aiLL 5/31/23 - Demystifying AI Tools at Your Fingertips

Transcript

1. 0:01foreign
2. 0:35sorry it's getting things set up again all right I reset my Wi-Fi router I'm back live again the hair is being special
3. 0:43um Kyle Shannon with the AI learning lab we're getting just getting going live again so welcome back I got ice water
4. 0:52resetting a router is hard um uh this is the AI learning lab if you've got questions about AI um drop me a uh a comment and I will do
5. 1:03what I can welcome back for those of you that were at openai.com that's the official chat EPT website if you have not played with
6. 1:12it you should absolutely play with it Go Tell it to write you a poem or write you a song greetings from Minneapolis hello Minneapolis
7. 1:21um also if you want to geek out on lots of different tools futurepedia.io I don't have any affiliation if you've got
8. 1:27questions about AI pop those in the comments uh hello everybody Welcome um I am a uh I'm just a guy I'm a I'm an
9. 1:37entrepreneur I'm a Founder uh and CEO of a video technology company I've been deep diving in AI for about the past
10. 1:45year and I started this Channel and I started an AI salon and I started an AI art project and I'm now completely reimagining my business
11. 1:55uh uh on AI and and uh it's every air the the more I learn about this stuff the more profound I think it is so
12. 2:04that's why I'm doing this that's who I am I'm happy to answer any questions I'm happy to chit chat um if you love the idea of a salon just
13. 2:13join if you go to the salon.ai there's a link tree there so we've got a meet up we've got we do meetings every two weeks
14. 2:20we've got a Discord server where you can geek out and you can do mid Journey images and ask questions and and you
15. 2:28know meet meet cool people that are doing things um so so definitely come join and uh it's a cool Community it's really neat so um
16. 2:40yeah so that's that um I try to keep up on the news although you know it's the stuff is moving so fast right now that if I take like a day
17. 2:50off like today I actually worked on work work you know like for my company instead of geeking out on AI and I feel
18. 2:57like oh my God what did I miss and it's a lot like there's just there's a lot every day right now uh the connection is
19. 3:03much better great yeah I reset the router so thank you for that I appreciate the feedback Emilio's wife
20. 3:09hello I'm back I I'm out of detention I was in detention for a while oh there's a whole bunch more people here now um if
21. 3:17you're new to this my name is Kyle Shannon this is the AI learning lab I talk a lot about AI here um that site is the official chat
22. 3:26website if you have not played with it just go there right now and while you listen to me ramble have it write a song for your
23. 3:33significant other or about your dog or or or whatever um ask it to write you a python application that my big Insight with chatgpt was
24. 3:43um asking it to program something for me um have it reformat some HTML have it analyze some code have it I don't know
25. 3:50it's just remarkable and so that's that and then futurepedia.io is just a site with lots of uh it's just a list of
26. 3:59different AI tools so so okay yeah I get put in timeout regularly and I don't know why because I've never read the
27. 4:07community guidelines but apparently I violate them a lot I think actually last time I got booted when someone asked me a question about
28. 4:16um bias in the models and I was explaining how the bias gets in there and how to deal with it and it just
29. 4:21booted me so I think the minute you start talking about polarizing issues it decides if it doesn't like the issue
30. 4:29you're talking about it says okay you're gone so I think that's what happened um so all right connection's much better that's great if
31. 4:43you have questions about AI pop those in the in the things below um I can I can do little show and tells
32. 4:51you know what I might do is I might pop open Photoshop beta and I might show you guys some of the generative fill stuff if you haven't
33. 5:05seen that stuff yet it's pretty pretty remarkable is this the beta yeah that's the beta um I think we could probably do it in
34. 5:13there couldn't we let's see mm-hmm [Music] go to layers where's layers
35. 5:36all right so I'm going to turn off the copy and then so the way this works let me grab the change my canvas size here
36. 5:54we'll go width we'll go a thousand oops thousand height will go 1400 all right so so now I've got this this
37. 6:11cool image that I made in stable diffusion that's sitting in the center of this big canvas and what you can do now is you
38. 6:21can just start to extend the canvas so if I just select the area to the right and the area to the left it now has it now pops up this little
39. 6:37um contextual menu it's this little contextual menu and it says generative fill and if I hit that I have to agree to their terms if I hit
40. 6:48that I can type in something or I can just hit generate and so I'm going to hit generate now let me try to get this so you can see it
41. 6:56get my fat neck out of the way all right so there's the image so I'm going to hit generate so now it's thinking
42. 7:10all right it's doing all that stuff and it didn't create anything I didn't create anything what is going on hold please
43. 7:22oh I might have been on the wrong layer I know it's on the wrong layer I'm a dum-dum I'm a dum-dum I don't know
44. 7:29if you knew this but I'm a dumb dumb okay hold please uh do I want to delete generative layer one yes I do I'm gonna
45. 7:37go to now I'm going to go down here okay sorry sorry you know I'm I'm photoshopping live here people what do you want from me what do you
46. 7:46want from me I'm trying to show you some cool stuff okay so I'm selecting the area to the right I'm shift selecting
47. 7:53the area to the left and now I'm going to say generative fill generate and you can extend the image automatically too I don't know how to do
48. 8:04that that's cool though all right let me see extend the image that's yeah that's what I'm doing oh wait I didn't hit generate again okay
49. 8:14all right there's our little progress bar now we're going this is going to be so good it's gonna be so worth the wait
50. 8:20I just went from 150 viewers to 50 people like what the hell is this this guy just why is it not showing my oh
51. 8:29it's not working now I don't know how to I can't do this live why is it not working why is it not working huh
52. 8:41I do not know why it's not working [Music] well anyway all right well that was one of those you know you know those things
53. 8:55where people are like oh I'm gonna show you this cool thing and then they just do what I just did yeah that was nice all right well whatever
54. 9:04um okay so if you're new here I will answer questions I can about um AI apparently my Photoshop beta is
55. 9:12not working for some reason it's that's very very weird I don't know what was going on with that what you doing I'm
56. 9:17just hanging out so if you if you're new here go to that URL that's chat GPT if you haven't played with chat GPT start that
57. 9:25um can you pull a photo out wait can you pull a photo out of a video using Photoshop yeah you can just do a
58. 9:34freeze frame within a video right I I just do screen captures I just pop things open in QuickTime and just do a screenshot
59. 9:42do I think a person or AI booted you I have no idea uh uh I assume it's the algorithm um do you think the excitement over llms
60. 9:53will lead to neglect in research of other methods hmm that's a good question I think llms are here to stay I think that
61. 10:09I I mean it's it's hard to know you know where things are going but I think the future is multimodal and and if you look
62. 10:17at the plug-in architecture of Chach EBT um right now these tools feel very single purpose right like touch ebts for words
63. 10:26and mid Journeys for images and photoshops for images and we're going to be moving into a much more multimodal
64. 10:33kind of future where you just ask it to do stuff and it will figure out do I need to put a plug-in here do I need to
65. 10:41make an image do I need to make a video does this person want to hear music do they want me to write them a symphony
66. 10:48do do they want me to print out the sheet music of the symphony right so I I think I think we're we're
67. 10:56I kind of feel like AI for for regular people which let's call the you know November 30th is when of of 22 is when chat GPT came out
68. 11:09this marks the the turning point where regular people get access to these incredible incredibly powerful tools and
69. 11:16and you know have kind of adopted them overnight um I think we've got training wheels right now and I think we're just kind of learning
70. 11:24what these things are and what what's even possible um I think things are going to get very transformative very quickly um so
71. 11:35I suppose focusing on large language models is gonna it might hamper other activities but if you look at if you look at what's
72. 11:46happening on GitHub every time a new concept comes out white paper comes out technology comes out it gets adopted by
73. 11:54by the the dev Community like that and you you're just there's just an explosion right now in Innovation and
74. 12:02some of it's really crappy and just thrown together but some of it's really exciting and you're seeing people kind
75. 12:07of mixing and matching the piece Parts as they come out so um I I don't I don't think so have I messed with any llms mixed with
76. 12:16extensions via uba uba Booga I have not um I don't know what that is but I'll check it out my thing is it forgets
77. 12:27things you set as non-variables uh Chachi BT does yeah well it's got a three thousand um word or a 4 000 token context window so
78. 12:39if you do anything longer than that it forgets stuff off the top there are going to be solves for that and and in
79. 12:46fact one of the solves is if you go to um po.com you have to pay for this but you get access to all these different
80. 13:01models right so the sage is there's gpt4 Claude Plus and then Claude instant 100K what that means is that's a hundred thousand token model
81. 13:13so you can you can put into your chat window 75 000 words you could literally copy and paste a long novel and start
82. 13:23asking questions on it or in your case if you if you're doing things where it's for getting context um you could uh you could use that for
83. 13:32you know for for better memory but you know there's things like Lang chain and and things where you can you can build
84. 13:40memory so so that it's remembering the conversations um but that's going to get better there's going to be tools that come out
85. 13:48um I don't know three months six months um uh what do I think about the royalty license based Falcon 40 billion large
86. 13:58language model outperforming llama 65 billion well I think that [Music] um I I don't know the royalty license based
87. 14:08one but I assume that means that um artists or or or data that it was trained on they get royalties or they
88. 14:16get or they get attribution which would be kind of cool um if it doesn't mean that then then that's just my ignorance of that but the
89. 14:23fact that you you're seeing smaller large language models outperforming larger ones um I think we're going to continue to see
90. 14:34that and that's there was a there was a Google there was a called we have no moat we have no moat Google we have no Moats
91. 14:49so this article so if you if you just Google we have no mode this this is an article let me see if I can get Focus to go
92. 14:56this is an article that that it's it's an engineer from Google this is not an official opinion but what he basically says here is that
93. 15:04um for a hundred bucks with with a with a small language model um you know scrappy little developers are creating
94. 15:13large language models that are that are as as good or close to as good as what Google and openai are spending millions of dollars producing
95. 15:23um and and so he says he thinks that's going to continue to happen and there's no real competitive Advantage for them
96. 15:30to to be doing these big models I don't know that that's necessarily true um because like Google owns all the data
97. 15:39right so I think Google's going to be able to to make these things Google's also very very aggressively starting to
98. 15:45fight back against open AI it's it's ironic that they're having to fight back against open AI since they invented the
99. 15:51technology that is the is the sort of key little nugget that's making all these things work this well um but but they are they kind of
100. 16:00mothballed something that was really cool um and then now they're now they're struggling to to catch up um just had chat EPT write me an SNL
101. 16:11Monologue for George Carlin Best Line Comedy is just tragedy plus time oh that's cool I think that's something that he actually said
102. 16:18I'm a big George Carlin fan I am using experimental Google Bart 2 this is much more advanced oh cool oh wait
103. 16:31say I missed whatever you said before that oh agent GPT can overthink about questions you ask it by itself yeah the the the autonomous
104. 16:41agents so so what she's talking about there is there's there's these things called autonomous agents one's called
105. 16:48Auto GPT there's another one called Baby AGI I'm sure there's a pile of them out there now that um rather than you having
106. 16:57to prompt it and and you know give it explicit an explicit request and get an explicit response you give it a goal you
107. 17:04say I want you to go find I don't know the revenue and marketing taglines of my competitors and they will just go figure
108. 17:12out what it needs to do it'll be like oh I gotta go search Google for you know what the competitors are and I've got to
109. 17:19go over here to find the financial information it writes its own prompts but they get a little lost right now
110. 17:24they're they're very early so so all of those autonomous agents right now are pretty janky but but they're still pretty cool um
111. 17:34Tick Tock experiment followed the live Creator that's cool that's a cool name how much is it how much is what chat GPT this
112. 17:43that's free or Poe I was talking about Poe I think it's 20 bucks a month for Poe what do I think of pie the mathematical constant is it a
113. 17:56constant [Laughter] the mathematical uh thingy I think Pi is awesome it's infinitely awesome never repeats never repeats
114. 18:14um that's the article that really excited me and made me interested in learning more about AI yeah the the we have no moat thing is a pretty
115. 18:21remarkable it's a pretty remarkable article um oh inflection AI Pi I don't know what that is let me go see if I can I'll look
116. 18:30real time uh inflection AI pi oh on pie your personal assistant is this built on top of
117. 18:47doing [ __ ] for me right now hey pie sorry you've been blocked oh I think I I think someone was trying to scam me damn you
118. 19:03don't go there all right you are welcome to leave if you were just trying to get me to [ __ ] up my machine um let me get rid of this window
119. 19:18all right there you go oh that's a cool paper so so this is a paper this is a paper called generative agents so
120. 19:27what the hell is going on oh it's reloading okay I hope that didn't just install some shitty something in my in my machine um
121. 19:36this is a really cool paper generative agents um these are so this is like a sims-like little world and
122. 19:44um these things use chat GPT to talk to one another and they they have memory from day to day so it's amazing so all right
123. 19:55let's see especially raspberry I've had it give me incorrect info when I questioned it it apologized and
124. 20:02answered again yeah Emilio's wife but it will apologize answer you again and often get it wrong the second time and third time
125. 20:10and fourth time as well I like to call Chachi BT mansplaining as a service because it will confidently answer you even if it's wrong
126. 20:18mid-journey alternative is leonardo.ai and it's free in basic yeah so um leonardo.ai is really really good um it's it's not mid-journey
127. 20:31um I think it's built on top of stable diffusion leonardo.ai but it's an incredibly powerful um yes I'm white listed it's an incredibly
128. 20:42powerful front end I gotta log in ain't doing this stuff live is pain in the booty but that's it's called live we're doing
129. 20:52it live isn't this exciting I could get infected with oh there's all sorts of crazy images what are you people making out there anyway um
130. 21:05yeah it's it's uh it's pretty cool here's this this is my feed on Leonardo receive status I wonder if I'm gonna
131. 21:14have to quit Microsoft edge here for a second because I think the memory just went kaboomy when I went to a shitty non-good site so let me
132. 21:26fire that back up and see if it's still malbehaving um I've heard having two AI essentially debate oh wait hang on essentially
133. 21:36debate answers leads to more accurate results yeah Jeff it that's a way to do it and then also um uh oh hang on a second here okay
134. 21:54oh please all right um I'm gonna go do something uh my machine just hung for a second am I still live am I live am I live
135. 22:10Edge all right let's see my back let me pin this open hello hello am I still alive am I here
136. 22:26yes okay great terrific all right back to questions I no longer have memory day to day I know welcome to my life too
137. 22:39um oh so Jeff asked the question I heard that having two AIS essentially debate answers leads to more accurate results
138. 22:45yeah they can do that I think there's going to be whole systems that are built to to make sure things are accurate um
139. 22:52but one thing that you can do with chat GPT is when it gives you an answer you ask it to evaluate its own answer and
140. 23:00tell you how it would improve its answer and then apparently it gets quite quite good after that I haven't played a ton with that
141. 23:08generative agents more than a paper anything to demo or interactive I just saw some video game I saw footage from a video game where
142. 23:21a character walked into a bar and started talking with one of the NPCs and it was an open AI conversation so um
143. 23:30I assume that stuff's being developed right now I think we'll see those generative agents quite soon but I I
144. 23:37don't know that there's anything live you can play with you can if you go to the paper there's a hugging face
145. 23:44instance of the paper I think you can go play with you can go play with the The Sims thing that they have set up and you
146. 23:51can watch it do its thing it's pretty remarkable would many small smaller llms debating answers be more accurate than a much
147. 23:58larger one probably yeah like I I think I don't know I think that I think the large language models are going to be kind of like
148. 24:12internet like like like right now we just like you have access to Internet like everyone's got access to it it's
149. 24:19kind of this ubiquitous thing I think that's how the large language models are going to be and then I think you'll see
150. 24:24tools stacked on top of them and and probably smaller more specialized llms or just little training sets that that
151. 24:32you know kind of Leverage both will be there so so I think so I think that's probably where we're headed is is to a more um a more um
152. 24:43modular kind of kind of future but I also think that that modularity developers are going to have to deal
153. 24:52with the modularity but I think for us us humans that modularity is going to drift into the background very very quickly it's
154. 25:00just going to be like these tools are just going to get more powerful and more mythical mystical what's the prompt to improve an answer
155. 25:10um I forget what it's called um it's not query you don't have a query I forget what it is I forget what it is
156. 25:21can it build a business plan sure let's go do that right now I'm just going to go to chat GPT and I'll say here's how we'll do a
157. 25:30business plan we'll do um what are three different business plan Frame Works
158. 25:45for a startup question mark let's go see what it gives us there are numerous business plan Frameworks The Lean Startup plan
159. 25:56blah blah blah the business model canvas that's a cool one I'm a fan of the business model canvas and the traditional business plan
160. 26:08so let's say can can you create a custom framework framework um
161. 26:22that is a hybrid of these three so it remembers it knows what I'm I'm referencing when I say these three so it says sure here's a potential
162. 26:36business plan that that combines aspects of Lean Startup business model plan and traditional plan the framework can be termed hybrid
163. 26:47business model hbm this is the this is the HPM this is awesome executive summary value proposition market analysis marketing strategy
164. 26:59customer relationships key Partnerships revenue streams and cost structure implementation timeline and key Milestones lean methodology
165. 27:08so is it just giving me 10 are we going deeper okay give me 10. all right so there's that so so then I'll say um can you ask me
166. 27:26a series of questions about my business so that you have enough information to
167. 27:41um write a plan based on that framework frame work um um
168. 27:57do not proceed to the next question until I say okay do you understand what that should
169. 28:12do that last little thing is rather than it just giving me a pile of questions it should ask them one by one
170. 28:18absolutely I understand let's proceed step by step one can you provide me a brief overview of your business outlining what you
171. 28:27wanted to do why it's Unique and how it will generate money I want to sell Frozen lizards on street street corners
172. 28:44I think frozen lizards are cool and people will pay a lot for them
173. 29:00interesting concept before we move forward could you clarify a few points are these lizards so so you know you you
174. 29:09see you know the the sort of smart-ass response I put in there but um here's here's the thing about about chat GPT that that
175. 29:20this is a subtle shift that took me a while to get my head around Google trained us for years to get good at prompting it right so we
176. 29:34had to kind of you know early in Google's days you had to learn Boolean logic and and you know underst they
177. 29:42trained us how to use it and the basic idea there is you want to get good at writing a single prompt and getting back
178. 29:47good results what happens with chat apt it's much more like a collaborative conversation and so what I just did
179. 29:54there with the business plan thing was just this kind of back and forth hey tell me about business plans and then
180. 29:59make me a hybrid and then you know ask me questions and you can really treat your interaction with it like you're
181. 30:08collaborating with with a really smart you know co-worker um and and that's that's kind of it I mean it's it's it's incredibly powerful
182. 30:17all right I'm going back here am I on YouTube I missed this stream I'm not on YouTube on Tick Tock and and uh
183. 30:28and and I I've been busy at work so I haven't been putting up as many tick tocks and then I was in detention for
184. 30:34for being a bad boy here on Tick Tock they kicked me off for a week this is my this is my second detention
185. 30:41um I have no idea what I said I don't really care um I'm just trying to help people here so there you have it all right um
186. 30:57in the salon space What are you asking um I don't know what you're asking okay have any recommendations on a program
187. 31:05that can make a 3D printing model from a text or image prompt it's coming um there was a thing out today
188. 31:17well yeah let me go look on Twitter here hang on um text to 3D um
189. 31:33yeah this thing come on focus focus focus um and I'm too bright there you go um so this is called prolific Dreamer
190. 31:46High Fidelity and diverse text to 3D generation with variational score distillation huh these Geeks know how to Market don't they
191. 32:00did you know it had a variational score distillation no I didn't does it really oh yeah what's that do I have no idea
192. 32:10um but it does something but these look like you know fully complete 3D 3D objects that they're rotating around here so
193. 32:19two or three months ago the text to 3D stuff was um you know kind of laughably bad now those things look like decently modeled
194. 32:293D models so I would say give give it you know maybe by the end of the summer um you should be able I'm sure there
195. 32:40will be a service that's that's you know type in your thing here and output your model there I would I would be not
196. 32:48surprised at all if if a site like thingiverse or one of the 3D model sites um just Incorporated text to 3D right in
197. 32:56it so I I would imagine very very quickly you're gonna you're gonna be able to either take a 2D image turn that 3D or
198. 33:03go text to 3D and go straight to the printer I I think that's coming quite quickly Pi is in training but never been blocked
199. 33:10is that chat gbt three four or five I don't know Pi kicked me off that like I got I got blocked so I don't know what that was about
200. 33:19ask it a question that needs answered in order to do a Biz plan for a biotech startup you didn't think my frozen
201. 33:26lizards was a compelling business idea wait ask it a question it needs answered in order to oh okay okay um okay the a joke let's
202. 33:42start over what um what if I want to start a AI powered
203. 33:59biotech startup can you help me refine my idea question mark
204. 34:13of course let's move forward with your AI powered biotech startup idea it's a very exciting field with great potential
205. 34:20what's the problem you're trying to solve what's a unique solution so it just goes through and asks you so so
206. 34:26yeah you can and what I would say is you know rather than having me do this here um just go to here chat.openai.com that's the official chat
207. 34:39GPT web start site and you know while you're listening to me here just start interacting with it start asking me
208. 34:44questions you will get to a business plan in if you've never done it before at all and you don't have any clue about it
209. 34:53probably within 45 minutes if you kind of have a sense of what you want and know what you want it'll help you focus
210. 34:59that very very quickly I hope I offer AI prompt engineering courses um I'm thinking about putting together a
211. 35:06workshop I haven't done it yet so if if that's something that you'd like you know give me a follow let me know in my DMs
212. 35:13um or let me know in here it's that's something I might put together um I love Frozen lizards dry ice costs a
213. 35:20lot that's good see I'm getting good business advice here for my frozen lizard business I don't know why you
214. 35:25guys think this is a bad idea I think it's pretty brilliant all right let's see um it writes amazing content for
215. 35:33marketing cop copy email copy ebooks if you give it the right info yeah there's an interesting thing there's a
216. 35:41if you ask it shitty questions you get shitty content if you ask it good questions you get good content and and one of the things that is
217. 35:50counterintuitive about it is if it gives you crappy content don't quit don't just go oh it gave me shitty content
218. 35:58um tell it it gave you shitty content and tell it why that was boring I want it to be more funny can you make me sound smarter um
219. 36:09you know give it more context than than you think is reasonable it's it's almost like like what I've found is it's almost like
220. 36:18if someone just introduced you to a brand new intern at work and the intern was really really smart and could kind of do anything
221. 36:27and you wanted to give that person a project what would you say well you'd go well hey there um I'm doing this kind of a project here's
222. 36:39what the project is about here's the specific problem I'm trying to solve like if you give it lots of context and
223. 36:46then ask it to do something it will give you much better answers it's it's pretty amazing that way um oh can't drink anything on screen I
224. 36:55think even water okay that's weird uh I should have an ALT account to bounce to when I go to jail yeah I
225. 37:03should probably do that I hadn't thought about that that's a good idea I will come up with an ALT account it'll
226. 37:09it's it's uh who's that guy Hawk Hawk's got like seven of them so I guess I'll have a different account oh man uh tick
227. 37:18tock tick tick tock experiment it will be like any big tech tech transformation eventually things just suggest who owns
228. 37:25the rights to the answers that chat gbt gives you do um but um cheche EBT can use them for training there's there's a bunch of security
229. 37:34issues so if you go to your settings there's a data controls thing and you can turn off your chat history
230. 37:42and training so so you're basically telling chat GPD not to use your data and not to save your chat history
231. 37:49because that's where a lot of people are losing their data um but if you if you want to be really paranoid about
232. 37:57um your data privacy and security um run a large language an open source large language model locally or on a virtual machine
233. 38:07um locally would be the most secure um so I would do that but that requires some some more geekiness but but you can
234. 38:15um you've got some level of data controls now within chat GPT and those are going to get better and I I'm pretty
235. 38:21sure that chat GPT is going to give um they're going to give um that they're creating a business a business level account I'm thirsty I'm
236. 38:33gonna go drink off camera here for a second if that's really a rule that's the dumbest thing I've ever heard
237. 38:40like what are people supposed to do they want you to stay on these things for more more than 30 minutes but you can't
238. 38:46drink water okay text to 3D sounds like it's got a lot of retro encabulator going on yeah exactly um has anyone interested in creating a
239. 38:57promo for my AI map uh please don't um uh you know promote business stuff in here like ask questions like I I really don't
240. 39:07want this to be a Spam spammy place I want this to be a place where we can just talk and and learn um where you know I'm I'm happy to talk
241. 39:15about your business but don't don't spam to the audience if you wouldn't mind prompt engineering is made unnecessarily
242. 39:22complicated no I I disagree with that I think prompt engineering or prompt crafting I prefer the term prompt crafting because
243. 39:31I think it is a really creative Endeavor um it's complicated right now because the tools are so new like we're just
244. 39:40getting our hands on this chat interface um and here's the thing you're you're not even querying a database you're querying a system
245. 39:54that essentially knows all of the internet so it's an incredibly vast data set that you're interacting with
246. 40:06so the reason that that prompts can get complicated is that if you want really specific results you kind of have to
247. 40:14steer the large language model into a semantic corner and so so that your answers are are more you know are in a semantic region so if you
248. 40:26wanted to be talking about you know um whatever quantum mechanics then you want it kind of in the region of of its
249. 40:36knowledge where there's lots of quantum mechanics knowledge in there you don't want it over in in the area where it's
250. 40:40talking about poster design um so so that requires you to to do multi-layered um kind of conversations that focus it
251. 40:52more and more and more and more um and that can also get frustrating because because it's got a limited context window
252. 41:01it will sometimes lose its way where you started a conversation in one place you gave it a bunch of information and then
253. 41:07as that kind of Scrolls off the top of its context window it will kind of forgot forget that it you told it that so so
254. 41:16current prompt engineering is complicated because we're early but this is just like the early days of the web
255. 41:22you had to manually type in HTML there weren't site Builder tools and there weren't automated any things
256. 41:29and there weren't e-commerce engines if you wanted any of that crap you had to build it yourself and so it was just
257. 41:34more complicated and so that's ah look at that lovely little heart there's a lovely little heart flying right on my
258. 41:40head that was lovely thank you and yeah I got I have a lot of fun working on prompts yes prompt writing is fun you'll get banned
259. 41:49for anything apparently chap GPT says it is not for commercial use when asked yep that's probably true I would you know
260. 41:58I would be careful with it you can um open AI um announce you can do Standalone machines on Microsoft Azure that are
261. 42:07that are like Standalone um open AI instances so if you're working for a big company um you can do that I mean Microsoft's going
262. 42:17to make it very very easy says AWS sews everyone everyone is going to make it easy to do um your your your own data
263. 42:30um we have a dehydrated King exactly I'm gonna go have a sip now [Applause] okay so if you're new here this is the
264. 42:42AI learning lab we are learning about AI um I'm I'm sharing what I can if I don't know I will I will uh
265. 42:50tell you that if I do know I'll tell you that uh if you're new go to that URL chat.openai.com that's the official chat
266. 42:58GPT website it's I don't get paid for it um just start playing with it it's it's I think it is super super important for
267. 43:07anyone who hasn't tried this stuff to try it start using these tool tools start getting educated about them get
268. 43:15curious about them see what they actually make possible because they are remarkable and they're going to be
269. 43:20everywhere and all of us are going to be required to use them so why not do that voluntarily what can you ask or not ask GPT well so
270. 43:32it's got a bunch of safety um guard rails built into it so if you ask it things like you know how do I build a dirty bomb or how do I
271. 43:43probably I'm probably gonna get kicked off because I said that how to how do I build um whatever um how do I do how do I harm
272. 43:50someone right all that sort of stuff it will um it will tell tell you that you can't ask that there are hackers that are you know
273. 43:59kind of hack around it and get around the guard rails the open source models don't have any safety um guardrails built into them so chat
274. 44:07gbd has them um you can you know hackers can kind of get around them um but but basically they're they're
275. 44:15putting those guardrails in place to try to make it a safe thing right because they're incredibly powerful tools so
276. 44:20they want people to use them and basically if you push the boundaries it will tell you and then you know you
277. 44:26decide if if you wanted to do something that it's going to not let you do then go get an open source model and and you
278. 44:32know set up your own server you can do that did I hear about the FDA approval of neurolink yeah so so neurolink if you
279. 44:39don't know is Elon musk's company where they have these remarkable they built this remarkable robot that that inserts these 60 electrodes in
280. 44:49um well in your brain they drill a hole in your skull put these electrodes in your brain and it's it's essentially a
281. 44:55direct brain to computer interface and they've opened it up for um human trials so yeah that's coming
282. 45:02Switzerland neuroscientists helped them in oh well the Switzerland thing so that was different that wasn't neural link so
283. 45:10there's a guy in Switzerland he was paralyzed in a motorcycle accident 14 years ago or something like that and
284. 45:16they used a large language model they imaged his brain had him think about moving his legs they imaged his brain as
285. 45:23he was thinking about that so they figured okay here's the part of the brains that activate when he tries to move his legs then they built a
286. 45:29Bluetooth Bridge they did oh maybe they did use neurolog I don't know they inserted something in his spine so
287. 45:36something in his brain something in his spine and a Bluetooth connection between them and then they used a large language model to um
288. 45:44to to basically pull the information from his brain send it directly to his spine and now he's walking again after
289. 45:5014 years of not walking it's I'm I'm telling you like we're we're about to enter this just an era of remarkable scientific progress
290. 46:02um I don't care when people find it so difficult about making a prompt it's just being specific it's listen here's
291. 46:08the deal if if you've got a that's really funny if you've got a faculty a good faculty with words um if you're good with words there's a
292. 46:17great Steve Martin joke some people have a way with words other people um not halfway um if you're good with words prompt crafting
293. 46:28feels fairly intuitive if you're not good with words it it doesn't um but but it's like I would say this
294. 46:36the people that are good with words and understand programming are going to be really good at this because chat GPT
295. 46:42right now is essentially programming using your native language um so if you're good with words and you're good with logical constructs then
296. 46:51you're going to be really good at prom prompt crafting um using Ai and brain implant via spinal reconnection yeah I know yeah uh yeah
297. 46:59paralyzed man walks again after 12 years might have been 12 years not 14 but yeah really remarkable thing
298. 47:04um they didn't give out much detail because of the efficacy of neural link I I yeah I mean when we're starting to do things like
299. 47:13curing paralysis um I I'm I'm you know I'm good I'm down with it I think it's amazing [Music]
300. 47:27by the way I set a goal because Tick Tock told me to and I listened to everything Tick Tock tells me except for
301. 47:34the community guidelines apparently but I set a goal for having um 13 subscribers to my live and I only
302. 47:41have four so if you want to subscribe to this you can help me hit my goal I don't know what it does but Tick Tock told me to do it
303. 47:49it's it's uh it's tutoring me it's probably using AI algorithms to tutor me on how to get more engagement so that
304. 47:57they can brainwash you all with my words all right hope for more funding for stem cell therapy me too a bunch of companies
305. 48:03are working on it yeah I think we're just gonna see um just incredible breakthroughs on the medical front the next couple of years
306. 48:10what kind of price are we looking at when chat EPT is finished with this trial period I think they're gonna I
307. 48:16think free will still be there Bing is free that's gpt4 um there are numerous open source large language
308. 48:27models so they're they're going to be all sorts of tools that are free I would be deeply deeply surprised if chat GPT
309. 48:36didn't keep a free tier deeply surprised let's see I've been teaching my mom to use these
310. 48:51tools because I don't want her left behind at work yeah that's awesome I listen I I'm this channel exists because I am really
311. 49:01passionate about teaching as many people as I can about these tools and and not even like teach them like I'm a teacher
312. 49:08and I'm an expert I'm not I'm just a guy that thinks these tools are going to change the world and they're gonna It's Gonna Change the
313. 49:18World so much faster than any of us are prepared for so I want as many people as possible to be as curious as possible I
314. 49:25think the people that do positive things right now are going to have a disproportionately um large impact on the future direction
315. 49:35of this stuff um it it that may seem naive and silly but I think it's it's really really important that people figure this stuff
316. 49:43out and start to get their heads around it so I think that's awesome that you're doing that if you know anything about
317. 49:48this share what you know with people I won't share a background into info about private people yeah that's smart I wouldn't do that either
318. 49:57um quick overview of Open Source models easiest ones to start with oh well so I've got for a second um this is free also I don't I'm not
319. 50:10charging for this so this is not a you know go here to make money thing um power to the people issue 15
320. 50:21um so this is now like a month month old um but this talks about four different large language models red pajamas stable LM
321. 50:31stable LM is from stability so that's the that's the company that did um that does stable diffusion um weblm Mini gpt4
322. 50:43um and then so that's that's issue 15 of everyday Ai and then issue 16 I think we had a we had another one in there
323. 50:55first things about palantir which is a terrifying military company um is there another
324. 51:15so what's the so so anyways that's where that's where I'd go just go good everyday AI issue 15. that's a good
325. 51:23place to start and then the other thing to do is just go to go to Twitter and start following people that are talking
326. 51:30about AI stuff you can follow everyday AI um you can follow Robert scoble is talking a lot about this there's a bunch
327. 51:38of people you can follow Andre carpathi from openai um and then and then if you go to GitHub or hugging face those are the code
328. 51:49repositories and you know hugging faces where a bunch of large language models live um just look for the trending projects
329. 51:55on GitHub and hugging face and that'll probably point you in the direction of things that people are excited about will neuralink cure sorry
330. 52:09lost my place I lost my place well neuralink clear my depression I don't know look you're mine look you're any of ours it's uh
331. 52:20we're just in this mental health morass aren't we I just feel like everyone right now is struggling on every
332. 52:26business that I'm interacting with people are just checked out and you know I think we're all trying to figure everything out and uh
333. 52:35yeah I I you listen I think I think that these tools are gonna are gonna have an incredibly um positive impact on Mental Health if
334. 52:46we use them right and whether or not neurolink is the interface or not I I mean I can't imagine so so right now
335. 52:56they've got it so that you know where elon's technology is is you've got to drill a whole hole in the skull and
336. 53:04implant these electrodes in the brain and that's you know that's instant connection to the brain but I can't imagine that AI isn't going
337. 53:13to help us develop a baseball hat kind of thing that's got you know brain reading technology in it that's actually quite fast
338. 53:20so so we'll see um I got an MS in computer science um at Georgia Tech I know a lot about this field that's cool yeah I have a
339. 53:30degree in acting from Penn State I you know the the I know I know conceptually a lot about this and and I I would say
340. 53:39my expertise is more in using technology um once the um computer scientists and technologists um have gotten it to a state that the
341. 53:50regular people can start playing with it I did that in the mid 90s with the world wide web and I'm doing it again here I think that um
342. 53:59the the this large language the large language model stuff and and AI stuff is is changing the game very similar to how
343. 54:07the World Wide Web didn't back then I have a way with words so llms and I are BFFs that's great yeah exactly if
344. 54:14you're good with words man this this stuff's a lot of fun um AI lab prompt crafting 101 okay I'll I'll maybe put together I'll put
345. 54:23together a workshop and and see if anyone's interested in it um I must do is The Tick Tock God's demand apparently
346. 54:31um thank you for your service thank you for thanking me for my service I you know I I enjoy this so I don't really
347. 54:37see this as a service this is fun um I've seen Twitter posts commenting on the seeming dumbing down of the results they get from chat GPT
348. 54:46um well certainly certainly um Microsoft dumbed down um Bing after they made the front page of the New York Times when when the uh
349. 54:59when the chat bot told the reporter he should leave his wife because it was in love with him um they dumbed it down after that um
350. 55:07they are they are definitely doing fine tuning and training so so these models are changing so there's some prompt
351. 55:16degradation that can happen where you have a prompt that works for a while and then it just kind of changes and gets worse so
352. 55:24um I wouldn't be surprised at that I haven't seen it personally but is python the standard when programming AI
353. 55:32what about CPP or even lisp I would go ask uh GPT for that I would use the one connected to t plus plan or
354. 55:41you're paying them they're 20 bucks a month fee you've got browse with Bing so so this is um chat GPT connected to the Internet so
355. 55:52I would go ask it um uh what language you know what what languages to use for AI and specifically what you want to do in AI because
356. 56:02Python's good for certain things and you know lisp or CPP might could be good for other things so I I would tell GPT what
357. 56:12you want to accomplish and then have it recommend what language you should be using and then have it write the code
358. 56:18for you because it does that and it does it remarkably well any opinion on graph Theory and information Theory also algebraic coding
359. 56:26Theory well as an actor no I don't I don't have any opinions on those because I don't know what most of those are um I do know that
360. 56:38the wolf from alpha plug-in for chat GPT allows you to kind of mix those two worlds together so and there's also some of the plugins for for
361. 56:49um chat GPT also have graphing and you know visualization capabilities so um so that's pretty interesting I would say
362. 56:59um when code interpreter comes out which is a another module that they're doing where you can do all sorts of fancy data analysis
363. 57:07um that that'll probably be pretty good at that stuff but I don't know anything about those sorry about that what are
364. 57:12our thoughts on fine-tuned Laura versus large General models I think these fine-tuned small models are are like
365. 57:20like the developer Community is pouncing on every time there's a new model out and they're getting really good really
366. 57:27fast so so I think that these small I I think probably within the year you'll be able to install something on
367. 57:37any phone any device locally that is in the neighborhood of gpt4 that's just a small teeny little thing that doesn't require
368. 57:48a massive chip to use um so yeah so I I'm I think those things are remarkable subscribe needs to find 5.99 a month I
369. 57:58hope you see a big part of that profit I don't see a part of any profit on any subscriptions so um and I don't know what you're talking
370. 58:07about subscribe means 5.99 a month oh oh for months here if I'm asking you to you described to my lives that's 5.99 a month to subscribe
371. 58:20to my lives I would not pay 50 cents for my lives I thought it was just if you subscribe you get notified GPT
372. 58:27gpd4 is incredible to me it's helping me create structure with my goals that's a great way to use it did you hear IBM research
373. 58:352033 quantum computer center and smaller Fusion energy production research 2033 no I haven't heard that
374. 58:45I've heard a couple of things that IBM's doing um like like they announced they're they're sort of dusting off Watson after after
375. 58:53it failed with Healthcare especially helpful with my ad HD makes it difficult to structure my goals oh that's cool yeah I'm a card carrying
376. 59:02member of the ADHD Club so welcome to the welcome to the club buddy um yeah um AI is both a blessing and a curse for
377. 59:10us adhders because man you can run down some [ __ ] rabbit holes [Laughter] all right they also lobotomized it yeah
378. 59:19they did they lobotomized uh what was her name uh Sydney they lobotomized Sydney once she when she got all
379. 59:27boiling rabbits on uh on that reporter all right we've got Collective fatigue from the past five years we do I think
380. 59:36long covet may be contributing to mass lethargy that's an interesting thought yeah I you know I think it's I think it's um
381. 59:47I think it's lack of human interaction I I think you know I like my company's a small company but you know I've got a
382. 59:54bunch of introverts in there that love working from home I hate working from home so we're getting things done but I've
383. 1:00:01noticed that just the connection to the company is not there I I don't think that's unique to us it's it's a drag
384. 1:00:08I have problems with the prompts so so here's the thing you can try is almost think of Chachi PT as your prompt
385. 1:00:18therapist actually let's go try this now this could be kind of interesting I'm going to use GPT I'm not good with words can you help me
386. 1:00:34learn how to prompt you better let me just leave it at that of course I'm here to help be clear and specific
387. 1:00:51clearly State what information breakdown complex questions provide context ask for clarification use examples so it'll teach you and say um
388. 1:01:02I don't know what any of that means um can you walk me through step by step
389. 1:01:18what to do question mark certainly step one identify your objective step two consider examples so so you can ask it
390. 1:01:32to do that so so one thing you can do is just start having it train you how to how to prompt better um there's another thing that you can do
391. 1:01:41within if you've got access to plugins let me let me create a new chat here so we're going to go to gpt4 plugins and then one of the plugins
392. 1:01:52is called prompt perfect so I'm going to turn on prompt perfect and the way this works is you type in the word prompt
393. 1:02:00is it prompt or perfect wait hang on oh you type perfect okay so I type in perfect at the beginning of a prompt and then I say
394. 1:02:16um teach wait teach me how to make better prompts for chat GPT
395. 1:02:33and so what this plugin does it's called prompt perfect what this plugin does is it will then it will rewrite your prompt
396. 1:02:41as a better prompt and then give you a better answer so let's see oh error communicating with the plugin service
397. 1:02:53regenerate response let me try it again maybe just add a hiccup now it doesn't like it it's not working right now
398. 1:03:00some some of these some of these uh plugins are a little wonky um um prompts.chat prompts dot chat which is a big long list why is it not
399. 1:03:15going come on prompts.chat there we go call it's called awesome chat gbt prompts it gives you a bunch of resources at the
400. 1:03:23beginning talks to you about prompt writing and then it's a whole long list of you're telling Chachi PD to act like a certain
401. 1:03:32um kind of like act like a travel guide act like an Advertiser act like a Storyteller act like a stand-up comedian
402. 1:03:39act like a motivational coach so it gives you prompts that the kind of focus chat gbt and make it um talk within a specific context um
403. 1:03:49I think I'm going to add that to my little uh my little photo shop thing here we're going to put in prompts dot chat
404. 1:04:02yeah look at that now we got now we're cooking with gas that's a very Gen X saying isn't it yeah we're cooking with
405. 1:04:12gas we're drinking out of the hose now all right hope that helped with the prompts is it on the stock market yet is open Ai No
406. 1:04:21they're not public they're a private company uh can it understand stocks yes it can write you trading Bots if you wanted to
407. 1:04:28they didn't dump it down they censored it whatever same thing the the effective result is the same they lobotomized it benefits of chat GPT
408. 1:04:37Premium versus Bing version um I think Bing is still a little neutered for my taste I don't I I think Bing is
409. 1:04:47is it's getting better but it still sucks a little bit I just think chat EBT is the most elegant of the large language models out there
410. 1:04:56what are you showing us on your PC screen I don't know what you mean um oh what are these things okay the top
411. 1:05:05one is chat GPT website the next one is that prompt thing and then future PD is just a list of tools
412. 1:05:14early GPT wrote my bio I've had it right my bio a dozen times in the past couple of months oh it said you were the CEO of a random
413. 1:05:26biotech company yeah chatgpt will lie to you make no mistake it is mansplaining as a service chat lies absolutely all right
414. 1:05:43thanks I've learned more than the year I've been using the a i the artist AI Creator thanks you're welcome thank you
415. 1:05:51very much I appreciate that yeah my purpose here is really just to do what I can to demystify these tools share anything that I know
416. 1:06:00point you in directions of things get you curious about this stuff because I think it's quite remarkable I gotta go
417. 1:06:05to bed soon too it is midnight and a half here in Denver Colorado makes I mean someone said something interesting and then I lost it because
418. 1:06:16Tick Tock is a it's a Fickle little minx Quantum makes AI small enough to fit in the phone oh interesting China's already doing this
419. 1:06:30to to to us everything I don't quite know what that means but but yeah I that I would I would think that we're gonna probably start to see
420. 1:06:40large language models on chips they'll build them into the hardware and we'll have lots of little devices that have incredibly
421. 1:06:48strong capabilities I think I'd use chat gbt to analyze my code for vulnerabilities and errors but not write
422. 1:06:53it for me it's great at analyzing code I had a we had for my app for my company um we had some obscure bug that was it
423. 1:07:01was an edge case where it just kept crashing on this one device and um our developer couldn't figure it out
424. 1:07:08couldn't figure it out pulled it into chat GPT asked what the problem was and it found it like one flipped you know
425. 1:07:15flipped something or other that didn't seem like it would be a problem flip basically flipped it back ran it
426. 1:07:22and it fixed it fixed the bug so it's it's remarkable at that stuff I let it do all my assignments why not [ __ ] it
427. 1:07:29half of long covid it I I would say so [ __ ] it on doing your assignments but you know learn what it's doing like like
428. 1:07:38the the the the the critical skill moving forward is not writing the paper right I feel like writing the papers days have have come and gone
429. 1:07:48chat GPT can write the paper but what the skill is is the creative the critical thinking and and the creative
430. 1:07:54thinking about well what's the paper you want to write why do you want to write that paper who's going to read that
431. 1:08:00paper why are they going to read it right all that sort of critical thinking stuff becomes way more important so if
432. 1:08:06you're just having it do doing your assignments and not using the Noggin you're you're screwing yourself all right half of long covet maybe
433. 1:08:13something else yeah I've got kids with chronic lyme disease so it could be that but hopefully AI will solve some of that [ __ ] too
434. 1:08:20because I'm crankier than [ __ ] about that um China's recent living off the land attacks are all AI driven living off the land of tax
435. 1:08:30powered by Quantum okay interesting yeah I don't know I'll look into Quantum that's really interesting um we're about to see a
436. 1:08:39Bonkers election season because of AI my goodness I love work from home as an introvert I live my life in my own head
437. 1:08:48yeah I listen I I appreciate the the introverts and that you're living your best life at home but us as extroverts
438. 1:08:56got our asses handed to us I did a I did a daily video diary for every day for a year for the first year of lockdown
439. 1:09:03called the extra extroverts View From Within it was just me whining for you can find that on YouTube uh it's just me
440. 1:09:12whining for a year about how I didn't like working from home um asked gbt how to cure Cova Chet GPT is going to cure a lot of stuff large
441. 1:09:24language models are just the large language models ability to parse data and allow you to just interact with it
442. 1:09:34like a [ __ ] magician is that's what it's like it's it's it's like magic it's it's insane I love AI telling teachers of works
443. 1:09:46this wait I love AI telling teacher oh if work is AI or not that's going to go away that that's such a short-term thing
444. 1:09:53because these these large language model tools are going to be built into every single writing tool every single
445. 1:09:58operating system every single app on your phone it's it's going to be at the operating system level so the the whole
446. 1:10:04thing about teachers resisting this is like when they resisted the calculator it's it's just kind of a useless
447. 1:10:10exercise right now what they should be working on is okay if large language models are going to be in [ __ ]
448. 1:10:16everything why are we teaching people little tactical skills that are going to be automated away why don't we teach
449. 1:10:21them critical thinking skills and like revenge of the liberal arts Majors where we need people that can think laterally not vertically
450. 1:10:30do programmers work on improving these models still or are the Bots doing it probably both well like a lot of the um
451. 1:10:37a lot of the open source large language models they're they're writing algorithms to use chat GPT to train to train
452. 1:10:47um the large language model so so it's both um if you code they prioritize you if you ask dumb questions then no yeah I
453. 1:10:55agree with that the robots are strong um I'm sure there's other categories they prioritize [Music] I'm a millennial I'm a gen xer
454. 1:11:08um they were public for a minute oh no I say that stuff I'm a millennial they were public for a minute open AI I
455. 1:11:16don't think was ever public no they were non-profit for a minute and now they are a capped for-profit and I heard Sam Altman in an interview
456. 1:11:27last week talk about the fact that he has zero percent ownership in open AI which I cannot wrap my head around
457. 1:11:35there there might be some other way he's getting compensated I can't imagine he's putting this much time into that
458. 1:11:41organization with no upside on the back end I mean he might just be doing it out of the good of his heart but I don't think so
459. 1:11:52um all right let me answer a few more questions and I'm gonna go to bed bye all right this guy reading comments from
460. 1:12:00three hours ago yeah that's what I do I try to I try to read everyone's comments and I get behind so whatever
461. 1:12:06you you are welcome to leave I encourage you to take off if you don't like it I've been on the best I can
462. 1:12:14just me and a phone and a glass of water I'm not allowed to drink on camera ooh tasty okay must be a lot of work filtering all the
463. 1:12:28garbage on the net because at least 25 percent of it is bad info or opinion yeah but you know the better you get it prompting the
464. 1:12:35better that you can extract actual value out of it and and you know think of it like this these large language models
465. 1:12:41get you to 80 really quickly it's still up to you to do the final refinement of the work has been my experience
466. 1:12:51uh uh uh I'm glad that you're helping one another in this that's super cool I like seeing that all right
467. 1:13:02all right so rather than reading dozens can provide information I'm looking for immediately wait my God so it's like 10 minutes I
468. 1:13:13why can't you drink on camera I don't know because well so I've been I've been kicked off of these lives twice now for
469. 1:13:20and it's seven like I go into detention for seven days where I can't go live um and I have no idea like why I went
470. 1:13:28like it doesn't care if I say [ __ ] or [ __ ] or piss or anything like that um but something I said or and then someone
471. 1:13:37told me in the community guidelines you can't drink on camera so now I'm drinking off camera because maybe that's
472. 1:13:42it I don't I don't know I don't I don't know and I don't really care so if anyone wants to learn the community guidelines and tell
473. 1:13:50me when I infract I'm fine with that but I don't really give a [ __ ] all right let's see I think I'm kind of getting to the end
474. 1:14:06of comments this guy my name is Kyle John this guy reading comments from three hours ago come on don't pick on me
475. 1:14:18this is me just sitting here in a chair talking words out of my mouth extra Google John Connor wants you to
476. 1:14:26calm down sir I will not calm down oh my machine went to sleep I will not calm down I'm all peppy and
477. 1:14:35fired up I'm I'm live again after a week away the AI and the chat GPT is the beginning of the end or it's the end of the beginning
478. 1:14:46come on gen xer2 Gen X bitter you betcha all right okay I can provide the info I'm looking for it's like 10 minutes
479. 1:14:57why can't you drink on camera I don't know he caught up laughing out loud all right wait we both what we both
480. 1:15:07this is great thank you so much thank you Sadie I appreciate it yeah I gotta go to bed um glad I stumbled across your live can
481. 1:15:13AI write me a killer resume and cover letter yes so here's a fun thing to do um just go assuming that you've put
482. 1:15:22together a LinkedIn profile you can just literally go copy and paste your LinkedIn here let's go do it I'll go do mine
483. 1:15:28filter I just paste it in everything I had it looks like it cut something off but whatever oh yeah that's just one of the
484. 1:15:36comments at the bottom okay so I'm gonna say um based on this right me a resume for a pose position
485. 1:15:57uh in the generative AI field um please use
486. 1:16:12resume best practices okay so then it took all that
487. 1:16:26and now it's writing me so Kyle Shannon I'm going to stop generating I'm going to say please use markdown um to make it look better
488. 1:16:42so markdown is just basically a it's kind of like HTML it's but it's it's like a more simple HTML just to do some basic formatting
489. 1:16:54and now it's stuck looks like it looks like chat GPT is a little sleepy tonight Kyle Shannon Denver Colorado objective
490. 1:17:06highly experienced so email you know put it here so it's formatting it for me it gave me an objective experience
491. 1:17:19so it's right now my experience so so you could you could do things like I'm just going to stop at generating here
492. 1:17:25you could do things like tell it to you could copy and paste in a format of a resume that you like tell it to match
493. 1:17:32that formatting I can now say um please um please write a cover letter for a senior position
494. 1:17:49leading a AI oops AI strategy for a Pharma company whatever
495. 1:18:07your name your address today's date hiring manager's name giving me all that dear hiring manager I'm writing to express my interest in a
496. 1:18:20senior leadership position my extensive experience in marketing strategy has equipped me with a multitude of skills I believe make me an
497. 1:18:29ideal candidate I've spent more than a decade in the AI field that's not true as Chief generative officer and founder
498. 1:18:36of the AI artist salon and storyvine have successfully applied my skills that's true right so it writes that I'm going to say
499. 1:18:45um can you make make it more conversational it will make it more conversational after it writes all that crap
500. 1:19:00um [Music] so the answer is yes it can I hope this letter defines you well to say I was excited would be an
501. 1:19:13understatement so I could understand you know sort of basic you know language requests and context and things like that
502. 1:19:21um if you're new here this is the AI learning lab I talk about there's much AI stuff as I can as coherently as I can even though it is
503. 1:19:30nearly 1am here in Denver um these uh URLs the top one is chatgpt the next one is prompts for chat gbt and
504. 1:19:41the bottom one is just AI tools if you want to look around I would strongly encourage you to go to chat CPT if
505. 1:19:47you've never used it just go there it's free sign up um start playing with it all right just saw your AI so cool my
506. 1:19:56friends asking VCS oops wait uh buffering uh Gen X you froze am I Frozen now am I good now hit the hit the screen let me know if I'm frozen
507. 1:20:10just some feedback you're providing an excellent Show Your Enthusiasm is really appreciated oh thank you I appreciate that very much
508. 1:20:19um he broke it did I break it okay and my back looks like I'm back this is awesome thank you I'm so glad you say
509. 1:20:26please to the yeah I do that too I know I find myself please you know please do this thank you for that
510. 1:20:32I'll if it gives you something good I'll say that was awesome now could you do this um it's I mean it's chat apt is programmed
511. 1:20:42to to give you what you want and so it's programmed to be polite to you so it it really is like interacting with a remote
512. 1:20:51employee that just happens to be way the [ __ ] smarter than you are um you should integrate aiprm check epd prompts to to I don't
513. 1:21:01know what those are I'll look those up I missed I missed it generative resume in best what kind of format um
514. 1:21:13I don't know whatever like go go find a resume format you like and you can just copy it in there I mean you're really
515. 1:21:19just looking for the the text of it um the web GPT extension okay I don't know if that is um you don't happen to have a video
516. 1:21:30breaking this down do you breaking what down I don't I don't know like my channel is full of videos breaking stuff
517. 1:21:36down I have videos talking about how chatgpt works I've got videos talking about how to build automations in zapier
518. 1:21:44um lots and lots of things thanks Emilio's wife um so I'm not quite sure what you're you're referring to there
519. 1:21:52um not frozen okay good you're good now good now good awesome how can I make Chachi PD to summarize a large PDF um so if you go
520. 1:22:05to gpt4 go to plugins plugins that I have been using recently is this thing called ask your PDF and so if I put in
521. 1:22:19let me go let me go find a white paper uh uh um let's do attention is all you need so attention is all you need is the
522. 1:22:34famous white paper that where Google introduced the concept of the Transformer and the Transformer is the thing that
523. 1:22:45all of these Cool Tools that we're using right now are built on top of so I'm going to grab that URL of this PDF that's out on the web
524. 1:22:56and I'm going to go in here to gpt4 and I'm going to say um please summarize this PDF in Five Short
525. 1:23:16bullet points and then I put a colon and then I paste in my URL and I hit return and now what the
526. 1:23:30um plug and do is it should fire off the plugin it's not like in plugins tonight for some reason I I think their servers are overloaded
527. 1:23:38right now is my sense yeah it's not it's just not happy with plugins um anyway assuming they work then then
528. 1:23:46it will go it'll it'll go read that PDF and and write I like I created a uh I created a prompt to generate a Twitter
529. 1:23:55thread for my newsletter so I can take any white paper or any article that I like and it will automatically break it
530. 1:24:02down into 10 Tweets in a in a thread um and and that's reading the thing I don't know how if it if the PDF is longer than um
531. 1:24:144 000 tokens I don't know if it breaks it up or not my guess is that these things don't because they're not
532. 1:24:19sophisticated but that's that's the way you can do PDFs all right let's see what else have we got here the resume sorry it should have been
533. 1:24:30more specific uh what was the question damn it um I don't know what the question was regarding the resume sorry about that all right
534. 1:24:45I want some passive income to get the wait oh if you want some passive income get those videos on YouTube oh that's
535. 1:24:52good thank you oh you've got a good content I appreciate that yeah I you know I'm doing this you know as a I wouldn't
536. 1:25:01call it a hobby because it's there's sort of a professional angle to this but I I'm not really thinking of this
537. 1:25:06content as a business I should probably think of it more as a business hey I'm the head of Applied ml stability
538. 1:25:13would love to chat with you about our models dude let's let's chat I would love to I would love to talk about
539. 1:25:23um about what stabilities up to I think I think what stability is up to is really interesting I think the fact that
540. 1:25:28you're open sourcing everything is fascinating um so so I would love to talk to you so um Joe Pena so um let's connect and I'm
541. 1:25:41going to write down your name so that I can go find you um but yeah I would love to talk maybe we can even have you come talk at the AI
542. 1:25:50salon so we we uh the AI Salon is at the salon.ai is the URL there's a link tree there and we meet every um every other week and we
543. 1:26:00have an online Discord um we had the head of AI from Adobe spoke really early on before they put out Firefly
544. 1:26:07um he came and showed us some of the some of what ended up ended up becoming Firefly and I just talked to him today
545. 1:26:13and uh he's going to come back and talk about the new generative AI um the generative fill tool they put in
546. 1:26:20Photoshop because it's like that's blowing up and so I said hey why don't you come back and tell us what you're learning so
547. 1:26:26um would love to chat chat with you about that stuff so I appreciate it appreciate you making the comment Joe Joe Pena cool that's exciting
548. 1:26:35see that's why I'm doing this I'm doing this because we're just at an exciting time right now and and you know
549. 1:26:45learning as much as you can about how these things work what they do why they do what they do why stability is open sourcing with their open
550. 1:26:54sourcing you know um I think if it weren't for companies like stability and open AI that just um are asking for forgiveness rather
551. 1:27:05than permission in terms of all the training of the models if if people had dealt with the the intellectual property
552. 1:27:12issues that are absolutely there right you know um if they had dealt with this proactively like can we have permission
553. 1:27:18to do this the answer would have been no none of this stuff would have been created so so all of this stuff exists
554. 1:27:24because of a little bit of wild west attitude that's just where we are right now we're very early in this part of the
555. 1:27:29industry and it's super super exciting um a fan in the chat pogies cool um I use an ad Google Chrome so I don't
556. 1:27:41need type questions and hit spacebar for recording I don't know what that's talking about but sounds like you're
557. 1:27:45helping one another good I like that if I have a pretty video breaking down the process I have a video breaking down
558. 1:27:53breaking down the process of what of prompting or of how chat EPT works I can I can talk about that I've got to in fact
559. 1:28:03no so somewhere in my in my videos is a video that talks about how chat GPT Works um thank you for the kind words Sadie I
560. 1:28:14appreciate that I obviously don't participate in lives do you have a video on how to do a resume oh I don't I was just kind of
561. 1:28:21doing that live on the Fly um no I would um I just go to YouTube and say um you know um using chatgpt to create a resume there's
562. 1:28:36tons of them there um I was just kind of doing it live on the flight like my interest in in this is you know rather than showing you
563. 1:28:43exactly how to do a resume my advice would be go start playing like just copy and paste um your work experience either from
564. 1:28:52LinkedIn or just you know sort of type it in and then have it help you organize it um and just tell it to
565. 1:28:59like like what you might want to do is start out by Say by saying you know chat EBT you're an expert you
566. 1:29:09know resume writer and employment coach tell me four different um types of resumes that I can put together have it teach you about resumes
567. 1:29:20and then ask it um you know based on the one that you like to write you a resume based on your work experience and it'll do it
568. 1:29:31all right what are what am I primarily using chat GPT for I'm using it for many many things at work like almost everything that I do now um
569. 1:29:44I'm having a rewrite things I'm having it help me organize things up front um we're using it for Creative
570. 1:29:54development for the video templates that we build I'm creating content generation engines um I'm completely rethinking
571. 1:30:03how my video platform does what it does and what it even outputs um I use it for fun I use it to help
572. 1:30:16organize my thoughts for the newsletter um I don't know like it's almost like the the shift that I've successfully
573. 1:30:27made this took me a while was whenever someone presents me with some sort of problem or if I come up with a problem I want to solve
574. 1:30:35where I've historically gone is okay here's how I Kyle would think about solving that problem and I don't do that
575. 1:30:43now I'm like oh is this something chat GPT could help with yes okay let's just go right to chat EBT and I start to work
576. 1:30:51with it like a creative collaborator can you use chat EBD to craft a legal argument you can you need to be careful
577. 1:30:58some lawyer just got busted because he he wrote he wrote some sort of legal opinion and he didn't check the stupid
578. 1:31:05references the idiot and the judge caught him these your your the things that you cited for this legal argument are made
579. 1:31:16up so you've got to be careful with that but yeah it's really good at understanding legal documents and Concepts it's just crazy
580. 1:31:25um all right let me answer a few more then I gotta go to bed I gotta go to bed I keep saying that it's one o'clock but
581. 1:31:32this is fun I like I like chatting with you folks um I was trying to oh you were trying to take screenshots as I was going through
582. 1:31:42the the steps with the resume yeah I would say just go start playing it like have it teach you have it teach you how
583. 1:31:48to do what you wanted to do it's it's really good at that um me too I use it for nearly all my work tasks I'm in a Management
584. 1:31:55Consultant yeah like Consulting right now it like there's a double-edged sword I think that the challenge with being a consultant
585. 1:32:03right now is that you know c-suite people that hire Consultants are gonna start answering their own questions they're
586. 1:32:12not going to need to call the consultant but Consultants that are leveraging these tools and can turn around their thinking
587. 1:32:22faster and more comprehensively than their competitors you're going to have an advantage so if you if you're getting
588. 1:32:28your head around these AI tools right now um I would dig I I think it's I think it's wicked smart and I you know I think
589. 1:32:36you can I think Consultants can change the game the Smart Ones um new here what's your video platform
590. 1:32:42so my company is called storyvine it's a uh it's an automated video storytelling platform we do a lot of work in
591. 1:32:49healthcare and Pharma we've been around for 11 years we don't use a lot of AI um in how we put our videos together we
592. 1:32:57use more Aristotle it's more storytelling based um but I've been layering in AI tools to the platform and I'm completely
593. 1:33:05rethinking what the platform is now so storyvine.com is the uh is the uh is the company we do you know like if
594. 1:33:15like if Pfizer wants their patients to talk about their disease Journey our app guides that patient how to tell their
595. 1:33:22story goes up to the cloud five minutes later you got a fully edited video all right uh I've been writing
596. 1:33:29commercial scripts with it and writing JavaScript for After Effects yeah the the whole thing of writing like little
597. 1:33:34little scripts little code Snippets for things like unity and after effects is so powerful muted am I muted can you hear me
598. 1:33:45if you can hear me tap the screen so I can see the things the lawyer got busted was too funny I know like like here's the thing use
599. 1:33:53these tools but also use your [ __ ] brain like look at the [ __ ] that it wrote especially if you're a lawyer
600. 1:34:00really you [ __ ] idiot it's like Jesus Christ okay we can hear you thank you all right um yeah so
601. 1:34:17all right I guess I'll leave on this so so first of all thank you all for hanging out with me you know just sort
602. 1:34:22of rambling here it's because it's late so I'm not all that coherent but but I'll leave you with this um if you haven't started playing with
603. 1:34:33these tools go play with these tools and like play play like like think of it like this we're lucky enough to to to have the most powerful
604. 1:34:44technology like in the history of mankind literally at our fingertips go play with it with that sense of
605. 1:34:53wonder and just start pushing it the more you push it the more it will reveal what it can do um and keep your brain engaged don't think
606. 1:35:04of it like this these tools are gonna do the work for me no no no you're still in charge of what your intention is what
607. 1:35:12you're trying to accomplish who you're trying to accomplish it for so so the the critical thinking and the the strategy
608. 1:35:20of why you would use these tools in the first place that's still critical like that's more critical than ever so don't
609. 1:35:27lose that piece but please start go play going and playing with the Tactical bits all right um I will think about putting together
610. 1:35:38um yeah yeah at our fingertips and free mostly yeah absolutely it's we're these are remarkable times we're at the
611. 1:35:43beginning of something incredibly um powerful here so thanks all for your time I'll be back I'm trying to do these
612. 1:35:51once a night now that I now that I'm not kicked off anymore and I'm not in detention um and uh I will
613. 1:35:58um I will see you tomorrow all right bye bye everybody thank you