around for decades and then this world wide web thing came along I
10. 1:40discovered it in 1994 and end up ended up starting one of the first online magazines and one of the first digital
11. 1:46agencies and this feels very much like we're in the in the same kind of uh same kind of era where where massive
12. 1:57change is coming not a lot of people know um what or how right now because there's so much excitement there's a lot of
13. 2:06people acting like experts and there are very few of them who actually are so one of the things I try to do on this
14. 2:13channel is is talk very straight about um what I know what I don't know um and uh and just trying to keep up
15. 2:21with everything quite frankly because the you know in comparing where we are with AI to um to the mid 90s with the with the
16. 2:30internet and the World Wide Web um things right now are moving way way dramatically faster and I think the um
17. 2:40the implications even though the implications for the popular internet um are pretty profound um they're
18. 2:46nothing like what's going on here so with that if you have any questions ask me questions let me see if there's any
19. 2:52here I haven't looked yet I someone just got the number one glitter badge good Lord congratulations that's exciting uh no one's asking any
20. 3:00questions if you have any questions let me know I can also bring people up on stage if you want to chat happy to chat about AI other stuff
21. 3:07um you know I could talk about what can I talk about I've started a lot of companies I can talk about that I can
22. 3:12talk about writing uh but but uh the thing this channel is all about is about learning about AI generative Ai and
23. 3:19what's going on so if you've got any questions ask away this is only my second live so I don't know what the
24. 3:24hell I'm doing here I am I am making it up as I go so so um hopefully we can have some fun uh talking about AI on a Sunday Sunday
25. 3:33morning I'm in Denver here so let me know where you are and I'll give you a shout out and what else um
26. 3:42trying to think I had a really good week last week it's taken me about three months to incorporate I just did a video
27. 3:49on this it's taken me about three months to incorporate generative AI tools with my current company which is storyvine it's a it's
28. 3:57an automated uh video creation platform and I finally came up with a vision for how to incorporate all this stuff and
29. 4:05one of the first conversations I had with was with one of our biggest clients which is a big Pharma company and they
30. 4:11got so excited about it they're like fly out look we should talk about this in person um so so that's really exciting uh it's
31. 4:17taken me a while to get there though gpt4 worth it yes absolutely worth it it is it is markedly better at a lot of
32. 4:25things coding reasoning creative writing Nuance then GPT 3.5 is that said GPT 3.5 is very very fast and it's much cheaper
33. 4:38so if you're doing API calls and things like that um there's reasons to use 3.5 or even three um just just because of things like
34. 4:46speed and and cost but if you're just talking about chat GPT is it worth paying you know the 20 bucks I mean there's a couple of things
35. 4:53there's there's an app and a website called po po is the app Poe and po.com poe.com um it's made by uh the company quora the
36. 5:04question and answer company and um there you can use multiple models so you can use GPT 3.5 you can use gpt4 you
37. 5:12can use anthropic's Claude model uh and maybe some others you can also create your own chat bot um I made eight or
38. 5:20nine chat Bots like uh I am Shakespeare if you go to po.com I am Shakespeare you can interact with a chat bot that
39. 5:27responds in iambic pentameter and writes like Shakespeare it's pretty fun um so so Poe is a good one um how can I create an app from the
40. 5:36scratch so so if you're not um a programmer a coder I'm not um I understand code and and I have coded in the past but like way in
41. 5:47the past and and never uh of of any um significance because I have ADD and I just don't have the attention span to really do coding
42. 5:58um but you can use no code or low code tools um to sorry this thing is teaching me how to use it well whatever I can't figure it out
43. 6:10um you could use a low code tool like zapier which allows you to um to string together a set of actions it's an automation tool
44. 6:21um and and with that I've built a number of apps now um where I create a Google form uh ask people a couple of questions in a Google
45. 6:28form I then take that and and incorporate their answers into a prompt send that to open ai's API and get back
46. 6:37um responses and it's it's pretty cool you can email them or you can create a Google doc or something like that so
47. 6:42that's how you can create apps from scratch there's a there's a no code if if you want to create true apps there's
48. 6:48one called Bubble which is still pretty technical you know all of these things are a little janky and Technical even if
49. 6:54they're if they're no code or low code um but but yeah I would just say just start playing with it figure out what you want to do all right
50. 7:03are there best practices I found with prompting um yes many a lot of the things that you'll see are um The Prompt chat gbt to act like a
51. 7:15certain thing so act like a um a coding expert or act like a marketer or act like a you know controller or whatever it might be
52. 7:26and then you can ask it questions uh and and it will it's it's answers will improve dramatically so that that's
53. 7:34that's that's just a little trick but here's here's the biggest thing that I found is that um shifting your mental thinking your
54. 7:44mental thinking shifting um how you think about prompting from what we do with Google right now where Google is kind of this single
55. 7:55um prompt single response you know find me whatever cheap flights and then the list of things comes up
56. 8:03if you treat chat GPT like that it's not as good if if you just say you know I want this and it gives you that if you
57. 8:10shift your mental model to go this is a conversation it's called chat EBT think of it like a conversation and think of
58. 8:17it more like less like this is a tool that's going to give me the exact answers and more like you're working
59. 8:22with like a really smart intern like like the smartest intern you've ever worked with because it essentially knows all of the internet
60. 8:29right so it knows lots and lots of stuff but if you think about it like that so you ask it for something and it doesn't
61. 8:34give you what you want you say hey that's not quite right do it like this oh do it like that you know what I
62. 8:39really want is this can you turn that into a table so so almost treat it like you know a co-worker that you're working
63. 8:46with remotely and um and that makes a big big difference so hope that helps let's see there's nothing to learn
64. 8:55it does everything for you I completely disagree with that it does a lot of things for you but the intention of the
65. 9:01user the intention of the prompter is everything and then the curation of the answers is even um more important
66. 9:09um so so you know when someone shares a piece of artwork on Twitter that they've done I guarantee you that there's been
67. 9:15you know a hundred images that they tossed in the trash uh before that one so um so this is there this is very much
68. 9:22listen these large language models check EPT and these image tools and the video tools and the music tools all this sort
69. 9:30of stuff they're just tools right we get to choose how to use them so the intention of humans and then what we
70. 9:37choose to curate and refine and put out to the world that's um that's ultimately uh what this stuff's about so
71. 9:45um they are incredibly powerful tools that said are there portals in space and time uh if you watch the Skinwalker Ranch show
72. 9:55you would certainly think so there's something above the triangle area um and and you know our large language
73. 10:02model is going to help us solve that and get there I don't know maybe uh let's see where are we here are there portals
74. 10:09is face and time you guys are funny um can you please llm also in these models are independent later layers and what
75. 10:20they do I don't quite know what that means if you could ask that again could I please explain an llm so llm stands for large language model
76. 10:28I'm assuming this is what you're asking um large language model is is what llm stands for and it basically means
77. 10:37a an AI system that uses the Transformer architecture so the the Transformer architecture was invented by Google in 2017
78. 10:48it is the big breakthrough that all of um the kind of modern AI stuff that we're talking about and playing with
79. 10:55right now is based on and basically what it allowed is it allowed for these large for these AI systems these machine
80. 11:02learning algorithms to be trained on massive data sets so so gpt3 was essentially trained on all of the public
81. 11:11internet right so all of Wikipedia all of you know websites that are kind of accessible and freely accessible without
82. 11:19being behind you know passwords and paywalls and things like that um so it is remarkably deep in the
83. 11:26amount of knowledge it's been trained on now um they stopped training at September 2021 so chat GPT is is only good up to
84. 11:35there if you use the GPT chat within Bing that's connected to the internet and there's plugins coming for chatgpt
85. 11:44that are connected to the internet that'll fill that Gap that time Gap um and then um and then the large language model is
86. 11:52basically just that that underlying model that when you're interacting with chat GPT and you ask it a question it's
87. 11:59then you know going into its database of the knowledge of all of the words that it knows um and it's it's figuring out what is
88. 12:08the most likely answer that you want based on your prompt so so that's that's what they are so the llm is the is the
89. 12:14underlying layer and then there are lots of applications going to be built on top of that so you ask about the layers so
90. 12:22at the at the core level you've got um all of these words in context of you know what they're written in all the
91. 12:31meaning of all the words basically goes into this database and then there are applications that sit on top of that so
92. 12:37if you use Jasper or copy AI those are an application layer above the large language model which is GPT and
93. 12:46basically there's a series of prompts that are pre-written for you so if if in Jasper you use the blog writer
94. 12:54um they have prompted chat GPT to to be really good at writing blogs for that particular tool so it's a layer on top
95. 13:01of the core large language model now the large language models there are lots of them last week there were four
96. 13:07of four open source models were were announced or released um there's the Lambda model which came from meta um there's
97. 13:17um uh stability just announced theirs Google's got barred open AI has uh chat or has GPT three and four three three point five and four
98. 13:28so I asked you yesterday but didn't uh what did you ask me yesterday you didn't answer and today too all right hang on
99. 13:37let me go see what you asked me how can I create an app from scratch I did answer that use things like zapier
100. 13:45low code or no code tools like zapier um z-a-p-i-e-r-dot-com or bubble if you want to create an actual application
101. 13:54zapier is an automation tool for creating web-based uh tools uh you might be able to create an actual application with it I think
102. 14:01with bubble you can do actual apps do I think time travel is possible I don't flip a no I have a degree in acting um
103. 14:13I can pin a comment now are there portals in space and time all right ask any questions I think I've got all the questions how hard is it to
104. 14:22build an extra later layer for example um crypto GPT uh if you're a programmer it's not hard at all uh I I think that layering
105. 14:35different tools is is going to be uh incredibly powerful if you look at if you if you haven't um watched videos on or or looked into
106. 14:45these things called autonomous agents you should they're incredibly new so they're very very raw and and they don't
107. 14:53work all that well right now but they're they're very interesting so one of them is called Auto GPT there's another one called Baby AGI
108. 15:03and what these tools do is they stack together different different tools to create powerful things and they also
109. 15:11prompt themselves so you say to Auto GPT um go find me the I don't know the profitability of the top five
110. 15:20um you know Global corporations in 2022 and it will give itself a task list it will say okay go find the revenue you
111. 15:28know go find the top companies um you know buy revenue or whatever and then it will say go go find their
112. 15:35profitability and then it will say make a table with that so it will actually prompt itself and it will figure out do
113. 15:41I need to do this on Google do I need to send this to Jet chat GPT um so a lot of developers are playing with
114. 15:49these autonomous agents right now figuring out what combinations of tools they can put together the other thing that's happening is um
115. 15:57chat GPT has introduced a plug-in architecture so I've got access to the plugins right now they're they're very
116. 16:04very thin and and what I would call very Alpha meaning you can't play with a lot of them they're hard to discover they're
117. 16:11hard to share I think you can only share a plug-in with up to 15 people or something like that but what it allows
118. 16:17is if you have one of these plugins turned on for example um let's say there's there's one of the plugins is for OpenTable you could say
119. 16:26to chat GPT I want to have a good steak dinner on Saturday night you know for a date and it will go find you
120. 16:34a restaurant that serves good steak dinners and even book you a um a table using the OpenTable plug-in
121. 16:43right so so by adding plugins to these models what you're going to see is is lots of combinations that create very
122. 16:50sophisticated applications hope that helps let's see um have I done any movies or television work a little bit I was mostly a stage
123. 17:01actor but um I was I was white boy number two in New Jack City I was I was like a kid from Connecticut buying drugs in Harlem that was fun
124. 17:14it was fun to shoot I also um my best friend Steve session um he and I met filming an episode of America's Most Wanted he was the Killer
125. 17:24and I was the best friend of the Killer and I remember one of my lines was uh Adam about this gun thing right before you shot someone in the
126. 17:34back um let's see how do you generate wait that's nice I remind someone of their homeroom teacher in high school I I
127. 17:46guess that could be a good or bad thing um shouldn't there be more custom data sets data sets like Bloomberg GPT would
128. 17:55that bring more value yes I think private data sets uh are gonna be are going to be where the primary value is
129. 18:02because I think the large language models are going to be ubiquitous they're going to be everywhere and so I
130. 18:07think being able to search on data that you trust and data that is uh specialized to what you want to solve I
131. 18:14think it's going to be quite powerful so I think Bloomberg GPT is is really interesting I was not a teacher ever um foreign
132. 18:27sorry going through here um um how do you generate actually wait how how to do it to generate actually applicable code I assume you
133. 18:42mean chat GPT um if you so so this was one of the first breakthroughs I had um with chat GPT was um I had heard it was good at code and I
134. 18:54said write me a python script that takes an input and does something with it and so it generated all of this code and I'm
135. 19:02like I realized after it generated the code I don't know how to run python code so then I told chat gbt how do you run
136. 19:10python code and it did uh it it said uh if you're on a Mac you go to the terminal and you type in this command
137. 19:20line I'm like I don't want to use command line can can I do it on a website it said oh sure here's some
138. 19:24websites that can run python code and then it told me how to do it paste it in the left-hand window hit the play button
139. 19:30and it'll generate the code so I copied the code pasted it in the window hit the play button and it gave me an error and
140. 19:35I'm like here's the error it gave me and it said well you need to install this library and I said how do you install
141. 19:40that library and it said um you click in the lower left hand corner of the window and then you search for the library and you install it and
142. 19:48within 90 minutes I had written a very sophisticated python application that was functional having never written a line of python in
143. 19:57my life so so basically you just you you use chat EPT to teach you how to generate applicable code and even how to run it
144. 20:07um the question gets mixed in the Transformer I I don't quite know technically how all the stuff works the Transformer
145. 20:14um I don't quite know how it technically works but what happens is the data that you put into the large language model I assume via the
146. 20:23Transformer gets embedded and basically what embedding does is it turns every word into a vector and a vector is think
147. 20:31of it just like an XY chart when you plot a point on an XY chart that's a vector so every word gets plotted in not
148. 20:40two-dimensional space but like thousand-dimensional space it's incredibly complicated mathematical space but all of these words are sort of
149. 20:47plotted in groupings of semantic meaning so if there's multiple meanings for a word it might live in multiple places within all
150. 20:55these word clouds so so when you tell chatgpt to act like a certain thing it's actually kind of narrowing What's called
151. 21:03the latent space it's it's narrowing sort of the word choices or the semantic meaning choices um within this thing so it's fascinating
152. 21:12how it works all right did I teach in school no um see what else we have have I done any movies covered that one
153. 21:21what is embedding and how does that tie into the llm so that's what I just described so embedding is taking a
154. 21:27document or a series of documents or a series of data and encoding them into these it's basically it's wild it's a
155. 21:34string of numbers and the string of numbers represents some position in in What's called the latent space so this multi-dimensional
156. 21:42thousand-dimensional um space of meaning with all the words and then when you search what it's doing
157. 21:50is it's literally looking for what's the next most probable word in the answer based on all of the stuff that was submitted
158. 21:57and it just does that over and over and over again so when you see it respond with a whole paragraph it's literally
159. 22:03you know typing that out word by word just using probability of of what's the most likely word to come next it's wild
160. 22:11the the more you understand this the more remarkable it actually is auto GPT is very good it's it it's new
161. 22:19um but it's it's quite profound what opportunities do you see with AI websites tools and that everyone is
162. 22:27missing and not paying attention to um all of this stuff is quite new um but I think I think the idea that you have to
163. 22:38learn how to code a website that you have to learn how to design a website I think that's going to be a thing of the past relatively quickly
164. 22:48um and and similar with with software certainly simple software more complex software I think is still going to
165. 22:55require a fair amount of engineering but I think we very quickly move into a world where with any kind of knowledge
166. 23:03and creative work I did a video on this called are you a virtuoso or are you a conductor and what I meant by that is if if you
167. 23:12just do one specific thing like I design images or I write you know copy or whatever it is um those individual tasks are going to
168. 23:21get automated very very quickly and so I think the skill moving forward is you know what's the strategy what am I trying to accomplish
169. 23:30um what tools do I need to accomplish that what does good look like what does success look like all of that remains
170. 23:35really important and I think that becomes our role and then what you will do is you will have tools out there
171. 23:42where you say you know design me a landing page or design me a a an application and it will just design it
172. 23:49and then you say make it function like this and it will just do that and if you look at these Auto GPT tools I even heard a concept of
173. 23:58you know the idea of a single use application where you ask the tool to do something it figures out if there's an
174. 24:04app for that if there's not an app for that it will just build you one and then you use it you do your thing and then
175. 24:09that app goes away like that's that's a possibility um so so I think I think website design and building is going to get quite automated
176. 24:18um but but you know how we put them together and what we get them to do that remains you know we're using these tools
177. 24:25so in terms of specific tools I don't know it's things are changing so fast right now I would say just if you're
178. 24:32interested in website building tools just start go to futurepedia.io um and and look for tools that are doing
179. 24:40web development stuff and just start playing but they're changing so fast that I wouldn't get too married to
180. 24:45anything do I know Steven from zero shoes yes in fact he was the guy that I met he and I met on an episode of
181. 24:53America's Most Wanted where he played the Killer and I played his best friend so yeah we've been and and what was cool
182. 24:58was we filmed that in um in Connecticut but we were both living in Manhattan at the time and it turned out we lived on the same street one
183. 25:07block from each other so we became good friends and we've been friends ever since he's actually one of the big
184. 25:12reasons why I moved out to Boulder and now I'm living in Denver all right can I ask a question about AI image Generations sure ask away
185. 25:23um what's your background since you know so much about AI programmer no I have a background in acting and storytelling
186. 25:31um but in the mid 90s I um I was doing desktop publishing as my day job and I discovered this new thing in 1994 new to me
187. 25:41um called the world wide web and I started one of the first online magazines an art and culture magazine that won a lot of awards and then I
188. 25:49co-founded agency.com which was one of the first digital agencies so I learned a lot about um storytelling at scale about coding
189. 25:58about innovating about strategy and how you take a big company like British Airways and and translate that brand into an online
190. 26:07environment in fact we built British Airways first ticketing system we built serious satellite radios first music
191. 26:14streaming system like this was we were developing websites back in the time where none of that stuff existed it wasn't really possible
192. 26:22um so so I kind of earned my chops just building stuff for clients uh uh what are some good AI sites to help start
193. 26:31to start an online business uh if you haven't played with it yet um check out GPT boss that's from McKinsey Bose I think is his last name
194. 26:40he's got a tick tock Channel um go to GPT boss uh The Tick Tock Channel and you'll see that um he's a great guy and GPT boss is
195. 26:49specifically designed to be a series of virtual employees that help you start a business so there's you know
196. 26:56you hire these different functions so you can hire an accountant and you can hire a marketing strategist and you can
197. 27:01hire a copywriter and and basically these are just little pre-prompted personas that specialize in certain
198. 27:09things so if you want you know mental health coaching for for running your startup which you will probably need
199. 27:17um you basically just ask the you know ask your questions to the to the psychologist there so that's a fun one
200. 27:22you can start with but but there's lots I would say you know just using chat GPT think of it like a business partner or a co-founder
201. 27:32might be an interesting way to start a business um how can you train GPT on your data set that's a huge one I think it's
202. 27:40incredibly important I think you know Apple's got a big announcement coming in June I have a sneaking suspicion that
203. 27:48apple is going to be all about um using i a i on your own data um I might be wrong but I think I think using AI on your own data sets either
204. 28:02your personal data set or your business data set um is going to be where the real power is in this stuff moving forward right now
205. 28:11if you want to do that there's a thing out there called Lang chain l-a-n-g like language chain that chains together a
206. 28:18bunch of things and what you do with Lang chain is you can point let's say you've got a directory of PDFs or you've
207. 28:25got a directory of Word documents you sort of Point Lang chain at that directory and then you have it use open
208. 28:33ai's embedding engine so you embed all of those documents you store all those vectors from the embedded documents in a
209. 28:42vector database like pine cone and then you can basically chat GPT search on your own data so Lang chain is a way to
210. 28:50do that there there are lots of tools Now sort of combining that with other things like Auto GP Auto GPT and baby
211. 28:58AGI incorporate Lang chain so what's happening with the developing Community right now is they're starting to combine
212. 29:07lots of different tools to create increasingly sophisticated capabilities so so that's a way to do that
213. 29:13oh Steven is awesome and a great friend yeah Steve's great I haven't talked to him in a bit let's see um what are my go-to AI apps I use
214. 29:24mid-journey mostly for images although dream Studio from stabilities good um I did the the image that you see on
215. 29:34screen right there is from stable diffusion so I set up a stable diffusion collab thing so I was kind of running it
216. 29:42running it in a virtual machine um for the for the language stuff I use open ai's playground or chat GPT mostly
217. 29:52chat GPT I would say any anything where I'm trying to solve a problem I'm generally doing it in in chat GPT
218. 30:00um if I want to codify things and make a tool I use zapier which is low code it's a little janky like zapier's feels a
219. 30:09little dated to me there's one that I haven't tried yet but looks pretty good called make um make dot something make dot IO maybe
220. 30:16but it's an automation tool you can you can find it by Googling it um and those are kind of the biggies right now um
221. 30:25uh you know if if I if I need to create background music for for videos moving forward I think I'm going to use sound
222. 30:33draw which if you haven't seen sound draw dot IO that's pretty interesting for music um none of the video stuff right now is
223. 30:40is all that interesting you can only do three second clips and and they're kind of like the video stuff right now is
224. 30:45kind of where images were a year ago they're not quite good enough to they're just weird if you've seen that there's a
225. 30:50pizza Commercial someone made using AI you know text to video stuff and it's kind of hilarious but a year from now
226. 30:57it's not going to be hilarious it's going to be you know very um very useful why does everyone insist on pushing
227. 31:05forward with AI That's a great question so AI has been progressing um I would say you know incrementally for
228. 31:18um a few decades a couple of decades a few decades probably a few decades machine learning's been around for for a
229. 31:23long long time and neural networks I forget when they were created maybe the 60s but but it's it's been going for a while and then in 2020
230. 31:34um gpt3 came out and gpt3 was the first of these large language models to really start you know doing something good and
231. 31:42and shortly after that um stable diffusion image generation got good and mid-journey started getting
232. 31:49good and we started hearing about all of these um these image generation tools which went from kind of comically bad to really
233. 31:58really good in a very short amount of time and then um November 30th um 2022 is the date that in my opinion changed
234. 32:10everything and that's the date that chat GPT was released why did that change everything well chechi PT itself wasn't
235. 32:19all that interesting actually from a technical perspective in fact this is this is kind of a cool thing the instructions for how to create
236. 32:28chatgpt we're actually in the documentation of gpt3's documentation for from open AI so so the developers
237. 32:38um thank you for the hat that was lovely um the the developers at open AI basically gave you the instructions for
238. 32:45how to create chat GPT and Sam Altman said we assume someone would make it and no one made it so so I think about a month before
239. 32:53Microsoft was going to make their big announcement they that they invested a billion dollars in open AI
240. 32:59um Sam Altman said you've got two weeks go make chat GPT and so they built that thing in two weeks um to to put the impact of chat GPT in
241. 33:10perspective so the World Wide Web which I was I was you know a part of the early days of the World Wide Web
242. 33:16to get to a hundred million users it took eight years uh for the World Wide Web to get to 100 million users chatgpt got to a 100
243. 33:26million users in um six weeks month and a half 100 million users so so because of that a lot of things have happened there's all these
244. 33:38cascading effects so because so many people used it like it just hit the cultural Zeitgeist right all of a sudden
245. 33:43everyone was talking about it so this is Tech that had been out you could play with gpt3 for a year and a half before
246. 33:49that I played with it I was like yeah this is cool but there was something about chat EPT that just caught the
247. 33:54imagination and then because it got to 100 million users so quickly it forced the hand of everyone else Google
248. 34:00invented the technology that is underneath chat EPT but they didn't release it because they knew if they released it it would threaten their
249. 34:08search business well they were forced out of the Shadows um Amazon has now been forced out of the Shadows meta has now been forced out of
250. 34:15the Shadows apple is kind of on the verge of making a big announcement so all of these companies that that had a
251. 34:23lot of AI technology nice hat thank you um a lot of these companies that had this technology were kind of keeping it
252. 34:31behind the firewall chat gbt forced it in the open so that that was one thing the other thing was this the developer Community because it it
253. 34:39you know had all of this Buzz about it the developer Community right now it's it's all they're focused on um so like
254. 34:47Auto GPT came out that was this thing that strings together these actions it got to I think it's a hundred thousand
255. 34:53Yeah a hundred thousand likes or stars or parts or whatever on GitHub um it did that like in two weeks right
256. 35:00so so incredibly quick adoption and what I'm seeing very consistently is people are um someone will put a paper out and
257. 35:10within two or three days someone will launch a Technology based on that paper or they'll put another paper out that
258. 35:15incorporates that paper into two other papers and they'll create this new hybrid kind of thing so the amount of
259. 35:21development happening right now is insane I've never seen anything like it so um I don't know that anyone is insisting
260. 35:29on AI pushing forward I think it has just broken through in a way that means it's not going away and the acceleration
261. 35:36is not going to slow down the call for a six-month pause on on AI Innovation from Elon Musk and others is
262. 35:44interesting I don't think it's got a chance of happening um I also I I view it a little cynically with musk in particular a week after he
263. 35:52called for a six-month pause he announced the creation of x dot AI his own AI company to compete with openai so
264. 35:59he helped fund open AI he doesn't like the direction it's going so he's going to compete with it so I was a little
265. 36:05cynical about that six-month pause anyway I don't think it's got a a realistic chance in hell um are there issues that need to be dealt
266. 36:15with that a six-month pause might help with absolutely there's lots of issues that need to be dealt with but there's no stopping it
267. 36:23um let's see their xlab 5000 is a no code solution for pretty much everything I don't know who there is let me know who that is I'll I'll
268. 36:37say that where are we here isn't it dangerous for our economy and our way of life uh I don't know if it's dangerous but it is going to be
269. 36:48disruptive on a scale that I don't think we can imagine right now and it's coming way faster than we think
270. 36:53I think short term it's going to be destructive but I think I think um I think long term it's it's potentially
271. 37:03interesting I just think lots of things change a lot of our economy right now is based on complex systems everything's
272. 37:09really complicated what these AI tools is going to allow us to simplify things and make things more efficient and so
273. 37:16it's going to eliminate some jobs that are are there to sort of do the complicated stuff um but it's gonna it's gonna create new
274. 37:24jobs that we can't even imagine what they are yet um so but I but I do think the next two or three years is going to be very very disruptive
275. 37:32um you know I'm I'm personally very excited I it's funny the more I use these these tools that you know in
276. 37:40theory automate creativity or automate knowledge work I find myself being more creatively engaged than I've ever been
277. 37:49um because what what the tools allow me to do is operate at kind of a higher creative level where I'm just thinking about the
278. 37:57Creative Concepts and the tools are doing the execution so before if I had an idea I would go to do something and
279. 38:05I'd go like oh I need to use a software tool and then I'd have to go you know sign up for that tool and then
280. 38:12I'd have to put in my credit card and I and I have to do this and then I'd have to go watch a YouTube video to learn the
281. 38:17tool and then I'd realize the tool didn't do what I wanted and 45 minutes later I don't even know what my creative
282. 38:22idea was anymore what's happening with the AI stuff is I'll have kind of a vapor of an idea and within five minutes
283. 38:28I can sort of bring that thing to life and so that gives me another idea so I find myself kind of executing at a much higher
284. 38:36um I don't know strategic creative level with these tools so as opposed to it making me less creative I find it making
285. 38:43me more creative so I think there's a silver lining to it but it is absolutely going to be disruptive for I I I can't think of
286. 38:53um a profession right now that isn't going to be impacted let's see what do I think is dangerous there may be risks but oh oh well I mean
287. 39:04the risks are so like the the auto GPT thing like you know someone could very very easily without them not without a
288. 39:12lot of technical knowledge right now could use something like Auto GPT to say go find all the systems you can that
289. 39:19have this particular exploit exploit it get me Rude access to those servers and then send me a list of servers I have root access to
290. 39:27stuff like that so so so but you know the the the white hat black hat um battle between good and evil that's
291. 39:36going to continue they just they're just gonna have very uh much stronger tools now I think the more people that get
292. 39:42curious about Ai and start using it for good I think it's going to have a disproportionately positive impact in
293. 39:47the future so if you're not curious and digging into AI right now I strongly encourage it let's see if I'm gonna laugh search for our AI
294. 40:01generation images of people eating spaghetti it's the best I know the hands are hilarious snails for some reason it
295. 40:07can't do snails very well snails are hilarious um all this stuff will get better but I I think I think we're gonna
296. 40:13look back at this era of image generation and you'll be able to know oh that was a 2022 image because you know
297. 40:18it's got 27 fingers on on a hand and things like that um I think we'll we will look back lovingly at this era the internet was
298. 40:28disruptive we just hardly remember how much this is the same just way more powerful totally agree with that um let's see okay
299. 40:43um yeah I totally your your name is whisper in whisper Aaron I don't know um yeah I agree I I think that the when I first got exposed to the
300. 40:55internet 94 was when I discovered the the web um it was very easy for me to see like what what the future would look like because
301. 41:04it was it was like you know okay everything's instantly connected geography um kind of collapses time collapses like
302. 41:12everything's kind of instant like you kind of see where it was going and you could kind of you could kind of predict
303. 41:17what the implications would be with this AI stuff I'm having a much more difficult time figuring it out because it's so profound right it's
304. 41:25you know it's not just we ask a tool to do something and it does something it's like these tools are you know in a lot
305. 41:34of ways even in their current state more capable than us like I'm not good at creating images um you know with Photoshop and illustrator
306. 41:44um I am now right I can now create world-class imagery um I don't have the attention span to to do much writing because I have ADHD and
307. 41:54whatever I do now right I can I can whip up an outline of something I can flesh out that outline I can turn that into a
308. 42:02written piece and edit it um in 15 minutes right where historically that would have taken me three hours and because of my 80d it
309. 42:10would have taken me three weeks um so so um so this stuff is way more profound and seeing what the future looks looks
310. 42:20like is is very very difficult so yeah I think that's a huge point um have I named my AI bot yet mine is
311. 42:28named Dinah I created a bunch of bots on Poe if you go to poe.com I am Shakespeare I am po um I am Steve Jobs I created a Steve Job
312. 42:41spot so I created a bunch of lots of historical fig figures I did Virginia Woolf I did um Emily Dickinson I did
313. 42:50who else did I do I don't know I did a book so I can go show you let's see let me go to here I am going to go to poe.com
314. 43:05and let me switch these and then I gotta log in hold please okay so I've got all these bots so like
315. 43:19here's I am Shakespeare so let me reset this it says Hail to the I don't know if you can see that it's a little hard to read but
316. 43:28whatever hopefully it focuses a little it says Hail to the I The Phantom Bard of Avon stand to serve thee well right at thy command so I say
317. 43:38um tell me about the beauty of Quan quantum mechanics and then it writes me in iambic
318. 43:53pentameter quantum mechanics doth hold Mysteries profound it's truths to be unraveled yet Unbound it scales both so so this is one where
319. 44:03you can talk to Shakespeare I've got if you go to poe.com I am Shakespeare you can also go to Steve Jobs so here's
320. 44:10Steve Jobs who kind of coaches you how to be be a business person I did Jay-Z so you can you can uh have Jay-Z right you write you a flow
321. 44:21um I did uh Emily Dickinson so poe.com Emily Dickinson um poe.com Maya Angelou Mary Shelley Virginia Woolf so those are kind of fun [Music]
322. 44:34um okay cool let's see great comment I don't know what my great comment was I use it for an idea and
323. 44:45then just ask AI to expand on it enhance it jump starts my creative yeah yeah I like this is the thing that most
324. 44:52excites me about these tools right now is you can go from just having an idea to something really fleshed out in like
325. 44:58five or ten minutes and it you can kind of figure out if it was a good idea or not and and it will often Inspire lots
326. 45:05of other ideas it's it when you start thinking about chat GPT in particular as a creative collaborator I think the image tools as well um
327. 45:17where you're actually kicking ideas back and forth with it that's everything changed for me when I started thinking
328. 45:23of it like that not like a Google search engine where I get the answer but more like I'm having a back and forth
329. 45:29um it gets it gets really it's really powerful when you start thinking about it like that um you're welcome I'm I'm glad me
330. 45:36blabbing about this stuff is valuable um do I have any tips with python python and GPT API integration
331. 45:45um I wrote a very sophisticated python application it was it was my breakthrough moment of what chat GPT
332. 45:51made possible I still haven't written a line of python in my life I don't know how to write python I don't particularly
333. 45:56have interest in learning it because I think all this stuff's going to go under the hood very very quickly
334. 46:02um I'm more interested in what do you do with it so what I've been using is zapier the automation tool where I'll create like a
335. 46:10Google form is my input and then I take the input from that form I roll that into a prompt that I then send out to
336. 46:17open AI to to GPT 3.5 or whatever I'll do that multiple times if I need to I then take the responses throw them in an
337. 46:26email or throw them in a Google doc and that that's how I've been using it so um I would say my tips with python and GPT
338. 46:34API integration is have chat GPT be your um your engineering partner um I did a video there's a cool tip that
339. 46:43I saw that you can take the readme file from GitHub from so so you go to a GitHub repo that you want to do
340. 46:51something with and maybe you're not as advanced technically like I'm not you can take the readme file and
341. 46:57basically start at the table of contents of the readme file you know down through it paste that into GPT and then say I'm
342. 47:04on a Macintosh please give me step-by-step instructions how to you know X you know turn like install this
343. 47:12GitHub repo and then you could even say you know give it to me one step at a time and don't move on to the next step
344. 47:19step until I've said okay and it'll give you the first step and then you go do that and if you have questions you ask
345. 47:25it the questions and you it'll literally walk you step by step through installing a GitHub repo so I've done that a couple
346. 47:31of times so that that might be a tip to use is use chat EPT as your coding partner um or your engineering partner because
347. 47:39you can ask it questions I mean I ask it really basic questions like what does this acronym stand for and you know what
348. 47:45does this tool do and then it explains it and that helps me get my head around what's actually going on um
349. 47:52do I have a link page where all of these useful links are collected I do not yet I should but I'm a loser um go to if if uh
350. 48:05let me see wait let me throw up a a presentation that I just gave um let's see so feel free to screenshot this so
351. 48:21um come on focus focus focus why is that not focusing this drives me crazy all right there you go uh let's see
352. 48:34I'm in the way here hold let's see oh God damn it I have no idea why this isn't focusing oh wait maybe I do is that the reason
353. 48:48it's kind of useless isn't it sorry for wasting your time um well I'll read them out so on the on the words thing
354. 48:58um chat EPT open AI playground um Bing with the with the new chat is gpt4 and then Poe is is another one on
355. 49:07images mid-journey stable diffusion I know this is really driving me crazy focusing anyone have any ideas that's a little better
356. 49:22um the other thing is prompt crafting there's a thing called prompt book which is really really good for image prompt
357. 49:29crafting and there's also the bottom thing awesome chat GPT prompts is really good um and then in terms of audio synthesis
358. 49:39uh and and animation stuff did Synthesia 11 Labs uh are all really good and you should be playing with those if you're
359. 49:48interested in animation Flawless AI That's going to do some really remarkable stuff it's just demoware
360. 49:54right now but I think what they're doing is looks really remarkable um on on Music Creation and video editing there's just a ton of things
361. 50:04going on there so go to futurepedia.com and that will um that will get you somewhere so anyway I hope all that stuff helps
362. 50:15let me flip back over to this thing um yeah so uh yeah I should put together a site with with kind of a the problem with putting
363. 50:27together a site of uh hey Mac how's it going I was actually just talking about you earlier someone asked about um
364. 50:35what did they ask you about oh you know how do I how do I start a business with um with AI I said you go to GPT boss and
365. 50:43let Mackenzie figure it out for you just hire all of his dudes uh um all right let's see um I should put together is there an AI
366. 50:58app that helps you focus your camera no like what the what the hell I I don't I honestly don't understand why it
367. 51:05can't focus it's making me not yeah it's just weird I don't know it's obviously seeing something maybe something in the
368. 51:14distance I don't know I don't know I nothing yeah AI needs to help with that there's some of the core technology that
369. 51:22we have to deal with that just sucks um any other questions [Music] I think that's it um I can wrap this up here I think I have a
370. 51:35newsletter I gotta go make but um yeah I'll keep doing these if this is a value any link for creating a
371. 51:43trading bot stocks crypto so I've seen a bunch of I haven't done this myself but I've seen a bunch of of tick tock videos of um
372. 51:53I've done the beginning part of this go to chat GPT I would say go to gpt4 because it's just better at this
373. 51:59um if you've got it but even even just regular chat EBT go to chat EBT um tell it to recommend um stock trading sites where you can do
374. 52:10paper trading which is basically you can do trading without spending real money um that that also have Bots that that
375. 52:17have you know automation within them and then um it will know what those Bots are programmed in so you can do one of two
376. 52:26things you can you can have it recommend different trading models and then you can go find a bot that has that you can
377. 52:33copy that code paste it into chatgpt and then say write me a bot that does something else or you can just say
378. 52:41um you know explain to me the different um I don't know trading strategies that Bots use and then find one that you find interesting
379. 52:49have it write you a bot and then go to that site that it recommends and sort of paste it in the bot maker and and start
380. 52:56playing with with fake money and I like I think within an hour you could probably build one and have it up and running
381. 53:03um and like the ones I've seen that are kind of interesting is they'll take two different models that that they they
382. 53:09like this model and that model and they'll sort of put them both in until chat GPT to combine it into a single
383. 53:14into a single model or bot that does it uh and then it will do that pretty remarkable stuff would I rather gain muscles or green
384. 53:25gain ideas uh I am not a muscle guy so igiers I suppose I'm I'm a I'm a food guy not a muscle guy any link to creating a training bot uh
385. 53:40let's see people I think AGI is already here behind the scenes I don't know if AGI is already here yet
386. 53:48um it might be but I you know I think I think some of the auto GPT some of these autonomous agents are
387. 53:58pretty interesting if you haven't seen there's there's a paper out right now called um age um autonomous agents yeah it's called
388. 54:07autonomous agents um and you know the little NPC characters in video games so NPC are non-player characters so like when you
389. 54:15go into the into the tavern and there's the barkeep and the barkeep you know has their whatever their seven lines that
390. 54:22they're programmed to say well this autonomous agents paper they basically um created these characters that use Chachi
391. 54:31PT to interact with one another with with one another and they have memory and they have they have jobs so like an
392. 54:38artist will get up they'll make themselves breakfast they'll go start doing art when they interact with someone with one of these other
393. 54:45characters they'll start a conversation and they'll remember that conversation so one of the things in the paper was
394. 54:52um one of the characters learned it was one of the other characters birthday and they planned a birthday party for that
395. 54:58character so 30 other characters were talking about the uh the uh the birthday of this one character so so I don't know if we're at
396. 55:07AGI yet but but we're starting to get to the point that these things start to look and act more human sorry the dog's barking um
397. 55:15is the mailman here someone's here all right um oh good I'm glad I made your drive go a little bit quicker I know I'm just kind
398. 55:26of rambling here but you know I think that's what these things are would I rather watch a tree go grow or a negro
399. 55:33um I don't have the patience for either I think knees are more interesting but trees let us breathe so let's go with
400. 55:40trees how should someone with no experience learn about AI this is a great one so I did a video you can go watch it
401. 55:47um called the the 2023 smart challenge and and here's the challenge I think figuring out how to use these tools in
402. 55:56the abstract is really hard um but where they start to make more sense is where you actually apply them to solve a problem so the 2023 smart
403. 56:05challenge I put out was come up with two projects um one personal so maybe it's like I don't know put together a
404. 56:16um exercise plan or of meal planning or something like that uh and then one for someone else so either your work or your community or
405. 56:27your family you know plan a birthday party or plan an outing or or solve a problem for your community center whatever it might be
406. 56:37um and then make at least 10 images with an image generating tool like you know mid Journey or stable diffusion or any one
407. 56:44of these tools so so just understand what it's like to make those images and then use chat EPT to solve a problem
408. 56:51um and I think that's the that that's the best way to get in is just pick a specific project and start using chatgpt to solve that
409. 57:01problem and you know you might like the the challenge you might come up with is I want to write a a kid a children's book from a kid
410. 57:09um and then you you'll find out really quickly oh you can run you can write a children's book and then you say oh I
411. 57:14want images for that and you go find an image tool and you make some images for it in doing a specific project with these
412. 57:21tools you'll learn a lot about what these AI things are and where your assumptions about what they are will change dramatically
413. 57:29um I think that's the best way to do it and then and then once you do that once it's it's kind of like crack it'll just
414. 57:35it'll just hook you it's it's very addictive uh because they're so powerful it's um Sam Altman described these AI tools as um
415. 57:43putting magic in the hands of people to see what they'll do with it and that's what it feels like and I've been you
416. 57:48know I've been kind of deep in this stuff for five or six months now and I just it still feels like magic to me it's remarkable
417. 57:58um is there any GPT that doesn't have a limited input of data now limited to September 2021 now the base models the base model stopped there
418. 58:08because creating the base models is very very expensive I think gpt3 cost them five million dollars or something like that so it's very very
419. 58:16um expensive to build that base model um but but with with Vector data databases and the ability to call like
420. 58:24like combine Google searches with these base models you can kind of fill the gaps um so if you use if you go to bing.com
421. 58:32and they have a chat section there thatchat bing.com is gpt4 so gpt4 is the underlying model but Bing is connected
422. 58:41to the internet so you can use Bing's um uh chat service um and it will fill in those gaps the other thing that we'll see is we'll see
423. 58:52um specific data sets like there's a thing out called Bloomberg GPT where it's a large language model but it's
424. 58:58been trained on fine-tuned on all of Bloomberg's specific you know financial data so I think you'll start to see that as well
425. 59:07all right I think I'm gonna uh wind it up here I gotta go do some work I gotta put out a an everyday AI newsletter
426. 59:14thank you all for your time and uh I'll be back at some point but take care talk soon