1 00:00:00,000 --> 00:00:03,540 some of the gentlemen over at the company known as minstrel AI. 2 00:00:04,140 --> 00:00:05,700 These folks here to 70. 3 00:00:05,700 --> 00:00:06,940 Not minstrel. 4 00:00:06,940 --> 00:00:07,540 Yeah. 5 00:00:07,540 --> 00:00:08,580 Mistral. 6 00:00:08,580 --> 00:00:30,240 Minstrel would be a bit something else. 7 00:00:30,240 --> 00:00:38,280 Ladies and gentlemen, boys and girls, children of all ages, dogs, cats, robots, and everybody in between, especially you, five day old AI startups with a multimillion dollar 8 00:00:38,280 --> 00:00:39,280 valuations. 9 00:00:39,280 --> 00:00:41,960 This is HTTTA. 10 00:00:41,960 --> 00:00:42,800 How to talk to AI. 11 00:00:42,800 --> 00:00:45,200 I am your host, Wes the SynthBind and SynthBind Wes. 12 00:00:45,200 --> 00:00:56,760 And as always, in a galaxy where gigabytes govern and algorithms guide our growth, one voice gallantly gears up to demystify the grander, strip down the gnarly knots and steer 13 00:00:56,760 --> 00:01:01,000 through the unexplored grid of artificial intelligence. 14 00:01:01,000 --> 00:01:01,920 Welcome. 15 00:01:01,920 --> 00:01:03,080 Who may you ask? 16 00:01:03,080 --> 00:01:07,040 The gateway to all things AI, gracing the border between humanity and technology. 17 00:01:07,040 --> 00:01:12,240 She is your guru, your guide, your good friend, ladies and gentlemen, Miss GoToGo. 18 00:01:12,240 --> 00:01:13,720 G, how are you this week? 19 00:01:13,720 --> 00:01:15,760 Hi, I am amazing. 20 00:01:15,760 --> 00:01:21,960 I'm so happy to be back, but I'm also sad that I missed you and Keegan last week. 21 00:01:21,960 --> 00:01:25,760 Hey, we had a good old time, but we do appreciate your contribution. 22 00:01:25,760 --> 00:01:26,240 Oh my God. 23 00:01:26,240 --> 00:01:33,080 When I was listening in the edit, I was like, oh, I was getting like, I'm hungry. 24 00:01:33,080 --> 00:01:34,880 I literally went to Kitten.AI. 25 00:01:34,880 --> 00:01:37,680 I was like, this is incredible. 26 00:01:37,680 --> 00:01:39,560 I think he's really onto something. 27 00:01:39,560 --> 00:01:40,760 He is indeed. 28 00:01:40,760 --> 00:01:44,760 And just the future plans are also fascinating. 29 00:01:44,760 --> 00:01:54,480 And that being said, like when I was listening to both of you, I was almost getting the same vibes as watching my really good friend from Italy. 30 00:01:54,480 --> 00:01:57,800 So Italian cooking and you both nerding out. 31 00:01:57,800 --> 00:02:00,640 I was like, oh my God, this is great. 32 00:02:00,640 --> 00:02:07,960 By the way, I have to share with you that same my friend sent me a message, but he's now taking CS50 course. 33 00:02:07,960 --> 00:02:08,840 Thanks to you. 34 00:02:08,840 --> 00:02:10,440 Yeah, awesome. 35 00:02:10,440 --> 00:02:12,800 Hey, I'm totally glad to hear that. 36 00:02:12,800 --> 00:02:20,440 I think it's, you know, I can't say enough about how that helps connect some of the dots for all the stuff that underpins the world around us these days. 37 00:02:20,440 --> 00:02:21,320 Absolutely. 38 00:02:21,320 --> 00:02:28,480 But yeah, let's jump into talking a little bit about the business side of AI and what kind of craziness is happening. 39 00:02:28,480 --> 00:02:31,240 That's something we can we can definitely talk about. 40 00:02:31,240 --> 00:02:41,560 I'm sure from every side, people kind of get the sense that everybody is excited about AI and we have a unique perspective in the sense that we probably follow it a little closer than most. 41 00:02:41,560 --> 00:02:52,640 But we found this video by the corporate bro on TikTok that I think perfectly sums up the interest in AI, specifically generative AI that businesses have. 42 00:02:52,640 --> 00:02:54,480 So I'm going to just play this right now. 43 00:02:54,480 --> 00:02:56,560 And I think everyone will get the sentiment. 44 00:02:56,560 --> 00:03:06,520 OK, so you've all seen the numbers and it's bad, but don't worry, because even though we burned through almost 30 million dollars in just under six months, I have a plan to raise more money and save this company. 45 00:03:06,520 --> 00:03:08,560 Oh, God, please don't say crypto again. 46 00:03:08,560 --> 00:03:11,480 No, this is even better. 47 00:03:11,480 --> 00:03:14,120 A.I. 48 00:03:14,120 --> 00:03:17,640 Come on in, AI. 49 00:03:17,640 --> 00:03:22,600 Greetings, humans. 50 00:03:22,600 --> 00:03:23,600 I thought you might like that. 51 00:03:23,600 --> 00:03:24,800 So like, what's it do? 52 00:03:24,800 --> 00:03:27,760 To be honest, we don't know, but it is generative. 53 00:03:27,760 --> 00:03:32,400 We can call it business GPT GPT. 54 00:03:32,400 --> 00:03:39,840 While an excellent, excellent ad there for sales GPT, it does capture the sentiment. 55 00:03:39,840 --> 00:03:41,120 That is a real thing. 56 00:03:41,120 --> 00:03:42,120 Sales GPT. 57 00:03:42,120 --> 00:03:43,680 Yeah, I'm Googling. 58 00:03:43,680 --> 00:03:48,240 So you go to feverishly type and looks it up right now for us. 59 00:03:48,240 --> 00:03:48,760 Really? 60 00:03:48,760 --> 00:03:51,200 I'm free, free plug care of that video. 61 00:03:51,200 --> 00:03:52,280 What are you finding here? 62 00:03:52,280 --> 00:03:53,000 Let me. 63 00:03:53,000 --> 00:03:54,840 But it is so true. 64 00:03:54,840 --> 00:04:02,600 And oh my God, it's just you replace in this video AI with anything else and you still 65 00:04:02,600 --> 00:04:06,640 get the same script, blockchain, crypto, NFTs. 66 00:04:06,640 --> 00:04:10,160 But the significant difference is in the technology. 67 00:04:10,160 --> 00:04:16,280 And, you know, if you look at NFTs versus AI, the idea of artificial intelligence been 68 00:04:16,280 --> 00:04:19,480 for hundreds of years around and people have been working on it. 69 00:04:19,480 --> 00:04:25,320 So it's just that it exploded right now and I think the big impact maybe historically 70 00:04:25,320 --> 00:04:28,680 is going to be that chat GPT was given to the public. 71 00:04:28,680 --> 00:04:34,680 Therefore, like before, like, you know, that was GPT three was in the playground, but it 72 00:04:34,680 --> 00:04:40,360 was not really widespread because it was not placed or packaged in a way that most users 73 00:04:40,360 --> 00:04:41,160 can use. 74 00:04:41,160 --> 00:04:42,920 But this technology existed. 75 00:04:42,920 --> 00:04:48,280 Well, I think to me, the distinction between any of the generative AI technologies and 76 00:04:48,280 --> 00:04:54,520 something like blockchain, NFTs is the barrier to entry for like the casual person tends 77 00:04:54,520 --> 00:04:59,640 to be much higher, kind of hard to conceptualize what a net fungible token is, what it even 78 00:04:59,640 --> 00:05:03,080 grants you, how it exists, how you even have it. 79 00:05:03,080 --> 00:05:07,240 You need to get a special wallet on a special type of website with keys that you have to 80 00:05:07,240 --> 00:05:09,640 write down and never lose or something like that. 81 00:05:09,640 --> 00:05:11,080 This is generative AI. 82 00:05:11,080 --> 00:05:12,040 Just take that. 83 00:05:12,040 --> 00:05:13,960 Type me a letter, write me a resume. 84 00:05:13,960 --> 00:05:17,640 You know, I think it's a lot easier for people to kind of understand and think about some 85 00:05:17,640 --> 00:05:24,280 of the possibilities, but the kind of underlying thesis of that video is not at all far removed 86 00:05:24,280 --> 00:05:25,480 from reality. 87 00:05:25,480 --> 00:05:30,200 Take, for example, some of the gentlemen over at the company known as minstrel AI. 88 00:05:30,200 --> 00:05:32,200 These folks here to 70. 89 00:05:32,200 --> 00:05:32,920 Not minstrel. 90 00:05:32,920 --> 00:05:33,480 Minstrel. 91 00:05:33,480 --> 00:05:33,980 Yeah. 92 00:05:35,080 --> 00:05:35,580 Sorry. 93 00:05:37,160 --> 00:05:38,200 Mistral. 94 00:05:38,200 --> 00:05:40,600 Minstrel would be a bit something else. 95 00:05:40,600 --> 00:05:41,160 Mistrial. 96 00:05:43,400 --> 00:05:44,440 I like minstrel. 97 00:05:44,440 --> 00:05:47,960 Like, isn't that like a wandering minstrel, a court jester? 98 00:05:47,960 --> 00:05:48,920 That makes sense to me. 99 00:05:49,640 --> 00:05:49,880 Okay. 100 00:05:49,880 --> 00:05:50,360 My girls. 101 00:05:50,360 --> 00:05:50,760 Okay. 102 00:05:50,760 --> 00:05:51,240 Mistral. 103 00:05:51,240 --> 00:05:55,000 If there is any girls listening, they will get why I'm laughing. 104 00:05:55,000 --> 00:05:56,760 But yeah, let's keep going. 105 00:05:56,760 --> 00:05:57,260 Yeah. 106 00:05:57,260 --> 00:05:58,600 So, mistral. 107 00:05:59,320 --> 00:06:04,920 These folks gained a $74 million investment or evaluation of their company rather. 108 00:06:04,920 --> 00:06:06,200 I think it's six people. 109 00:06:06,200 --> 00:06:07,640 They have zero users. 110 00:06:07,640 --> 00:06:10,600 They have zero MVP. 111 00:06:10,600 --> 00:06:12,040 They have nothing built. 112 00:06:12,040 --> 00:06:17,400 They have a seven page memo written on a Google Doc that secured them $74 million in funding. 113 00:06:17,400 --> 00:06:19,160 Now, here's the thing. 114 00:06:19,160 --> 00:06:24,520 This is, I think, an instance where it's just like, all right, yes, there is fever, craziness 115 00:06:24,520 --> 00:06:27,800 happening to invest in anything generative AI. 116 00:06:27,800 --> 00:06:33,560 But if you look at some of the people that are associated with this, they are six folks 117 00:06:33,560 --> 00:06:40,280 that have some compliance, some deep experience with LLMs, things like DeepMind, the llama 118 00:06:40,280 --> 00:06:41,800 model at Meta. 119 00:06:41,800 --> 00:06:47,000 So, it's almost just like, all right, this is evaluation based on your resume, what you've 120 00:06:47,000 --> 00:06:52,600 done in the past, which is an interesting note to me, at least, because it goes, all 121 00:06:52,600 --> 00:06:56,680 right, well, this is a case of where people's like, all right, the tech is, it's going to 122 00:06:56,680 --> 00:06:57,560 work out. 123 00:06:57,560 --> 00:07:00,200 So, I'm just investing in you. 124 00:07:00,200 --> 00:07:04,680 Is that something that, did you perceive it the same way when you heard about this story? 125 00:07:04,680 --> 00:07:06,040 In a way, yes. 126 00:07:06,040 --> 00:07:09,960 And I worked in Accelerator and Venture Capital before. 127 00:07:09,960 --> 00:07:16,600 So, the underlying thesis when investing from, you know, especially Accelerators is that 128 00:07:16,600 --> 00:07:21,880 you're investing in the founders, investing in a team who is going to be behind the startup. 129 00:07:21,880 --> 00:07:25,000 So, in this way, yes, it reflects. 130 00:07:25,000 --> 00:07:32,280 Another aspect is how to say this, that same as you are investing in some startups who 131 00:07:32,280 --> 00:07:38,200 are promising to build models and the majority of your investment is going to computing 132 00:07:38,200 --> 00:07:44,360 power, so you're literally buying startup who will give money to the NVIDIA. 133 00:07:44,360 --> 00:07:51,240 I think in this way, this is just puts a price tag on this type of profile people or developers 134 00:07:51,240 --> 00:07:52,600 with this type of resume. 135 00:07:52,600 --> 00:07:59,000 So, I 100% agree with you, but it's very interesting dynamic going forward that if 136 00:07:59,000 --> 00:08:06,440 we start seeing people who work in these fields at such scale value, if they just come together, 137 00:08:06,440 --> 00:08:13,080 and that's enough to raise money, the future will show and will tell that if I would be 138 00:08:13,080 --> 00:08:15,320 the investor, I would never do this. 139 00:08:15,320 --> 00:08:18,360 I would never put my money just on resumes. 140 00:08:18,360 --> 00:08:19,000 I completely agree with you. 141 00:08:21,000 --> 00:08:22,280 On resumes alone. 142 00:08:22,280 --> 00:08:22,920 You're totally right. 143 00:08:22,920 --> 00:08:26,760 I'm glad you brought up compute because, I mean, I think what was it? 144 00:08:26,760 --> 00:08:33,320 OpenAI said that GPT 3.5 was $100 million to train in compute time. 145 00:08:33,320 --> 00:08:36,760 So it's just like, all right, well, you have a $75 million valuation. 146 00:08:36,760 --> 00:08:42,840 This company, Mistral, is trying to be a competitor to Bard and Chad GPT, specifically for the 147 00:08:42,840 --> 00:08:45,000 European market, from my understanding. 148 00:08:45,000 --> 00:08:50,680 Yes, they already know that it's going to be a huge monumental task to train, but I also 149 00:08:50,680 --> 00:08:55,560 think it's interesting that part of their whole business model is this is going to be 150 00:08:55,560 --> 00:08:57,160 an open source model. 151 00:08:57,160 --> 00:09:01,320 So to me, it's like, all right, you're going to put this out there, which I agree with 152 00:09:01,320 --> 00:09:05,720 to a point, and I think we've talked about this on past podcasts, but one of the problems 153 00:09:05,720 --> 00:09:09,880 with that is then you're going to have a harder time monetizing it. 154 00:09:09,880 --> 00:09:13,640 Stable Diffusion and Stability AI is a perfect example right now. 155 00:09:13,640 --> 00:09:20,840 They have incredible genre shattering technology, but they don't have a customer base the same 156 00:09:20,840 --> 00:09:21,240 way. 157 00:09:21,240 --> 00:09:27,400 If I put my marketing hat on, this move was brilliant from that point of view because 158 00:09:27,400 --> 00:09:28,840 now we are talking about it too. 159 00:09:28,840 --> 00:09:34,760 But I don't know if you remember then there was this conversation that Chad GPT is direct 160 00:09:34,760 --> 00:09:41,560 competitor to Google and then web browsing search is going to be next big thing and Google 161 00:09:41,560 --> 00:09:42,680 didn't do it yet. 162 00:09:42,680 --> 00:09:44,440 Bing was not yet there. 163 00:09:44,440 --> 00:09:49,960 I saw so many quite notable investors talking about this search engine, Neva. 164 00:09:49,960 --> 00:09:53,000 Did you know that they just got, you know, bankrupt? 165 00:09:53,000 --> 00:09:57,800 I don't want to say bankrupt because it's placed on mergers and acquisitions, so it's 166 00:09:57,800 --> 00:09:58,120 fine. 167 00:09:58,120 --> 00:10:04,040 But they are not anymore in the field and you don't see this in the news that hey, 168 00:10:04,040 --> 00:10:09,800 the startup came, collected investment, tried to do something, but then big players just 169 00:10:09,800 --> 00:10:12,520 a month later maybe boom have it. 170 00:10:12,520 --> 00:10:18,520 So I think there is two coins that we don't hear enough about all the startups, plugins. 171 00:10:18,520 --> 00:10:23,720 I think I talked about it in my video mentioning that all this, you know, email extensions 172 00:10:23,720 --> 00:10:24,600 and stuff like that. 173 00:10:24,600 --> 00:10:29,000 We are charging money, but how they are going to compete. 174 00:10:29,000 --> 00:10:30,760 I want to see that strategy. 175 00:10:30,760 --> 00:10:36,840 I want to see and probably in this document is outlined how they are planning to do that. 176 00:10:37,560 --> 00:10:43,160 But this amount of money on something that we plan to do something like that, there is 177 00:10:43,160 --> 00:10:48,280 probably another 50 startups who have exactly the same strategy and the plan, maybe not 178 00:10:48,280 --> 00:10:52,840 even Google Doc, maybe at least Pitch Deck talking about Pitch Deck. 179 00:10:52,840 --> 00:10:55,000 Yeah, maybe they at least printed it out. 180 00:10:55,000 --> 00:10:57,000 But yeah, that's actually a great point. 181 00:10:57,000 --> 00:11:01,000 You know, I mean, it even says in the bottom here, they expect to need to raise another 182 00:11:01,000 --> 00:11:06,360 200 million on top of their evaluation just to be able to create a model that can compete 183 00:11:06,360 --> 00:11:07,560 with GPT-4. 184 00:11:07,560 --> 00:11:11,000 To raise cash like that, you probably need a Pitch Deck, am I right? 185 00:11:11,720 --> 00:11:12,040 Yes. 186 00:11:12,680 --> 00:11:20,600 So pivoting to the Pitch Decks, the interesting finding was done by Clarify Capital that GPT-4 187 00:11:20,600 --> 00:11:25,560 outperforms humans in Pitch Deck effectiveness among investors and business owners. 188 00:11:25,560 --> 00:11:31,240 So basically, if you are a startup and you are making your Pitch Deck, your likelihood 189 00:11:31,240 --> 00:11:36,200 to get an investor is higher if you use GPT-4 to make it and write it. 190 00:11:36,200 --> 00:11:42,440 And it's like one in five investors and business owners pitched by GPT-4 would invest 10,000 191 00:11:42,440 --> 00:11:43,400 or more. 192 00:11:43,400 --> 00:11:51,080 And with this Pitch Deck from Mistral, I actually went and copied their Google Doc text. 193 00:11:51,080 --> 00:11:54,840 So I took the text, ran it by AI content detectors. 194 00:11:54,840 --> 00:11:59,880 And yes, I know they are not good, but I was like at least looking like, okay, what's going 195 00:11:59,880 --> 00:12:01,000 to be the spectrum? 196 00:12:01,640 --> 00:12:02,760 And it was human. 197 00:12:02,760 --> 00:12:03,480 What did you get? 198 00:12:03,480 --> 00:12:04,520 You did the same, right? 199 00:12:05,240 --> 00:12:10,440 Well, I did it from the sense of, you know, for my business, I'm like, let's take this. 200 00:12:10,440 --> 00:12:16,360 If they're getting 74 million, we will take the same format and use it for white papers 201 00:12:16,360 --> 00:12:18,040 for some of our tools we're building. 202 00:12:18,600 --> 00:12:20,520 I mean, that's just the math right there. 203 00:12:20,520 --> 00:12:21,480 You know, all right. 204 00:12:21,480 --> 00:12:27,560 If this, then that, you know, if this gets you a million dollar investment without a 205 00:12:27,560 --> 00:12:33,000 product and some smart people behind it, hey, I know some smart people and we do have some 206 00:12:33,000 --> 00:12:38,600 things, you know, so maybe it'll work the same, but I'll probably need a GPT-4 written Pitch 207 00:12:38,600 --> 00:12:40,440 Deck to go along with it. 208 00:12:40,440 --> 00:12:47,720 If we take a step back and on the full serious note, there is going to be so much money lost 209 00:12:47,720 --> 00:12:48,840 from the investors. 210 00:12:48,840 --> 00:12:51,160 And this is what is my fear. 211 00:12:51,160 --> 00:12:56,840 The part like the sentiment AI gets, if that happens and the startups start falling behind 212 00:12:56,840 --> 00:13:03,240 and bankrupting and all that, the media will take it, hopefully, or I expect that media 213 00:13:03,240 --> 00:13:04,920 will make a big deal out of that. 214 00:13:04,920 --> 00:13:10,520 And at the end of the day, it's going to be the casual user sentiment towards the technology, 215 00:13:10,520 --> 00:13:12,040 which is going to be damaged. 216 00:13:12,040 --> 00:13:19,480 And that's a bit sad because this technology is incredible and making it so sensationalized. 217 00:13:19,480 --> 00:13:22,520 I don't think it's ever a good thing to do. 218 00:13:22,520 --> 00:13:23,160 Yeah. 219 00:13:23,160 --> 00:13:27,800 I mean, it's seldom do you have something, though, that literally everyone can use or 220 00:13:27,800 --> 00:13:29,400 benefit from in such a way. 221 00:13:29,400 --> 00:13:33,560 So, I mean, it's only natural to have dozens of people trying to plant their flag on the 222 00:13:33,560 --> 00:13:35,000 same thing. 223 00:13:35,000 --> 00:13:38,680 It's definitely a gold rush, but not everyone can strike it rich. 224 00:13:38,680 --> 00:13:44,280 And I think the unfortunate thing for us as consumers is if everyone's using the same 225 00:13:44,280 --> 00:13:50,840 technology, but for different services, hey, to make a pitch deck, to generate some content 226 00:13:50,840 --> 00:13:53,080 for a blog, to write marketing emails. 227 00:13:53,080 --> 00:13:58,680 If everyone's using the same technology, it's not going to be the person with the best tech 228 00:13:58,680 --> 00:14:02,680 or the best implementation that rises to the top and becomes the market leader. 229 00:14:02,680 --> 00:14:07,400 It's probably going to be the person that can market it the best and tell their story 230 00:14:07,400 --> 00:14:13,400 the best, which might mean we as consumers are going, all right, well, I'm conditioned 231 00:14:13,400 --> 00:14:18,680 now to sort by five stars and pick the thing with the most downloads or the most reviews. 232 00:14:18,680 --> 00:14:23,320 I'm going to pick that one, even though it doesn't do what I need it to. 233 00:14:23,320 --> 00:14:28,440 We'll be the first one to say in the generative presentation market, there's things like 234 00:14:28,440 --> 00:14:33,080 Gamma, there's a few other different competitors. 235 00:14:33,080 --> 00:14:35,480 I think it's read.ai. 236 00:14:35,480 --> 00:14:40,360 We'll add them in the newsletter, some of the different AI-based generative slides, 237 00:14:40,360 --> 00:14:42,240 generative presentation markets. 238 00:14:42,240 --> 00:14:47,560 The market leaders got 20 or 30 times more than the one that I think is the best, which 239 00:14:47,560 --> 00:14:48,880 is Gamma. 240 00:14:48,880 --> 00:14:53,080 And that's a perfect example of a case where the people with the best technology, the people 241 00:14:53,080 --> 00:14:56,680 leveraging the AI the best aren't the ones winning at this point. 242 00:14:56,680 --> 00:15:02,880 Right, I hope that at the end of the day, market wins and market tells the truth, like 243 00:15:02,880 --> 00:15:05,200 what people use and value. 244 00:15:05,200 --> 00:15:09,560 And from the video we shared, there is this aspect like, but what does it do? 245 00:15:09,560 --> 00:15:11,240 What can they actually do? 246 00:15:11,240 --> 00:15:17,560 And they hear it so many times that even as basic as charge BT, that yeah, it's great. 247 00:15:17,560 --> 00:15:22,640 Like everyone talks about it, even here in Germany, there is even in the local newspaper 248 00:15:22,640 --> 00:15:25,040 title all over the place, right? 249 00:15:25,040 --> 00:15:30,640 And then people test it and they go like, yeah, I didn't get what I expected. 250 00:15:30,640 --> 00:15:33,160 So what should I do with this? 251 00:15:33,160 --> 00:15:36,000 Is it even good enough or is just the hype? 252 00:15:36,000 --> 00:15:37,680 And then just leave it at that. 253 00:15:37,680 --> 00:15:43,160 So I think there is going to be some time in the market to see who is emerging at the 254 00:15:43,160 --> 00:15:44,160 end. 255 00:15:44,160 --> 00:15:48,200 I like to compare this as something to the Excel sheet. 256 00:15:48,200 --> 00:15:49,800 Everyone knows Excel sheet. 257 00:15:49,800 --> 00:15:55,240 But everyone is an expert in the Excel sheet, but if you have to pull data or do some math 258 00:15:55,240 --> 00:16:00,480 or, I don't know, just store some information, you will go through probably Excel sheets 259 00:16:00,480 --> 00:16:02,680 or Google sheets, right? 260 00:16:02,680 --> 00:16:08,320 So I'm curious which of these models will become this tool in your stack deck that every 261 00:16:08,320 --> 00:16:10,920 business wants to come to work for a new company. 262 00:16:10,920 --> 00:16:13,140 They will be like, okay, we use Google docs. 263 00:16:13,140 --> 00:16:14,720 We use this AI tool. 264 00:16:14,720 --> 00:16:17,960 We use this, I don't know, like conferencing call. 265 00:16:17,960 --> 00:16:21,360 I'm not sure if chat GPT is one of those. 266 00:16:21,360 --> 00:16:22,360 What do you think? 267 00:16:22,360 --> 00:16:25,680 I think we talked about it a little bit in one of the prior podcasts, but they did a 268 00:16:25,680 --> 00:16:34,240 study of several different AI tools and terms of the big language models against a large 269 00:16:34,240 --> 00:16:35,480 subsection of the population. 270 00:16:35,480 --> 00:16:38,440 So which ones of these have you even just heard about? 271 00:16:38,440 --> 00:16:44,320 From StableDiffusion to GPT-3 to Bard to Lama, any of these other ones. 272 00:16:44,320 --> 00:16:50,160 Chat GPT, by far and away, is the one that's most proliferated through the social zeitgeist. 273 00:16:50,160 --> 00:16:54,360 I think 60% of people had heard about it or used it. 274 00:16:54,360 --> 00:17:02,000 I think it was around 50 to 58 that heard of it and 14 used it, which was a huge surprise 275 00:17:02,000 --> 00:17:03,360 to me, right? 276 00:17:03,360 --> 00:17:04,360 Wow. 277 00:17:04,360 --> 00:17:12,560 What I think is interesting about that is chat GPT is a fine-tuned version of the GPT-3.5 278 00:17:12,560 --> 00:17:14,280 or 4 model. 279 00:17:14,280 --> 00:17:20,760 It's tuned in such a way to elicit these very natural sounding responses, and that's it. 280 00:17:20,760 --> 00:17:24,600 You don't get some of the other controls you get when you're actually trying to pull the 281 00:17:24,600 --> 00:17:27,920 levers of the GPT-4 model itself. 282 00:17:27,920 --> 00:17:33,520 I'm amazed when I talk to potential clients and say, hey, we started to build this AI 283 00:17:33,520 --> 00:17:34,520 tool. 284 00:17:34,520 --> 00:17:38,840 We use the chat GPT API, and then we can't get it to output factual information. 285 00:17:38,840 --> 00:17:46,080 I'm like, well, why are you using the thing that's tuned to be more interesting and human 286 00:17:46,080 --> 00:17:49,800 sounding when you're asking for this deterministic factual response? 287 00:17:49,800 --> 00:17:52,760 It's because that's just the thing that they know about. 288 00:17:52,760 --> 00:17:55,640 Hey, I want the AI shiny thing. 289 00:17:55,640 --> 00:18:01,160 Give me the chat GPT API into my business and make it so. 290 00:18:01,160 --> 00:18:05,840 It means we got plenty of work coming down the pike. 291 00:18:05,840 --> 00:18:12,160 Yeah, you have a really good point because here in Germany, when I was in this tech conference, 292 00:18:12,160 --> 00:18:17,680 one of the speakers said that they were actually in conversations with German government for 293 00:18:17,680 --> 00:18:19,560 developing their own model. 294 00:18:19,560 --> 00:18:21,500 And government was like, we are in. 295 00:18:21,500 --> 00:18:24,440 We want to do this and we have money. 296 00:18:24,440 --> 00:18:30,400 But one prerequisite for that is we need the biggest companies to buy in that they will 297 00:18:30,400 --> 00:18:34,360 prioritize using this model over anyone else. 298 00:18:34,360 --> 00:18:35,680 And that didn't go through. 299 00:18:35,680 --> 00:18:40,640 At least at that point, because I don't want to name the biggest companies in Germany, 300 00:18:40,640 --> 00:18:48,440 but I guess you know, automotive banks, very famous banks that they were like, no, we will 301 00:18:48,440 --> 00:18:53,960 use chat GPT for open AI, but we don't want our own model. 302 00:18:53,960 --> 00:18:55,580 That was very interesting for me. 303 00:18:55,580 --> 00:18:58,360 And he was like completely disappointed, shocked. 304 00:18:58,360 --> 00:19:03,320 And they are trying to push this forward and kind of change the sentiment. 305 00:19:03,320 --> 00:19:09,080 But maybe there is certain fear of being too late to the game, especially also when GPT 306 00:19:09,080 --> 00:19:11,960 for is marketed as the best, the best. 307 00:19:11,960 --> 00:19:17,560 And you want the best because maybe if you wait for government, German government to 308 00:19:17,560 --> 00:19:22,880 develop their own model, go through the whole set of regulations, compliance, you know, 309 00:19:22,880 --> 00:19:27,880 we could be looking at two, three years, including training at best. 310 00:19:27,880 --> 00:19:33,840 Yeah, so if they have a product right now in the market, one aspect is to jump on it. 311 00:19:33,840 --> 00:19:37,720 Another aspect is to shop around what are the alternatives. 312 00:19:37,720 --> 00:19:42,920 This is a perfect example though of like, why people are excited to just, you know, 313 00:19:42,920 --> 00:19:44,620 they feel like they don't want to miss the boat. 314 00:19:44,620 --> 00:19:45,980 I see it all the time. 315 00:19:45,980 --> 00:19:49,920 You have probably a very significant first mover advantage. 316 00:19:49,920 --> 00:19:55,280 If you just kind of find your niche and exploit it and get your tool, get your model, get 317 00:19:55,280 --> 00:19:59,360 your AI idea kind of built and made and out there. 318 00:19:59,360 --> 00:20:05,180 And that is a pervasive sentiment that I think a lot of other businesses and business owners 319 00:20:05,180 --> 00:20:09,400 are sharing that is not something that government needs to embrace. 320 00:20:09,400 --> 00:20:16,040 I remember I saw this with first hundred OpenAI enterprise customers and you have the biggest 321 00:20:16,040 --> 00:20:20,000 players like I saw Salesforce, LinkedIn there. 322 00:20:20,000 --> 00:20:24,200 So there is also this, you know, selling point, who is your customer. 323 00:20:24,200 --> 00:20:28,360 So if I am as a business and they know that, okay, my competitors in the market are using 324 00:20:28,360 --> 00:20:35,320 OpenAI, so do I go the same route or do I actually look for alternative? 325 00:20:35,320 --> 00:20:38,040 But what if there is no better alternative right now? 326 00:20:38,040 --> 00:20:39,040 But there will be. 327 00:20:39,040 --> 00:20:43,840 Like I am very hopeful that this is not going to be market capitalization. 328 00:20:43,840 --> 00:20:49,360 And as a business owner, if I would be sitting in those shoes, I would be very closely also 329 00:20:49,360 --> 00:20:54,520 looking at what's happening with regulations, but also what's happening with lawsuits. 330 00:20:54,520 --> 00:21:00,100 This is what I shared with you that Child GPT Maker OpenAI faces class action over how 331 00:21:00,100 --> 00:21:02,340 it used people's data. 332 00:21:02,340 --> 00:21:08,040 And there's a couple of these coming out and it's not from Europe, it's from California, 333 00:21:08,040 --> 00:21:12,880 the firm called KlarSyn and they have experiences with these types of lawsuits. 334 00:21:12,880 --> 00:21:15,800 So that came out just a couple of days ago. 335 00:21:15,800 --> 00:21:22,560 So how all of this pans out and this, I don't know how connected this is, but Child GPT 336 00:21:22,560 --> 00:21:25,860 removed web browsing from the plus users. 337 00:21:25,860 --> 00:21:31,160 And I'm very curious to hear your opinion on that because I was filming a video and 338 00:21:31,160 --> 00:21:33,800 I filmed the video over two days. 339 00:21:33,800 --> 00:21:38,680 And on the one day I was like, oh, let me show you how to make a LinkedIn post where 340 00:21:38,680 --> 00:21:43,640 you insert link, use web browsing and then it researches, pulls data and creates your 341 00:21:43,640 --> 00:21:44,760 LinkedIn post. 342 00:21:44,760 --> 00:21:46,400 The next day I want to film, it's gone. 343 00:21:46,400 --> 00:21:47,960 And I was like, oh, right. 344 00:21:47,960 --> 00:21:51,320 I'm paying money and I purchased it. 345 00:21:51,320 --> 00:21:55,840 Then GPT-4 came out, but of course web browsing was one of those features too. 346 00:21:55,840 --> 00:22:00,020 And now it's just removed with nothing, with no lowering price, nothing. 347 00:22:00,020 --> 00:22:01,520 So what is your take on this? 348 00:22:01,520 --> 00:22:02,520 Yeah. 349 00:22:02,520 --> 00:22:08,000 I mean, I think you brought up the most crucial part of it from a consumer's perspective where 350 00:22:08,000 --> 00:22:12,440 it's just like, hey, I'm paying for this service where this is advertised. 351 00:22:12,440 --> 00:22:14,120 Why am I not getting it? 352 00:22:14,120 --> 00:22:18,960 I think OpenAI though is pretty transparent that like, Hey, this is a lot of this stuff 353 00:22:18,960 --> 00:22:20,960 is very experimental. 354 00:22:20,960 --> 00:22:24,520 Some of this is even in like alpha stages of development. 355 00:22:24,520 --> 00:22:25,800 I mean, I don't know. 356 00:22:25,800 --> 00:22:31,320 To me, I probably have not found all the best use cases for web browsing right within chat 357 00:22:31,320 --> 00:22:32,320 GPT. 358 00:22:32,320 --> 00:22:38,280 If I have workflows or use cases for that, there's other ways to build that type of stuff 359 00:22:38,280 --> 00:22:45,040 in with things like web scrapers or using some Python code in conjunction with the GPT-4 360 00:22:45,040 --> 00:22:46,480 API. 361 00:22:46,480 --> 00:22:49,120 So I mean, like it's, is it going to come back? 362 00:22:49,120 --> 00:22:51,160 Yeah, it's probably going to come back. 363 00:22:51,160 --> 00:22:53,200 It's just how long, who knows? 364 00:22:53,200 --> 00:22:54,200 Yeah. 365 00:22:54,200 --> 00:22:58,200 I actually agree with you that my experience with web browsing was half of the time it 366 00:22:58,200 --> 00:23:04,200 says can't read content, can't read content, failed, failed, scrolling down, failed. 367 00:23:04,200 --> 00:23:11,680 So if we are pulling it out, fixing it, but then there is another point of this paradox 368 00:23:11,680 --> 00:23:17,160 that we have Bing, which runs on GPT-4 and it has web browsing and then there's chat 369 00:23:17,160 --> 00:23:19,860 GPT who does exactly the same, but you pay. 370 00:23:19,860 --> 00:23:23,920 So I'm curious on three fronts that what you are saying. 371 00:23:23,920 --> 00:23:27,400 I think it's very optimistic view that we are nice guys. 372 00:23:27,400 --> 00:23:28,640 We're testing the product. 373 00:23:28,640 --> 00:23:30,480 We're pulling it back to improve. 374 00:23:30,480 --> 00:23:35,440 Then there's politics between two companies and two competing kind of offerings. 375 00:23:35,440 --> 00:23:39,600 And then there's another aspect of regulations and conversations on that front. 376 00:23:39,600 --> 00:23:40,600 So I don't know. 377 00:23:40,600 --> 00:23:41,600 We'll see. 378 00:23:41,600 --> 00:23:46,120 I mean, the difference though being when you have Bing, your underlying architecture is 379 00:23:46,120 --> 00:23:48,040 that of a page rank algorithm. 380 00:23:48,040 --> 00:23:52,520 That's what Google got famous for back in the day where it's measuring the importance 381 00:23:52,520 --> 00:23:54,520 of the information on pages. 382 00:23:54,520 --> 00:23:59,840 In addition to the keywords that you're putting in, it's just overlaying the language models 383 00:23:59,840 --> 00:24:06,560 capability to interface with that page ranked service that is Bing, that is Google. 384 00:24:06,560 --> 00:24:12,400 I think the difference from OpenAI's perspective, if you're in chat GPT, you're using the language 385 00:24:12,400 --> 00:24:17,980 model that has this little capability of potentially going out and searching the web, but probably 386 00:24:17,980 --> 00:24:19,280 not in the same way. 387 00:24:19,280 --> 00:24:22,480 Because it's not, I would venture to say it's not using a page rank algorithm. 388 00:24:22,480 --> 00:24:27,460 It's using stuff like keyword or depth for search or some of these other ways that we've 389 00:24:27,460 --> 00:24:33,960 come to index data because in and of itself, page rank is a very, very compute heavy, very 390 00:24:33,960 --> 00:24:39,560 taxing type of thing to do because Google's the number one web browser because they have 391 00:24:39,560 --> 00:24:44,240 gone out and developed these tools that index every page on the internet and find it and 392 00:24:44,240 --> 00:24:45,240 index it. 393 00:24:45,240 --> 00:24:49,840 So to be able to look through that quickly is not something a language model is built 394 00:24:49,840 --> 00:24:50,840 to do. 395 00:24:50,840 --> 00:24:55,400 I just thought about another interesting thing, prompt hacking. 396 00:24:55,400 --> 00:25:00,460 I remember I came across this research paper or made the whole video about it where they 397 00:25:00,460 --> 00:25:07,240 inserted prompt hack and prompt injection into that page and then Bing went on that 398 00:25:07,240 --> 00:25:14,680 webpage and got infected and then the hell broke loose and I started asking for users' 399 00:25:14,680 --> 00:25:20,760 bank information for email, for name and then created a scammy link where you click and 400 00:25:20,760 --> 00:25:22,820 you're basically done. 401 00:25:22,820 --> 00:25:30,840 So our aspect is something also could be happening behind the doors but kind of the dark side 402 00:25:30,840 --> 00:25:36,720 of all of this, what if more and more we started finding out these injections on the market? 403 00:25:36,720 --> 00:25:43,020 Well you would hope that you wouldn't have to worry about these big, supposedly ethical 404 00:25:43,020 --> 00:25:49,400 parent companies not really kind of engaging in activity that could be considered a little 405 00:25:49,400 --> 00:25:51,320 skeezy or dark side of the web. 406 00:25:51,320 --> 00:25:56,240 I think this is a perfect opportunity to discuss some of the changes that Google made to their 407 00:25:56,240 --> 00:25:58,000 privacy policies this week. 408 00:25:58,000 --> 00:26:04,580 Yes, it is a couple of days ago Google changed their privacy policy and I can just read one 409 00:26:04,580 --> 00:26:11,480 statement which let's just jump into and discuss that Android maker will use this data which 410 00:26:11,480 --> 00:26:16,720 is that users share with others for the sake of training its AI model. 411 00:26:16,720 --> 00:26:22,360 So this is not the data you kind of leave on the internet but also the data that you 412 00:26:22,360 --> 00:26:28,080 share with other people and they placed everything in a positive light because of course we know 413 00:26:28,080 --> 00:26:34,280 you better as a user therefore we personalize the services and we improve our services. 414 00:26:34,280 --> 00:26:42,600 So on that spectrum yes but on another spectrum what choice as a user do we even have anymore? 415 00:26:42,600 --> 00:26:44,400 I would hope they give us the choice. 416 00:26:44,400 --> 00:26:45,920 Yeah it's totally lose-lose. 417 00:26:45,920 --> 00:26:50,760 I would hope they at least give us the choice like OpenAI does where it says you say hey 418 00:26:50,760 --> 00:26:54,920 I don't want to use any of my responses for future training data. 419 00:26:54,920 --> 00:26:59,800 There's a little lever you can flip to do that inside of chat GPT. 420 00:26:59,800 --> 00:27:04,000 I would expect that has to be a required feature going forward. 421 00:27:04,000 --> 00:27:08,120 I mean I don't know how I feel about this because it has to be because then you're going 422 00:27:08,120 --> 00:27:13,220 to anytime you have a lot of data and you can extract machine learning type insights 423 00:27:13,220 --> 00:27:17,340 from it I mean that's what this is going to become and then they're going to sell those 424 00:27:17,340 --> 00:27:23,160 insights to advertisers and it'll be even more like hey I just had a thought about this 425 00:27:23,160 --> 00:27:25,760 thing and you're going to get an Instagram ad for it. 426 00:27:25,760 --> 00:27:27,080 I just had a thought about it. 427 00:27:27,080 --> 00:27:33,720 Well it's because they know that someone with your age race gender financial situation you 428 00:27:33,720 --> 00:27:37,240 know because you saw this thing on the news a week ago you're going to be thinking about 429 00:27:37,240 --> 00:27:41,520 this product now you know like you're not be able to draw those kind of connections 430 00:27:41,520 --> 00:27:45,160 which is kind of wild I would say. 431 00:27:45,160 --> 00:27:47,800 It's happening to me now all the time. 432 00:27:47,800 --> 00:27:54,160 I don't know what I've done because I usually try never accept anything reject everything 433 00:27:54,160 --> 00:28:00,320 but now we have a conversation and go if I go on Instagram that's it I get ads for exactly 434 00:28:00,320 --> 00:28:01,320 the same thing. 435 00:28:01,320 --> 00:28:02,320 It's spooky. 436 00:28:02,320 --> 00:28:09,280 One of the most intrusive ways of those those those Google search ad words coming back in 437 00:28:09,280 --> 00:28:11,320 the form of advertisements. 438 00:28:11,320 --> 00:28:16,680 If you ever want to really send an indication to like some like your partner or boyfriend 439 00:28:16,680 --> 00:28:22,600 girlfriend just get on their phone and just search for diamond engagement rings. 440 00:28:22,600 --> 00:28:30,160 You will get so ad spammed on every single platform from brilliant earth rare diamonds 441 00:28:30,160 --> 00:28:32,160 just to no end. 442 00:28:32,160 --> 00:28:34,640 I was looking for a piece of jewelry for my wife. 443 00:28:34,640 --> 00:28:37,120 It's our 10th anniversary coming up later this year. 444 00:28:37,120 --> 00:28:42,080 I made that mistake just even just searching jewelry and it's been months of just that's 445 00:28:42,080 --> 00:28:45,240 all I it's every other thing on Instagram. 446 00:28:45,240 --> 00:28:48,640 So I mean I'm almost like well I don't have a choice at this point. 447 00:28:48,640 --> 00:28:55,680 I have a real story which is on a bit different spectrum and this is like probably 10 years 448 00:28:55,680 --> 00:28:57,960 ago in the architecture firm. 449 00:28:57,960 --> 00:29:03,460 One of the guys we had workshop one of the guys came and used you know office computer 450 00:29:03,460 --> 00:29:12,640 but he searched about babies because wife who is in the same firm is pregnant. 451 00:29:12,640 --> 00:29:18,920 So he just did some search and then basically the office on the same network started getting 452 00:29:18,920 --> 00:29:25,160 pumpers and then it was literally like as a joke like hey like is anyone pregnant and 453 00:29:25,160 --> 00:29:30,280 as being a girl in architecture firm that's you know you have maybe two girls in the firm 454 00:29:30,280 --> 00:29:36,960 and it's kind of evident that hey what's going on and then yeah that was the case that one 455 00:29:36,960 --> 00:29:39,400 of the girls was pregnant. 456 00:29:39,400 --> 00:29:43,640 The firm knew it before they came announcing that. 457 00:29:43,640 --> 00:29:47,880 There's a famous article famous situation that is talked about in a lot of data science 458 00:29:47,880 --> 00:29:50,320 classes for machine learning. 459 00:29:50,320 --> 00:29:55,880 Target was one of the first bigger businesses to implement a lot of different ML tactics 460 00:29:55,880 --> 00:30:00,660 and they basically got it down to the point where if you're a woman and you're buying 461 00:30:00,660 --> 00:30:06,800 any combination of these 20 products you know two or two or three or more of these products 462 00:30:06,800 --> 00:30:12,660 there is a very high likelihood chance that this person is pregnant or has a new baby. 463 00:30:12,660 --> 00:30:18,280 So Target had a lawsuit because they started sending a bunch of ads to a family and a father 464 00:30:18,280 --> 00:30:23,360 was getting these ads for hey for your come do your baby registry at Target hey here's 465 00:30:23,360 --> 00:30:27,920 a bunch of things like that well it turns out his teenage daughter was with child and 466 00:30:27,920 --> 00:30:32,640 he didn't know yet but Target did because of some AI so and that's that story is from 467 00:30:32,640 --> 00:30:33,640 10 years ago. 468 00:30:33,640 --> 00:30:39,280 Right so these tools definitely know is better than probably our closest people but if you 469 00:30:39,280 --> 00:30:45,240 pull what I sent to you I found this graph to be incredibly interesting. 470 00:30:45,240 --> 00:30:51,480 This is kind of like a quick I would say sketch grading foundation models providers compliance 471 00:30:51,480 --> 00:30:54,200 with the draft EU AI Act. 472 00:30:54,200 --> 00:30:59,840 If you look at the scaling so this is like okay there is 48 parameters and you can see 473 00:30:59,840 --> 00:31:05,120 each company on what kind of level we are so the interesting thing was Bloom. 474 00:31:05,120 --> 00:31:11,200 I'm really curious to look into them more how Google compares to OpenAI which is absolutely 475 00:31:11,200 --> 00:31:17,680 close 27 out of 48 versus 25 which is OpenAI at 48. 476 00:31:17,680 --> 00:31:25,480 Autropic cloud seven out of 48 the ones who we see in the market as the ones claiming 477 00:31:25,480 --> 00:31:32,040 that this is you know one of those models which are exactly thank you very much. 478 00:31:32,040 --> 00:31:37,480 I don't know what what is your thoughts and also luminous which is I think in Germany 479 00:31:37,480 --> 00:31:42,520 based and completely not compliant at the lowest level. 480 00:31:42,520 --> 00:31:47,160 To the flip side of that coin I think is well there is no standard to measure this stuff 481 00:31:47,160 --> 00:31:52,960 against this might be the first standard so when you have kind of that that yardstick 482 00:31:52,960 --> 00:31:58,600 okay then people can start to track to that you would hope that a lot of businesses and 483 00:31:58,600 --> 00:32:03,000 these founders and creators of these language models would have all of these things at the 484 00:32:03,000 --> 00:32:07,320 forefront of their mind but I imagine they're also trying to get the science and the tech 485 00:32:07,320 --> 00:32:12,400 right and compete with one another in this arms race basically and I think they also 486 00:32:12,400 --> 00:32:18,880 recognize it is probably a trade-off whereas if you have all of these types of compliance 487 00:32:18,880 --> 00:32:24,920 levels built into your model is it as capable as it was before or do you have to develop 488 00:32:24,920 --> 00:32:29,720 the technology to be compliant which doesn't push the tech forward. 489 00:32:29,720 --> 00:32:36,440 I was a couple years ago where I wanted to say like I think it was Apple there was one 490 00:32:36,440 --> 00:32:42,100 year where it crossed over where they had to spend more money on lawsuits against people 491 00:32:42,100 --> 00:32:47,880 like suing them like patent trolls and people saying that they had infringed on their patents 492 00:32:47,880 --> 00:32:52,880 or their copyrights some may be the case absolutely true they just spend more money on lawsuits 493 00:32:52,880 --> 00:32:57,320 than they did on research and development in any given year so like to me like that's 494 00:32:57,320 --> 00:33:01,960 a indication of like all right this is not moving the technology forward when we're having 495 00:33:01,960 --> 00:33:06,760 to deal with these type of compliance things so I imagine the founders of a lot of these 496 00:33:06,760 --> 00:33:11,740 companies and the people working there they want to push the tech forward but and kind 497 00:33:11,740 --> 00:33:16,000 of view some of these things as secondary the compliance aspects. 498 00:33:16,000 --> 00:33:21,960 Yes but also like just for you know listeners you can see this graph on YouTube and we will 499 00:33:21,960 --> 00:33:26,820 include it in the newsletter but this is done by Stanford research on foundational models 500 00:33:26,820 --> 00:33:33,480 and for me what was interesting with this they rank each major lab on multiple different 501 00:33:33,480 --> 00:33:39,240 factors so for example their data source data governance copyrighted data compute energy 502 00:33:39,240 --> 00:33:43,760 capabilities and limitations which is kind of like a pattern where majority of these 503 00:33:43,760 --> 00:33:51,840 models actually stand out this also shows for example copyrighted data only two switches 504 00:33:51,840 --> 00:34:00,040 bloom and gpt neox which have compliance at some sort of level for me it shows that what 505 00:34:00,040 --> 00:34:05,920 it is possible to have clean data sources data governments copyrighted data you know 506 00:34:05,920 --> 00:34:10,960 in your model and still have a really great model because bloom is one of those models 507 00:34:10,960 --> 00:34:17,840 which have almost if not every single language inside the model it's it's tough because 508 00:34:17,840 --> 00:34:22,800 cleaning cleaning that data doesn't make you any money doesn't help get the thing to market 509 00:34:22,800 --> 00:34:27,280 any faster you know it just doesn't it's kind of one of those things we're like hey it's 510 00:34:27,280 --> 00:34:32,800 better to ask for forgiveness than permission a lot of the times I love how from two different 511 00:34:32,800 --> 00:34:37,960 worlds we are this is great to have these discussions with you because it's required 512 00:34:37,960 --> 00:34:43,720 to have these different perspectives what I see looking at this graph versus you so 513 00:34:43,720 --> 00:34:48,560 I bet you see also like lobbying exactly what you said like all this loss that's coming 514 00:34:48,560 --> 00:34:56,000 out like just draining money from the innovation from my side I'm like okay so can we do innovation 515 00:34:56,000 --> 00:35:01,560 you know considering users considering copyrighted data is it actually possible and if it is 516 00:35:01,560 --> 00:35:07,560 possible can we sustain this slower rate because we can't keep up with it anyways to be honest 517 00:35:07,560 --> 00:35:13,200 no yeah I suppose I mean I think it does highlight maybe a couple things because as someone who's 518 00:35:13,200 --> 00:35:19,480 trying to play with these tools and do new and different stuff in a professional capacity 519 00:35:19,480 --> 00:35:26,520 semi-professional maybe capacity to me I don't think there's ever been a major technological 520 00:35:26,520 --> 00:35:32,720 innovation where government put out their regulations first and then you know we've 521 00:35:32,720 --> 00:35:38,520 reached far beyond it I'm just trying to think of the last couple major major ones you know 522 00:35:38,520 --> 00:35:43,500 I mean shoot we still haven't even regulated or done anything about social media or the 523 00:35:43,500 --> 00:35:48,800 internet probably a couple things there's still copyright laws that govern tons of internet 524 00:35:48,800 --> 00:35:55,520 content that are from the time when we only had records and radio broadcasts I appreciate 525 00:35:55,520 --> 00:36:01,840 your perspective and yeah that's true you know same as well yours is ours well thank 526 00:36:01,840 --> 00:36:08,760 you okay so Wes is this place any as good to end the podcast I think this is a place 527 00:36:08,760 --> 00:36:14,280 as good as any so for go to go this is Wes Synthmind saying happy prompting everybody 528 00:36:14,280 --> 00:36:21,880 happy prompting everybody thanks for listening to how to talk to AI with your hosts go to 529 00:36:21,880 --> 00:36:31,720 go and Wes the synth mind as always you can check out the show notes and links at howtotalkto.ai 530 00:36:31,720 --> 00:36:52,240 that's all for this week's episode happy prompting everyone.