The honest scorecard for what AI can actually do

00:00:00,200 --> 00:00:35,699 [Eric] [upbeat music] Welcome back to the Token Intelligence Show. John, one of the questions that you and I get a lot is, what can AI actually do? Which sounds like a really simple question, and I think to you and I, because we use AI all day, every day, is easy to gloss over. But it's actually kind of a hard question to answer, and you've been answering it a lot this week. 00:00:35,700 --> 00:00:35,940 [John] Right. 00:00:37,080 --> 00:00:47,100 [John] It's super hard to answer. And, and the difference between what people think it can do and what it can actually do varies to a huge degree, and- 00:00:47,100 --> 00:00:47,560 [Eric] Yes 00:00:47,560 --> 00:00:53,780 [John] ... it varies on both sides of overconfidence and not either under-confidence or just not knowing. 00:00:53,780 --> 00:00:54,000 [Eric] Yep. 00:00:54,000 --> 00:00:55,700 [John] So it's, it's so interesting. 00:00:55,700 --> 00:01:02,940 [Eric] So what are we going to accomplish today? Are we going to teach people what AI can actually do and not do? Is that what we're gonna walk away with? 00:01:02,940 --> 00:01:05,740 [John] We're, we're gonna do something fun. I've got five points. 00:01:05,740 --> 00:01:05,780 [Eric] Okay. 00:01:05,780 --> 00:01:07,980 [John] I'm gonna have you react live to. 00:01:07,980 --> 00:01:08,560 [Eric] Okay, great. 00:01:08,560 --> 00:01:11,980 [John] And then, um, I'll tell you if I [laughs] agree or disagree with your- 00:01:11,980 --> 00:01:12,539 [Eric] Perfect 00:01:12,539 --> 00:01:24,480 [John] ... reaction. But no, I- this is, um, this will be really fun because I've got the whole spectrum from, for sure, doing this today, and both of us do it, to, like, some out there stuff. 00:01:24,480 --> 00:01:31,490 [Eric] Okay, great. Should we start with, uh, not to insult your intelligence- 00:01:31,490 --> 00:01:31,490 [John] [laughs] 00:01:31,490 --> 00:01:41,430 [Eric] ... or my intelligence, and especially our listeners' intelligence, should we start with trying to define what AI actually is? Like, what is the mechanism? 00:01:41,430 --> 00:01:41,460 [John] Mm. Yeah. 00:01:41,460 --> 00:01:43,500 [Eric] Do we wanna start there? 00:01:43,500 --> 00:01:48,200 [John] Yeah, I think it's worth starting there, um, because I think it's broad. 00:01:48,200 --> 00:01:48,350 [Eric] Yep. 00:01:48,350 --> 00:01:56,360 [John] 'Cause AI to some people is, um... I used to use Google, and I Google things in ChatGPT. 00:01:56,360 --> 00:01:56,600 [Eric] Mm-hmm. 00:01:56,600 --> 00:02:04,250 [John] And AI to other people is I have a army of agents or AI employees doing, you know- 00:02:04,250 --> 00:02:04,250 [Eric] Right 00:02:04,250 --> 00:02:05,200 [John] ... hundreds of tasks. 00:02:05,200 --> 00:02:06,140 [Eric] Right. Right. 00:02:06,140 --> 00:02:08,240 [John] Whether they're useful tasks or not, we'll get into in a minute. 00:02:08,240 --> 00:02:11,859 [Eric] Okay, great. Uh, do you wanna give your definition, or do you want me to start? 00:02:11,860 --> 00:02:12,080 [John] Yeah. 00:02:13,240 --> 00:02:29,160 [John] So I mean, I think the best, the best common definition that I can get to, um, is... I mean, the, the mechanics of it is where I would start. So you're sending text through a model provider, a large language model- 00:02:29,160 --> 00:02:31,420 [Eric] So you're sending it to OpenAI or Anthropic- 00:02:31,420 --> 00:02:31,900 [John] Or Anthropic 00:02:31,900 --> 00:02:32,720 [Eric] ... or xAI, et cetera. 00:02:32,720 --> 00:02:35,540 [John] Right. And then you're getting a response back. 00:02:35,540 --> 00:02:35,920 [Eric] Mm-hmm. 00:02:35,920 --> 00:02:43,960 [John] And there are a million different ways that... And two, and like, Claude, Gem- Gemini, ChatGPT, OpenClau, they're all different, 00:02:44,980 --> 00:02:46,240 [John] you know, versions of that. 00:02:46,240 --> 00:02:46,360 [Eric] Yep. 00:02:46,360 --> 00:02:54,420 [John] And then I think the places people are seeing, like, the value leveling up is connecting that to one or more other things- 00:02:54,420 --> 00:02:54,680 [Eric] Mm-hmm 00:02:54,680 --> 00:02:56,620 [John] ... like software, other pieces of software. 00:02:56,620 --> 00:02:57,340 [Eric] Yep. 00:02:57,340 --> 00:03:07,160 [John] And then the other interesting component there is, um, the ability that it can actually code really well. Um, so there's a... That, that's kind of another angle. 00:03:07,160 --> 00:03:07,640 [Eric] Yep. 00:03:07,640 --> 00:03:09,340 [John] Um, w- how would you define it? 00:03:10,540 --> 00:03:20,240 [Eric] The most helpful analogy that I've used in the many that I've tried has been starting with Google Translate, because I think that's- 00:03:20,240 --> 00:03:20,250 [John] Mm 00:03:20,250 --> 00:03:21,600 [Eric] ... a very accessible- 00:03:21,600 --> 00:03:22,360 [John] Yeah 00:03:22,360 --> 00:03:44,220 [Eric] ... uh, that's a very accessible experience for people. So you can start typing into Google Translate, you know, and, and traditionally on the web, you know, the mobile app's slightly different, but you would start typing a sentence, right? So if I'm traveling to France or Italy and I wanna ask where the bathroom is- 00:03:44,220 --> 00:03:44,340 [John] Yes 00:03:44,340 --> 00:03:53,579 [Eric] ... Google Translate, you can say, you know, "Where is the..." and you can type R, and it will generally, like, auto-complete the word restroom- 00:03:53,580 --> 00:03:54,560 [John] Hmm. Okay 00:03:54,560 --> 00:03:58,049 [Eric] ... and then translate that into Italian. And so a couple of things- 00:03:58,049 --> 00:03:58,080 [John] Right 00:03:58,080 --> 00:03:58,880 [Eric] ... are happening there. 00:04:00,060 --> 00:04:01,320 [Eric] One is that 00:04:02,360 --> 00:04:06,560 [Eric] Google Translate is predicting the next word- 00:04:06,560 --> 00:04:06,760 [John] Right 00:04:06,760 --> 00:04:08,040 [Eric] ... that you're going to say. 00:04:08,040 --> 00:04:08,380 [John] Right. 00:04:08,380 --> 00:04:09,810 [Eric] And how is it doing that? Well, 00:04:10,960 --> 00:04:19,519 [Eric] you know, they're, they have, you know, hundreds of millions, uh, or trillions of examples of people typing into Google, "How do you say, uh-" 00:04:19,519 --> 00:04:19,550 [John] Right 00:04:19,550 --> 00:04:20,300 [Eric] ... "Where's the bathroom-" 00:04:20,300 --> 00:04:20,790 [John] Yeah. Yeah. 00:04:20,790 --> 00:04:20,959 [Eric] "... in Italian?" 00:04:22,050 --> 00:04:22,060 [John] Right. 00:04:22,060 --> 00:04:27,810 [Eric] And so they have an immense amount of data that, that they can use to auto-complete that phrase for you. 00:04:27,810 --> 00:04:28,610 [John] Sure. Right. 00:04:28,610 --> 00:04:35,440 [Eric] And then they can essentially real-time translate that into Italian, right? And so there are a couple things happening under the hood. 00:04:36,580 --> 00:04:36,920 [Eric] And 00:04:38,340 --> 00:04:47,280 [Eric] you can think about AI as doing that same thing, but for almost any context. 00:04:47,280 --> 00:04:47,460 [John] Right. 00:04:47,460 --> 00:04:49,700 [Eric] Right? So, um, 00:04:51,300 --> 00:05:00,760 [Eric] let's say that I wanna write an email to my team talking about how we need to really buckle down for the end of the quarter, right? 00:05:01,880 --> 00:05:02,180 [Eric] Well, 00:05:03,480 --> 00:05:10,990 [Eric] the Frontier Labs have ingested hundreds of millions, maybe hundreds of mi- I don't know, hundreds of millions of emails- 00:05:10,990 --> 00:05:10,990 [John] Right 00:05:10,990 --> 00:05:17,870 [Eric] ... or examples of emails. Every blog post out there on how to write a good email to your team, they have ingested all of that information- 00:05:17,870 --> 00:05:17,870 [John] Yeah 00:05:17,870 --> 00:05:18,170 [Eric] ... from the internet. 00:05:18,170 --> 00:05:19,520 [John] All the HubSpot content about writing emails. 00:05:19,520 --> 00:05:27,910 [Eric] All of the HubSpot content about writing emails, leadership, uh, team building, clear communication, all of that. It's, they've scraped- 00:05:27,910 --> 00:05:28,000 [John] Yeah 00:05:28,000 --> 00:05:29,020 [Eric] ... the entire internet, right? 00:05:29,020 --> 00:05:29,980 [John] Right. Right. 00:05:29,980 --> 00:05:35,440 [Eric] And so now, instead of just, you know, "Where's the bathroom?" Um, 00:05:36,960 --> 00:05:42,920 [Eric] now instead of just, "Where's the bathroom?" They have the entire context of communicating with your team. 00:05:42,920 --> 00:05:43,110 [John] Right. 00:05:43,110 --> 00:05:49,900 [Eric] Right? And so when you say, "I wanna write an email to my team," they can essentially auto-complete that, like Google Translate, right? 00:05:49,900 --> 00:05:49,920 [John] Yeah. Right. 00:05:49,920 --> 00:05:52,500 [Eric] They're just using a huge amount of context, right? 00:05:52,500 --> 00:05:52,520 [John] Right. 00:05:52,520 --> 00:05:57,120 [Eric] And so that's kind of how I explain AI, and it's, it's really subject agnostic- 00:05:57,120 --> 00:05:57,130 [John] Right 00:05:57,130 --> 00:05:59,310 [Eric] ... because you have the entire knowledge of the internet, right? 00:05:59,310 --> 00:05:59,310 [John] Right. 00:05:59,310 --> 00:05:59,720 [Eric] And then 00:06:00,760 --> 00:06:07,636 [Eric] probably the way that I would complete the picture is that-That's a very broad level, right? 00:06:07,636 --> 00:06:07,656 [John] Right. 00:06:07,656 --> 00:06:15,786 [Eric] And so AI is very open-ended in that I can ask him how to write an email to my team. I can ask him how to write a blog post on, you know, how to write- 00:06:15,786 --> 00:06:15,815 [John] Right 00:06:15,815 --> 00:06:17,026 [Eric] ... an email to my team. [laughs] 00:06:17,026 --> 00:06:18,216 [John] Yeah. Right. There you go. 00:06:18,216 --> 00:06:20,916 [Eric] Uh, you know, or whatever it is, right? Um, 00:06:22,156 --> 00:06:27,816 [Eric] how to build a project on Vercel, how to, you know, wire an AI d- agent up to Snowflake- 00:06:27,816 --> 00:06:28,016 [John] Right 00:06:28,016 --> 00:06:32,216 [Eric] ... you know, to create, uh, you know, a data bot or something, right? 00:06:32,216 --> 00:06:32,936 [John] Sure. 00:06:32,936 --> 00:06:41,196 [Eric] And it has ingested so much information that it, it can auto-complete those things. But where it gets really powerful is when you give it specific context. 00:06:41,196 --> 00:06:41,376 [John] Right. 00:06:41,376 --> 00:07:15,456 [Eric] So you can say, "Hey, here's a, here are a bunch of conversations that I've had with my team. Here are, like, one-on-one transcriptions or meeting notes with everyone on my team. Here's the quarterly plan for the entire marketing organization that my boss sent me. So I want you to take that and then help me write an email to my team." And the results tend to be really, really good because you're not asking it to auto-complete on all of the information on the internet, you know, or in the world. You're asking it to auto-complete on things that are very, very contextual. 00:07:15,456 --> 00:07:16,316 [John] Right. 00:07:16,316 --> 00:07:18,496 [Eric] And it's extremely good at that. 00:07:18,496 --> 00:07:18,775 [John] Right. 00:07:18,776 --> 00:07:31,476 [Eric] Uh, and code is a great example, right? Or writing SQL or whatever, because it's very contextual. Here's a project, here's a report, here's a database with a bunch of tables, and the auto-complete, quote-unquote, gets really good. 00:07:31,476 --> 00:07:31,616 [John] Yeah. 00:07:31,616 --> 00:07:33,006 [Eric] So that was very long, but that's- 00:07:33,006 --> 00:07:33,006 [John] No, I like that 00:07:33,006 --> 00:07:33,956 [Eric] ... how I explain it. 00:07:33,956 --> 00:07:47,236 [John] The, so the thing that made me think of, you know, you going through your definition is one that I think both of us have heard before, which the translation absolutely. And think about English to code, code to English. 00:07:47,236 --> 00:07:47,316 [Eric] Mm-hmm, mm-hmm. 00:07:47,316 --> 00:07:51,306 [John] English to an- you know, to another language, to Italian, Italian back to English. 00:07:51,306 --> 00:07:51,376 [Eric] Yep. 00:07:51,376 --> 00:07:57,716 [John] Um, the other thing, in the most simple terms, is it's very good at reading really quickly and writing really quickly. 00:07:57,716 --> 00:07:59,596 [Eric] Yes. It can do it instantly. 00:07:59,596 --> 00:08:01,116 [John] Yeah, yeah. 00:08:01,116 --> 00:08:01,396 [Eric] So, 00:08:02,776 --> 00:08:11,496 [Eric] uh, that's a great point. I mean, on a daily basis, I will give an AI model 50 pages of source material, 00:08:12,916 --> 00:08:16,836 [Eric] and in 30 seconds it can, it can read the context- 00:08:16,836 --> 00:08:16,976 [John] Right 00:08:16,976 --> 00:08:21,186 [Eric] ... and then take, you know, take action on the context. 00:08:21,186 --> 00:08:21,216 [John] Right. 00:08:21,216 --> 00:08:22,236 [Eric] It's unbelievable. 00:08:22,236 --> 00:08:22,776 [John] Yep, yep. 00:08:22,776 --> 00:08:23,136 [Eric] Okay. 00:08:23,136 --> 00:08:28,856 [John] All right. So here's, here's, here's my five, um, things. The first one, so what can AI actually do? 00:08:28,856 --> 00:08:29,436 [Eric] Yep. 00:08:29,436 --> 00:08:34,716 [John] We're gonna start easy, okay? Um, and you react to it, and I'll fill in some gaps here. 00:08:34,716 --> 00:08:35,056 [Eric] Great. 00:08:35,056 --> 00:08:37,356 [John] Um, web research and deep analysis. 00:08:39,976 --> 00:08:43,796 [Eric] Excellent. I, I would... Can I give a rating scale here? 00:08:43,796 --> 00:08:44,396 [John] Yeah. 00:08:44,396 --> 00:08:45,136 [Eric] I would- 00:08:45,136 --> 00:08:45,776 [John] One to, one to 10? 00:08:45,776 --> 00:08:48,036 [Eric] Yeah. I would say, um, 00:08:49,336 --> 00:08:51,396 [Eric] I would give it an eight out of 10, 00:08:52,996 --> 00:08:53,216 [Eric] uh, 00:08:55,076 --> 00:08:56,046 [Eric] in the current state. 00:08:56,046 --> 00:08:56,056 [John] Yeah. 00:08:56,056 --> 00:08:59,196 [Eric] It used to be horrible, AI used to be horrible at web research- 00:08:59,196 --> 00:08:59,336 [John] Mm-hmm 00:08:59,336 --> 00:09:02,466 [Eric] ... but the tools that it uses have, have become very sophisticated- 00:09:02,466 --> 00:09:02,466 [John] Yeah 00:09:02,466 --> 00:09:04,716 [Eric] ... in the last year. Um, 00:09:06,176 --> 00:09:06,676 [Eric] and, 00:09:07,976 --> 00:09:20,196 [Eric] uh, the... So you can ask a pretty open-ended question, and it will, like a, the AI, you know, agent or Anthropic- 00:09:20,196 --> 00:09:20,206 [John] Whatever 00:09:20,206 --> 00:09:22,216 [Eric] ... GPT, whatever you're using- 00:09:22,216 --> 00:09:22,226 [John] Yeah 00:09:22,226 --> 00:09:25,316 [Eric] ... it will actually issue a bunch of queries. And so the way- 00:09:25,316 --> 00:09:25,326 [John] Right 00:09:25,326 --> 00:09:29,695 [Eric] ... that I think about this is... And you can actually look under the hood if you want to- 00:09:29,696 --> 00:09:29,706 [John] Right 00:09:29,706 --> 00:09:30,656 [Eric] ... to see what's going on. 00:09:30,656 --> 00:09:31,375 [John] Right. 00:09:31,376 --> 00:09:33,146 [Eric] But let's just say that, um, 00:09:34,796 --> 00:09:40,756 [Eric] I wanna ask a question about the MacBook Neo that Apple just released. 00:09:40,756 --> 00:09:40,976 [John] Yeah. Right. 00:09:40,976 --> 00:09:48,906 [Eric] Right? And so if I was going to do some research on that, right? Like, I wonder why it seems popular. Is it popular? 00:09:48,906 --> 00:09:48,916 [John] Right. 00:09:48,916 --> 00:09:50,536 [Eric] If so, why is it popular? 00:09:50,536 --> 00:09:50,726 [John] Right. 00:09:50,726 --> 00:09:52,196 [Eric] Right? Um, 00:09:53,676 --> 00:09:57,416 [Eric] I would normally just Google, like, I would just start on Google, right? In- 00:09:57,416 --> 00:09:57,616 [John] Sure 00:09:57,616 --> 00:09:58,975 [Eric] ... in sort of the old world, right? 00:09:58,976 --> 00:09:59,256 [John] Right, right. 00:09:59,256 --> 00:10:01,176 [Eric] Like, how many of the, of these are selling- 00:10:01,176 --> 00:10:01,186 [John] Right 00:10:01,186 --> 00:10:02,916 [Eric] ... and sort of issue a couple Google queries, right? 00:10:04,136 --> 00:10:06,616 [Eric] But if I ask an AI model, 00:10:07,736 --> 00:10:10,625 [Eric] you know, "The MacBook Neo seems to be popular. 00:10:11,796 --> 00:10:16,116 [Eric] Is that true? And then if so, why is that true?" 00:10:16,116 --> 00:10:16,125 [John] Right. 00:10:16,125 --> 00:10:21,556 [Eric] "And I'm interested in that, like, because of the current economic climate," right? 00:10:21,556 --> 00:10:21,576 [John] Right. 00:10:21,576 --> 00:10:30,816 [Eric] Like gas prices are rising or whatever it is. And instead of me doing three or four Google searches and trying to comb through links- 00:10:30,816 --> 00:10:31,156 [John] Right 00:10:31,156 --> 00:10:35,115 [Eric] ... it will issue 30 or 40 Google searches. 00:10:35,116 --> 00:10:35,236 [John] Right. 00:10:36,356 --> 00:10:37,566 [Eric] Google s- or let's say web searches. 00:10:37,566 --> 00:10:38,836 [John] Yeah, web searches. Yeah. 00:10:38,896 --> 00:10:45,626 [Eric] It scans all the links. It tries to find the most relevant links. It will actually visit those web pages. It will, like, read all of the content on each- 00:10:45,626 --> 00:10:45,626 [John] Right 00:10:45,626 --> 00:10:46,546 [Eric] ... of those web pages 00:10:47,676 --> 00:10:51,216 [Eric] and determine the most relevant content. 00:10:51,216 --> 00:10:51,636 [John] Right. 00:10:51,636 --> 00:11:03,756 [Eric] Uh, and then return that back as a summary and cite the sources, which is great. And so if you extrapolate that out for a given topic, deep research has become way, way faster. Um- 00:11:03,756 --> 00:11:04,176 [John] Yeah. 00:11:04,176 --> 00:11:04,816 [Eric] Way, way faster. 00:11:05,956 --> 00:11:20,226 [John] Yeah. Yeah. Yeah, I would definitely agree on that one. I actu- So when it first came out, like, people were really excited about it. Um, I would vote this probably around an, an eight as well, but as far as, like, capability. 00:11:20,226 --> 00:11:20,786 [Eric] Mm-hmm. 00:11:20,786 --> 00:11:25,555 [John] But I think it maybe is falling down a little bit as far as popularity. 00:11:25,556 --> 00:11:25,836 [Eric] Hmm. 00:11:25,836 --> 00:11:32,856 [John] Um, and this is just actually really simple. Um, so if you have, um, Claude or GPT on, on your phone- 00:11:32,856 --> 00:11:32,946 [Eric] Mm-hmm 00:11:32,946 --> 00:11:33,766 [John] ... or use it on your computer. 00:11:33,766 --> 00:11:34,476 [Eric] Mm-hmm. 00:11:34,476 --> 00:11:39,476 [John] Um, there's a button that you hit to tell it to spend more time researching a thing. 00:11:39,476 --> 00:11:40,396 [Eric] Yes. Yeah. 00:11:40,396 --> 00:11:47,676 [John] I, I, I, I mean, I know about this. I use it, uh, especially, like, at work. I forget to use it. 00:11:47,676 --> 00:11:47,726 [Eric] Yeah. 00:11:47,726 --> 00:11:51,036 [John] 'Cause it, 'cause if you tell it, like, search the web, like, it'll do it, but it'll- 00:11:51,036 --> 00:11:51,396 [Eric] Mm-hmm 00:11:51,396 --> 00:11:52,836 [John] ... like, just search a couple things. 00:11:52,836 --> 00:11:52,896 [Eric] Right. 00:11:52,896 --> 00:11:56,196 [John] But you have to, like, flip the switch to get it to, like, search a bunch of things. 00:11:56,196 --> 00:11:56,576 [Eric] Yep. 00:11:56,576 --> 00:11:57,476 [John] I forget to turn it on, honestly. 00:11:57,476 --> 00:12:01,216 [Eric] And that's not available on every plan. You actually have to pay a lot more- 00:12:01,216 --> 00:12:01,226 [John] Yeah 00:12:01,226 --> 00:12:01,956 [Eric] ... for that. 00:12:01,956 --> 00:12:08,026 [John] Yeah. So I think that's a really interesting thing. Like, if you're on a free tier, to your point, you might not have it. And even me- 00:12:08,026 --> 00:12:08,036 [Eric] Yep 00:12:08,036 --> 00:12:10,536 [John] ... who, like, is paying for it, totally forget to just- 00:12:10,536 --> 00:12:10,686 [Eric] Yeah. I agree 00:12:10,686 --> 00:12:11,856 [John] ... flip the switch. Um- 00:12:11,856 --> 00:12:21,416 [Eric] But yeah, that's interesting. And that's a, I think that's a really good point forAnyone who has tried this and felt like, "Eh, AI is not great at that." 00:12:21,416 --> 00:12:22,316 [John] Mm-hmm. 00:12:22,316 --> 00:12:27,086 [Eric] Uh, one is that on the, on the cheaper free tiers, it's gonna be worse. 00:12:27,086 --> 00:12:27,096 [John] Right. 00:12:27,096 --> 00:12:29,216 [Eric] And the reason is because it's an economic- 00:12:29,216 --> 00:12:29,366 [John] It's expensive, yeah 00:12:29,366 --> 00:12:31,776 [Eric] ... it, it's expensive for them to do that. 00:12:31,776 --> 00:12:37,875 [John] And just an order of magnitude thing, um, I, I hadn't used it as much recently. I've used it a couple things this week. 00:12:37,876 --> 00:12:38,636 [Eric] Mm-hmm. 00:12:38,636 --> 00:12:44,596 [John] It'll visit thru- I think I did something this week, it visited 383 individual websites. 00:12:44,596 --> 00:12:45,376 [Eric] Yes. It's incredible. 00:12:45,376 --> 00:12:46,596 [John] It's, it's incredible. 00:12:46,596 --> 00:12:47,296 [Eric] Yeah. 00:12:47,296 --> 00:12:51,556 [John] And, and last time I re- like remember looking... 'Cause often you just ignore what it's doing. [chuckles] 00:12:51,556 --> 00:12:52,036 [Eric] Right. Exactly. 00:12:52,036 --> 00:12:53,396 [John] The last time I dug into that- 00:12:53,396 --> 00:12:53,576 [Eric] Mm-hmm 00:12:53,576 --> 00:12:55,036 [John] ... it was, it was probably dozens. 00:12:55,036 --> 00:12:55,336 [Eric] Yep. 00:12:55,336 --> 00:12:56,796 [John] But I mean, hundreds of sites. 00:12:56,796 --> 00:12:57,796 [Eric] It's very useful. 00:12:57,796 --> 00:12:58,036 [John] Yeah. 00:12:58,036 --> 00:13:00,956 [Eric] But it's not cheap, and you do have to be intentional. 00:13:00,956 --> 00:13:01,956 [John] Yep. 00:13:01,956 --> 00:13:02,216 [Eric] All right. 00:13:02,216 --> 00:13:05,556 [John] All right. Next one. We're gonna go the full opposite end of the spectrum. 00:13:05,556 --> 00:13:05,816 [Eric] Great. 00:13:05,816 --> 00:13:15,476 [John] All right. Ready? So number one was web research and deep analysis. Think we both agree around an 8 out of 10. Number two, and this has been splashy press headline- 00:13:15,476 --> 00:13:15,756 [Eric] Mm-hmm 00:13:15,756 --> 00:13:22,155 [John] ... YouTubers, running a fully autonomous one-person company. [laughs] 00:13:26,396 --> 00:13:27,816 [Eric] O- one or two out of 10- 00:13:27,816 --> 00:13:28,026 [John] Okay. Yeah 00:13:28,026 --> 00:13:29,176 [Eric] ... I'm going to say. 00:13:29,176 --> 00:13:29,496 [John] Yeah. 00:13:29,496 --> 00:13:31,296 [Eric] One or two out of 10. Uh, 00:13:33,176 --> 00:13:33,646 [Eric] the, 00:13:35,876 --> 00:13:40,016 [Eric] the challenge actually goes back to... 00:13:42,936 --> 00:13:49,116 [Eric] It, it, it goes back to the input problem that, that we talked about when we were, when we were defining what AI is. 00:13:49,116 --> 00:13:50,236 [John] Right. 00:13:50,236 --> 00:13:56,096 [Eric] And if you give an AI model really specific context- 00:13:56,096 --> 00:13:56,656 [John] Right 00:13:56,656 --> 00:14:00,576 [Eric] ... it can do the, quote-unquote, "auto-complete" really, really well. 00:14:00,576 --> 00:14:01,016 [John] Yep. 00:14:01,016 --> 00:14:01,436 [Eric] And 00:14:03,096 --> 00:14:07,336 [Eric] the challenge is... I think there are two challenges. 00:14:08,866 --> 00:14:16,416 [Eric] The simple challenge is the economics of trying to ingest the entire context of the world into an AI model- 00:14:16,416 --> 00:14:16,426 [John] Right 00:14:16,426 --> 00:14:17,736 [Eric] ... and the cost of that. 00:14:17,736 --> 00:14:17,956 [John] Right. 00:14:17,956 --> 00:14:18,116 [Eric] Right? 00:14:19,336 --> 00:14:20,556 [Eric] So you 00:14:21,716 --> 00:14:48,066 [Eric] could theoretically, you know, have an agent that's scanning the news for a very specific set of topics and ingesting that context and looking at it, you know, and, you know, reading every transcription from every call in the company. And y- they're just, you know... But it... That would be extr- it would be pr- it would be so expensive that it wouldn't be worth the amount of, like, the, the cost wouldn't be worth it. 00:14:48,066 --> 00:14:48,276 [John] Yeah. 00:14:48,276 --> 00:14:48,516 [Eric] Right? 00:14:48,516 --> 00:14:49,396 [John] Right. 00:14:49,396 --> 00:14:55,635 [Eric] But all of that information is what humans can naturally ingest as we go throughout day-to-day stuff, especially if you're paying attention. 00:14:55,636 --> 00:14:55,765 [John] Right. 00:14:55,765 --> 00:14:55,795 [Eric] Right? 00:14:55,796 --> 00:14:56,316 [John] Yeah. 00:14:56,316 --> 00:15:05,966 [Eric] Um, and that's what it takes to make good decisions with a business, because, you know, like, people who are great businesspeople, 00:15:07,036 --> 00:15:15,376 [Eric] it... I mean, so often what I see is that they are just good at reading a lot of context and putting it together and making a great decision, right? 00:15:15,376 --> 00:15:16,276 [John] Right. 00:15:16,276 --> 00:15:19,436 [Eric] And AI cannot do that, 00:15:20,536 --> 00:15:22,476 [Eric] um, in the current state. 00:15:22,476 --> 00:15:22,896 [John] Right. 00:15:22,896 --> 00:15:24,135 [Eric] It just can't do that in the current state. 00:15:24,136 --> 00:15:24,816 [John] No. 00:15:24,816 --> 00:15:27,756 [Eric] And so that's like the basic, uh... 00:15:29,196 --> 00:15:42,736 [Eric] Yeah, that's the basic economic side of it, like, you, you just can't ingest all that information. And then I think the other, the other barrier is that, um, 00:15:44,496 --> 00:15:47,385 [Eric] the context within a business changes so, so quickly, 00:15:48,596 --> 00:15:53,716 [Eric] and that's a real challenge for AI models, right? Like, if you, 00:15:55,256 --> 00:16:00,566 [Eric] i- if you think about a one particular task that needs to be done, right? So, 00:16:02,576 --> 00:16:12,636 [Eric] uh, an AI agent needs to perform an analysis, or for me, an AI agent needs to, like, do a bunch of research and, like, generate a draft or an outline that I can review for, you know, whatever. Um, 00:16:13,796 --> 00:16:26,856 [Eric] and incredibly powerful within that sort of defined, like, narrow scope. But if you multiply out, that out across all of the decisions that need to be made, you know, across the roles in even a very small business- 00:16:26,856 --> 00:16:27,226 [John] Right 00:16:27,226 --> 00:16:34,836 [Eric] ... and the amount of context that's needed for that, beyond even the expense, like, the context with all these different variables is constantly changing. 00:16:34,836 --> 00:16:35,295 [John] Right. 00:16:35,295 --> 00:16:46,526 [Eric] The priorities of the business are changing. And that's really, really difficult, and I don't think that's gonna change. I, I don't, I don't necessarily see that as a problem that, you know, AI is going to overcome- 00:16:46,526 --> 00:16:46,556 [John] Right 00:16:46,556 --> 00:16:50,096 [Eric] ... one day. That's just the chaos of, of living in our world. 00:16:50,096 --> 00:16:54,936 [John] Yep. So I'm gonna answer at a one and a three, depending on the situation. 00:16:54,936 --> 00:16:55,336 [Eric] Okay. 00:16:55,336 --> 00:16:58,696 [John] One is almost every business, one or zero. 00:16:58,696 --> 00:16:59,276 [Eric] Mm-hmm. 00:16:59,276 --> 00:17:00,416 [John] Like, not a thing. 00:17:00,416 --> 00:17:01,236 [Eric] Yep. 00:17:01,236 --> 00:17:09,436 [John] The three, and this is the trouble, the, the three or four or something is if you are in a very niche business of 00:17:11,076 --> 00:17:16,596 [John] some sort of content business where you're, where you're doing, um, 00:17:17,836 --> 00:17:20,356 [John] uh, y- you know, like influencer type stuff. 00:17:20,356 --> 00:17:21,136 [Eric] Hmm. 00:17:21,136 --> 00:17:29,716 [John] So there's this weird cycle of, like, the people that are into AI and, like, do podcasts or do YouTube or do whatever- 00:17:29,716 --> 00:17:29,806 [Eric] Mm-hmm 00:17:29,806 --> 00:17:34,396 [John] ... as their job, and they're seeing that, like, oh, I could... Like, we just talked about research. 00:17:34,396 --> 00:17:34,716 [Eric] Hmm. 00:17:34,716 --> 00:17:35,416 [John] That, that, that being- 00:17:35,416 --> 00:17:35,716 [Eric] Yeah 00:17:35,716 --> 00:17:52,386 [John] ... an 8 out of 10 greatly benefits them 'cause all they're doing, all they've ever done is follow headlines, look at press releases, look at maybe, maybe some podcasts, look at, like, research papers, but mostly, like, press releases and scrape the internet, pull all the headlines in, create content, make video. 00:17:52,386 --> 00:17:52,396 [Eric] Yep. 00:17:52,396 --> 00:17:58,016 [John] Whatever they're doing, that motion is high, like, is way over here as far as- 00:17:58,016 --> 00:17:58,076 [Eric] Yep 00:17:58,076 --> 00:18:00,356 [John] ... like, autonomous company for a billion dollars probably high. 00:18:00,356 --> 00:18:01,766 [Eric] Agreed. Yeah, yeah. 00:18:01,766 --> 00:18:08,376 [John] But it is such an edge case. But all of them are personally experience how awesome it is, so it gets, I think, exaggerated for everyone. 00:18:08,376 --> 00:18:09,336 [Eric] I, I agree with that. Yep. 00:18:09,336 --> 00:18:09,716 [John] Yeah. 00:18:09,716 --> 00:18:20,564 [Eric] But I would... It's, it plays towards exactly what we were saying, which is that in a single dimension-The context is highly related. 00:18:20,564 --> 00:18:20,594 [John] Yeah. 00:18:20,594 --> 00:18:24,164 [Eric] And so, and so there's... You get asymmetric value there. 00:18:24,164 --> 00:18:24,344 [John] Exactly. 00:18:24,344 --> 00:18:25,513 [Eric] But those types of businesses- 00:18:25,513 --> 00:18:25,513 [John] Right 00:18:25,513 --> 00:18:26,764 [Eric] ... are very one-dimensional. 00:18:26,764 --> 00:18:27,384 [John] Right. Right. 00:18:27,384 --> 00:18:31,664 [Eric] As soon as you move to an additional dimension, which most businesses are extremely complex- 00:18:31,664 --> 00:18:31,674 [John] Right. Right 00:18:31,674 --> 00:18:33,504 [Eric] ... like, that's where things get hard. 00:18:33,504 --> 00:18:37,744 [John] Right. Right. And the only other thing is some... Well, actually, it goes into our next one, so I won't- 00:18:37,744 --> 00:18:37,914 [Eric] Okay. Yes 00:18:37,914 --> 00:18:43,384 [John] ... I won't start this. Okay. Number three. Um, so we're, we're pretty l- fully autonomous company. 00:18:43,384 --> 00:18:43,544 [Eric] Mm-hmm. 00:18:43,544 --> 00:18:49,604 [John] We're pretty low there. I agree with that. Um, number three, I think this is the most controversial one. We've got a great Reddit post- 00:18:49,604 --> 00:18:49,944 [Eric] Ooh. Okay, great 00:18:49,944 --> 00:18:56,244 [John] ... for this one. Creating professional grade creative work, images, graphics. 00:18:56,244 --> 00:18:57,984 [Eric] Wow. Okay. 00:19:00,164 --> 00:19:02,284 [Eric] Wow. This is a very- 00:19:02,284 --> 00:19:03,604 [John] I'm gonna, I'm gonna pull up this post 00:19:03,604 --> 00:19:04,194 [Eric] ... con- this is very controversial 00:19:04,194 --> 00:19:08,804 [John] ... because this is something that came up, um, this week or last week. 00:19:08,804 --> 00:19:12,604 [Eric] Do you want me to re- do you want me to respond before you read the Reddit post? Or- 00:19:12,604 --> 00:19:12,924 [John] Y- yeah. 00:19:12,924 --> 00:19:13,404 [Eric] Okay. 00:19:13,404 --> 00:19:13,764 [John] Yeah, yeah. 00:19:13,764 --> 00:19:14,244 [Eric] Um, 00:19:19,004 --> 00:19:19,704 [Eric] let's see. 00:19:22,444 --> 00:19:24,464 [Eric] I... Can I have a split answer? 00:19:24,464 --> 00:19:25,244 [John] Yeah. 00:19:25,244 --> 00:19:25,464 [Eric] So 00:19:27,184 --> 00:19:31,244 [Eric] I think that for... I'm gonna have a couple of hot takes here. 00:19:32,504 --> 00:19:34,224 [Eric] Okay? Um, 00:19:35,604 --> 00:19:39,284 [Eric] there's a lot of really bad creative work in the world. 00:19:40,324 --> 00:19:41,464 [John] Pre-AI? 00:19:41,464 --> 00:19:42,424 [Eric] Pre-AI. 00:19:42,424 --> 00:19:42,904 [John] Okay. Yeah. 00:19:42,904 --> 00:19:43,764 [Eric] Generally. 00:19:43,764 --> 00:19:43,904 [John] Okay. 00:19:43,904 --> 00:19:44,924 [Eric] Right? Um, 00:19:46,764 --> 00:19:54,384 [Eric] I mean, if you talk to... Uh, it, it's sort of the classic cliché of if you talk to a designer or even if you talk to someone who builds software- 00:19:54,384 --> 00:19:54,574 [John] Mm-hmm 00:19:54,574 --> 00:20:01,384 [Eric] ... they kinda say, like, "Yeah, I mean, most design is not great," right? Or like, "Most software is pretty bad." 00:20:01,384 --> 00:20:01,524 [John] Yeah. Yeah 00:20:01,524 --> 00:20:01,884 [Eric] You know? Um- 00:20:01,884 --> 00:20:03,684 [John] I would definitely say that, for sure. 00:20:03,684 --> 00:20:04,804 [Eric] Yeah. You've said that before. 00:20:04,804 --> 00:20:04,944 [John] Yeah. 00:20:04,944 --> 00:20:05,404 [Eric] Right? It's like- 00:20:05,404 --> 00:20:05,564 [John] Yeah 00:20:05,564 --> 00:20:07,834 [Eric] ... yeah, most software is, like, pretty bad. 00:20:07,834 --> 00:20:07,924 [John] Right. 00:20:07,924 --> 00:20:11,723 [Eric] And so when you have, like, an amazing experience or you see, like, truly amazing design, you're like- 00:20:11,724 --> 00:20:11,824 [John] Right 00:20:11,824 --> 00:20:12,994 [Eric] ... "Wow, that's pretty awesome." 00:20:12,994 --> 00:20:13,003 [John] Right. 00:20:13,004 --> 00:20:13,264 [Eric] You know? 00:20:14,284 --> 00:20:14,664 [Eric] Um, 00:20:16,884 --> 00:20:17,224 [Eric] so 00:20:18,404 --> 00:20:23,384 [Eric] I think that AI will help improve 00:20:24,444 --> 00:20:28,924 [Eric] the low water mark for design in the world- 00:20:28,924 --> 00:20:28,933 [John] Yeah 00:20:28,933 --> 00:20:29,604 [Eric] ... generally. 00:20:29,604 --> 00:20:30,064 [John] Yeah. 00:20:30,064 --> 00:20:30,544 [Eric] Um, 00:20:31,864 --> 00:20:35,784 [Eric] and because it's extremely time-consuming, it's- 00:20:35,784 --> 00:20:36,484 [John] Yeah 00:20:36,484 --> 00:20:48,104 [Eric] ... it requires a very specific skill set, and AI creates this form factor where... Because I think you and I are probably both this way, where I'm not a designer- 00:20:48,104 --> 00:20:48,534 [John] No. Yeah 00:20:48,534 --> 00:20:53,664 [Eric] ... but I'm decent at saying, "I think that this type of thing would look good." 00:20:53,664 --> 00:20:53,754 [John] Right. 00:20:53,754 --> 00:20:54,644 [Eric] Right? Our, our logo- 00:20:54,644 --> 00:20:54,653 [John] Yeah 00:20:54,653 --> 00:20:55,624 [Eric] ... is a good example of that. 00:20:55,624 --> 00:20:55,644 [John] Yeah. 00:20:55,644 --> 00:20:59,364 [Eric] Right? We wanted something that looked like it was printed on, like, a, you know- 00:20:59,364 --> 00:20:59,524 [John] Yeah 00:20:59,524 --> 00:21:00,634 [Eric] ... a brown bag. 00:21:00,634 --> 00:21:00,984 [John] Yeah. Uh-huh. 00:21:00,984 --> 00:21:01,424 [Eric] You know? And- 00:21:01,424 --> 00:21:01,614 [John] Yeah 00:21:01,614 --> 00:21:02,844 [Eric] ... it's like, okay, this is cool, right? 00:21:02,844 --> 00:21:03,464 [John] Right. 00:21:03,464 --> 00:21:06,704 [Eric] But it would be very difficult for me to actually create that. 00:21:06,704 --> 00:21:06,824 [John] Right. 00:21:06,824 --> 00:21:10,644 [Eric] But with AI, I can describe it, and it can produce- 00:21:10,644 --> 00:21:10,884 [John] Right 00:21:10,884 --> 00:21:12,554 [Eric] ... way, way, way, way better results than- 00:21:12,554 --> 00:21:12,554 [John] For sure 00:21:12,554 --> 00:21:13,324 [Eric] ... I would've been- 00:21:13,324 --> 00:21:13,534 [John] For sure 00:21:13,534 --> 00:21:14,584 [Eric] ... able to do on my own, right? 00:21:14,584 --> 00:21:15,304 [John] Yeah. 00:21:15,304 --> 00:21:19,364 [Eric] Okay. If we go to the other end of the spectrum and we think about brands that have 00:21:20,544 --> 00:21:31,464 [Eric] an extremely opinionated brand, uh, very, you know, designers who curate very carefully, um, 00:21:33,164 --> 00:21:38,664 [Eric] it, you know, is AI leverage there? I think it can create some leverage- 00:21:38,664 --> 00:21:38,744 [John] Right 00:21:38,744 --> 00:21:40,704 [Eric] ... but it's way, way less. 00:21:40,704 --> 00:21:41,104 [John] Right. 00:21:41,104 --> 00:21:42,284 [Eric] Way, way less. 00:21:42,284 --> 00:21:42,864 [John] Yeah. 00:21:42,864 --> 00:21:51,304 [Eric] I think it creates leverage in taking the design work from the designer and translating it into different formats, right? 00:21:51,304 --> 00:21:51,433 [John] Oh, definitely. 00:21:51,433 --> 00:21:53,984 [Eric] I wanna create an, a set of ads. I wanna- 00:21:53,984 --> 00:21:54,814 [John] Right 00:21:54,814 --> 00:22:02,484 [Eric] ... you know, create an image for these, you know, for a blog post or, like, iterate on billboard ideas but based on existing design. 00:22:02,484 --> 00:22:02,844 [John] Right. 00:22:02,844 --> 00:22:07,404 [Eric] And so there's process things that sort of makes those faster. But, you know- 00:22:07,404 --> 00:22:07,414 [John] Right 00:22:07,414 --> 00:22:13,394 [Eric] ... the, you know, at least, I mean, I work with some on, like, I mean, industry renowned designers- 00:22:13,394 --> 00:22:13,404 [John] Right 00:22:13,404 --> 00:22:14,494 [Eric] ... on a daily basis. 00:22:14,494 --> 00:22:14,504 [John] Right. 00:22:14,504 --> 00:22:19,264 [Eric] And, uh, it's still really hard to be a truly phenomenal designer. 00:22:19,264 --> 00:22:19,484 [John] Yeah. 00:22:20,524 --> 00:22:20,804 [John] Okay. 00:22:22,464 --> 00:22:30,944 [John] This one, this one I think I have an interesting take on. They're... One of the superpowers of being able to use AI is to be able to express things in words. 00:22:30,944 --> 00:22:31,724 [Eric] Yep. 00:22:31,724 --> 00:22:32,784 [John] Written or- 00:22:32,784 --> 00:22:32,884 [Eric] Mm-hmm 00:22:32,884 --> 00:22:41,964 [John] ... you know, via voice, which is, you know, coming along. But one of the things that a really good designer can do is express that. 00:22:41,964 --> 00:22:42,384 [Eric] Yep. 00:22:42,384 --> 00:22:49,954 [John] Use a logo, for example, in words. Um, and obviously the creative capacity of, like, what looks good- 00:22:49,954 --> 00:22:49,954 [Eric] Yep 00:22:49,954 --> 00:22:55,984 [John] ... together and all that stuff, too. Um, which that, I mean, is getting easier to steer- 00:22:55,984 --> 00:22:56,084 [Eric] Mm-hmm 00:22:56,084 --> 00:22:56,734 [John] ... in AI, but it's- 00:22:56,734 --> 00:22:56,734 [Eric] Mm-hmm 00:22:56,734 --> 00:22:57,964 [John] ... still, like, kind of hard to steer. 00:22:57,964 --> 00:22:58,324 [Eric] Yep. 00:22:58,324 --> 00:23:03,654 [John] Um, and, and it feels more intangible, like what looks good together, how, like proportions. 00:23:03,654 --> 00:23:03,674 [Eric] Totally. Totally. 00:23:03,674 --> 00:23:05,434 [John] Like, that feels way more intangible. 00:23:05,434 --> 00:23:05,464 [Eric] Yep. 00:23:05,464 --> 00:23:15,204 [John] But I think one of the interesting things that you can see anywhere in the content or design world is, um... Or nature. Like, des- like, really good designers copy nature. 00:23:15,204 --> 00:23:15,324 [Eric] Yep. 00:23:15,324 --> 00:23:16,064 [John] Right? Like- 00:23:16,064 --> 00:23:16,204 [Eric] 100% 00:23:16,204 --> 00:23:18,184 [John] ... patterns and proportions and all the things. 00:23:18,184 --> 00:23:18,604 [Eric] Yep. 00:23:18,604 --> 00:23:27,904 [John] So one of the cool things about AI, since image, imaging is, is a language that it speaks, is you can take... So you've got a poster behind you. 00:23:27,904 --> 00:23:28,264 [Eric] Mm-hmm. 00:23:28,264 --> 00:23:30,074 [John] You take a picture of that poster and be like, 00:23:31,174 --> 00:23:34,774 [John] "Describe this layout and fo-" like all this- 00:23:34,774 --> 00:23:34,774 [Eric] Yes 00:23:34,774 --> 00:23:35,293 [John] ... to me in words. 00:23:35,293 --> 00:23:36,994 [Eric] It's a translation tool. Yeah, yeah, yeah. 00:23:36,994 --> 00:23:38,424 [John] Yeah. You can, you can use the translation tool. 00:23:38,424 --> 00:23:39,424 [Eric] Wow. Yeah. Yep. 00:23:39,424 --> 00:23:42,174 [John] Take a picture of, like, the cover of that book or something you like. 00:23:42,174 --> 00:23:42,924 [Eric] Mm-hmm. 00:23:42,924 --> 00:23:45,874 [John] Or, like, or say I was decorating. I don't know any of decorating language. 00:23:45,874 --> 00:23:45,884 [Eric] Mm-hmm. 00:23:45,884 --> 00:23:47,524 [John] But I'm like, "I like this room." And they're like, "That is, like, 00:23:48,544 --> 00:23:51,854 [John] arc..." Like, whatever. Like, French modern country. Like, I would never- 00:23:51,854 --> 00:23:51,854 [Eric] Yeah 00:23:51,854 --> 00:23:52,424 [John] ... know those words. 00:23:52,424 --> 00:23:52,784 [Eric] Right. Right. 00:23:52,784 --> 00:23:54,344 [John] And then you can feed that back into the model and get a result. 00:23:54,344 --> 00:23:55,584 [Eric] Yeah, yeah, yeah. Totally. 00:23:55,584 --> 00:23:56,554 [John] So I think that's one of the most interesting- 00:23:56,554 --> 00:23:57,624 [Eric] It creates a feedback loop. 00:23:57,624 --> 00:23:57,944 [John] Yeah. 00:23:57,944 --> 00:23:58,904 [Eric] It's awesome. 00:23:58,904 --> 00:23:59,114 [John] Yeah. 00:23:59,114 --> 00:23:59,144 [Eric] Yeah. 00:23:59,144 --> 00:24:00,804 [John] It's a little non-intuitive. 00:24:00,804 --> 00:24:01,164 [Eric] Yep. 00:24:01,164 --> 00:24:08,324 [John] Um, but I agree that the, the... I think we brought the average up, but I still think there's a, there's a pretty sizable gap between- 00:24:08,324 --> 00:24:08,824 [Eric] Yeah, I agree 00:24:08,824 --> 00:24:09,164 [John] ... the best. 00:24:09,164 --> 00:24:10,084 [Eric] I agree. 00:24:10,084 --> 00:24:11,604 [John] All right. Reddit. Ready? 00:24:11,604 --> 00:24:12,244 [Eric] Yes. 00:24:12,244 --> 00:24:43,399 [John] Okay. So this is, uh, this is a Reddit post about X. So [laughs] process that for a minute. Um, okay. I just generated an image in the style of a Monet painting using AI.Please describe in as much detail as possible, describing an image, what makes this inferior to a real Monet, um, painting. And then just, like, wall of comments on X. I'll read a couple. "There's a certain harshness, no soft blending of colors, no depth, no symbiosis- 00:24:43,400 --> 00:24:43,460 [Eric] Hmm 00:24:43,460 --> 00:24:44,200 [John] ... of the elements." 00:24:45,300 --> 00:24:45,980 [John] Um, it's- 00:24:45,980 --> 00:24:47,140 [Eric] Was it an actual Monet- 00:24:47,140 --> 00:24:47,290 [John] It's- 00:24:47,290 --> 00:24:48,239 [Eric] ... and they were just trolling? 00:24:48,240 --> 00:24:55,440 [John] It's all Bork no- nonsense with no sense of space. Yeah, so the, like, millions of views, tons of comments, and it was just an actual Monet. Yeah. 00:24:56,680 --> 00:24:57,880 [Eric] [laughs] 00:24:57,880 --> 00:24:58,860 [John] [laughs] 00:24:58,860 --> 00:24:59,940 [Eric] God, I love that. 00:24:59,940 --> 00:25:03,220 [John] Which... But, but we are kind of there. 00:25:03,220 --> 00:25:03,440 [Eric] Yep. 00:25:03,440 --> 00:25:05,569 [John] Not fully symmetrically, but- 00:25:05,569 --> 00:25:05,780 [Eric] Mm-hmm 00:25:05,780 --> 00:25:08,900 [John] ... but especially on, like, something as, like, famous as a Monet- 00:25:08,900 --> 00:25:09,140 [Eric] Mm-hmm 00:25:09,140 --> 00:25:12,540 [John] ... that it really can, pixel for pixel, be- 00:25:12,540 --> 00:25:12,740 [Eric] Right 00:25:12,740 --> 00:25:15,960 [John] ... nearly indistinguishable, if not indistinguishable from the original. 00:25:15,960 --> 00:25:16,540 [Eric] Right. 00:25:16,540 --> 00:25:17,740 [John] Which is the point of the post. 00:25:17,740 --> 00:25:18,530 [Eric] Which is the point of the post. 00:25:18,530 --> 00:25:19,160 [John] Even though it's kind of trolling. 00:25:19,160 --> 00:25:24,260 [Eric] I actually w- made a joke to one of my friends the other day who's an, who's an artist, um, 00:25:25,500 --> 00:25:39,350 [Eric] digital media photography, et cetera. But, um, you know, I said, uh... We were joking about it. We were talking about paintings, and, uh, I said, "You know, Jackson Pollock is, uh, you know, the... It's kind of funny because 00:25:40,680 --> 00:25:47,600 [Eric] I, now that I have kids, I view Jackson Pollock differently because I'm like, kids just sort of create Jackson Pollock-esque- 00:25:47,600 --> 00:25:47,890 [John] Uh-huh. Yeah 00:25:47,890 --> 00:25:49,760 [Eric] ... you know, drip paintings." [laughs] 00:25:49,760 --> 00:25:50,120 [John] Right. Right. 00:25:51,700 --> 00:25:56,310 [Eric] And he made it... His response was really interesting. He doesn't have kids, but he- 00:25:56,310 --> 00:25:56,310 [John] Right 00:25:56,310 --> 00:25:57,970 [Eric] ... he kind of chuckled, and he said, "Yeah, I mean, 00:26:00,340 --> 00:26:05,990 [Eric] if you think about things like that, it's, it's the time in history 00:26:07,200 --> 00:26:09,580 [Eric] and the current, like, culture of art- 00:26:09,580 --> 00:26:09,590 [John] Right 00:26:09,590 --> 00:26:12,160 [Eric] ... that, that really imbues it- 00:26:12,160 --> 00:26:12,170 [John] Right 00:26:12,170 --> 00:26:13,120 [Eric] ... with importance." 00:26:13,120 --> 00:26:13,760 [John] Yeah. Yeah. 00:26:13,760 --> 00:26:14,880 [Eric] Um, you know, which is interesting. 00:26:14,880 --> 00:26:14,900 [John] Yeah. 00:26:14,900 --> 00:26:16,560 [Eric] So it kind of goes back to context anyways. 00:26:16,560 --> 00:26:23,190 [John] Yeah. And I even feel like, I don't know a lot about art, but we even refer to art in the context of history, like, like a period, and the art- 00:26:23,190 --> 00:26:23,190 [Eric] Yeah 00:26:23,190 --> 00:26:23,470 [John] ... is associated with the period. 00:26:23,470 --> 00:26:25,040 [Eric] Absolutely. Yep. Yep. 00:26:25,040 --> 00:26:26,940 [John] Okay. This one, this is a good one. Um, 00:26:27,980 --> 00:26:30,380 [John] uh, pretty obvious, I think. Coding. 00:26:30,380 --> 00:26:32,280 [Eric] Hmm. Yes. 00:26:32,280 --> 00:26:34,179 [John] Oh, do we give numbers on the, on the last one, though? 00:26:35,200 --> 00:26:36,260 [Eric] Well, we split it. 00:26:36,260 --> 00:26:36,860 [John] We s- yeah. 00:26:36,860 --> 00:26:37,460 [Eric] It's difficult. 00:26:37,460 --> 00:26:37,900 [John] Yeah, I agree. 00:26:37,900 --> 00:26:39,620 [Eric] It's probably 50/50. I don't know. 00:26:39,620 --> 00:26:40,040 [John] Yeah. I agree. 00:26:40,040 --> 00:26:41,420 [Eric] Um. 00:26:41,420 --> 00:26:42,450 [John] All right, coding. 00:26:42,450 --> 00:26:44,780 [Eric] Coding, I'm going to go... 00:26:44,780 --> 00:26:45,839 [John] Like software coding. 00:26:45,840 --> 00:26:46,000 [Eric] Yep. 00:26:46,000 --> 00:26:46,809 [John] Of course. Yeah. 00:26:46,809 --> 00:26:47,500 [Eric] Yep. Uh, 00:26:49,680 --> 00:26:50,580 [Eric] I'm gonna go... 00:26:52,280 --> 00:26:56,320 [Eric] Ooh, this is a hard one for me. Um, I'm gonna go 00:26:58,360 --> 00:26:59,230 [Eric] 7 out of 10, 00:27:00,920 --> 00:27:01,469 [Eric] and 00:27:02,740 --> 00:27:09,870 [Eric] the reason I'm gonna go 7 out of 10 is because... Well, actually, this is the reason I'm going 7 out of 10. It... 00:27:11,280 --> 00:27:15,520 [Eric] I get the chance to work with some unbelievable software engineers- 00:27:15,520 --> 00:27:15,700 [John] Right 00:27:15,700 --> 00:27:16,460 [Eric] ... in my job. 00:27:16,460 --> 00:27:16,499 [John] Right. 00:27:16,500 --> 00:27:20,540 [Eric] I mean, really phenomenal. Pre-AI, you know- 00:27:20,540 --> 00:27:21,220 [John] Right 00:27:21,220 --> 00:27:24,330 [Eric] ... renowned, uh, you know, in Silicon Valley. 00:27:24,330 --> 00:27:24,520 [John] Yeah. Right. 00:27:25,600 --> 00:27:26,060 [Eric] And 00:27:28,080 --> 00:27:29,380 [Eric] not one of them 00:27:31,180 --> 00:27:32,960 [Eric] still write software the old way. 00:27:32,960 --> 00:27:33,320 [John] Sure. 00:27:33,320 --> 00:27:35,720 [Eric] They only write software with coding agents. 00:27:35,720 --> 00:27:36,640 [John] Right. 00:27:36,640 --> 00:27:45,440 [Eric] That's just how they do it. And so it has f- AI has fundamentally changed the way that those people do their jobs. 00:27:45,440 --> 00:27:46,600 [John] Yeah. 00:27:46,600 --> 00:27:46,940 [Eric] Uh- 00:27:46,940 --> 00:27:49,800 [John] S- so they can do their jobs a lot faster. Do they all go home early now? 00:27:49,800 --> 00:27:52,000 [Eric] [laughs] Do you not go home early? 00:27:52,000 --> 00:27:59,320 [John] [laughs] That... I mean, that's funny, but, like, outsider looking in, I think a lot of people are like, "You know, they probably don't have anything to do anymore." 00:27:59,320 --> 00:28:05,290 [Eric] Right. Y- yes, which I, uh, which is very easy to think that, but, um, 00:28:06,700 --> 00:28:14,220 [Eric] I think it's very easy for us to make a lot of assumptions about the way that things were pre-AI that are skewed- 00:28:14,220 --> 00:28:14,660 [John] Right 00:28:14,660 --> 00:28:20,170 [Eric] ... because of, because of our experience with AI. Or, or sorry, actually even better, or, like- 00:28:20,170 --> 00:28:20,170 [John] Right 00:28:20,170 --> 00:28:21,460 [Eric] ... our perceptions of AI. 00:28:21,460 --> 00:28:22,380 [John] Right. 00:28:22,380 --> 00:28:23,040 [Eric] And so... 00:28:24,740 --> 00:28:26,780 [Eric] A- and, uh, this goes back to something that we said before. 00:28:28,040 --> 00:28:33,400 [Eric] Most software in the world is still not great. 00:28:33,400 --> 00:28:34,140 [John] Right. 00:28:34,200 --> 00:28:41,499 [Eric] And so because building truly great software or really become... like creating great art- 00:28:41,499 --> 00:28:41,509 [John] Yep 00:28:41,509 --> 00:28:55,000 [Eric] ... or writing or whatever is not finding a silver bullet. It's, it's, like, an immense amount of hard work and iteration to get to the point where you craft something that's meaningful. 00:28:55,000 --> 00:28:56,100 [John] Right. 00:28:56,100 --> 00:29:18,970 [Eric] And, uh, and so when it comes to generating software, if you talk to someone who was previously a phenomenal engineer, uh, and, and now they use coding agents, what's happening is that they are cycling through all of the bad ways, bad ideas, you know, experiments way, way, way, way faster. 00:29:18,970 --> 00:29:19,260 [John] Right. 00:29:19,260 --> 00:29:30,139 [Eric] And so I think what's going to happen is that the quality of software will actually increase over time, kind of the same way with design. Um- 00:29:30,140 --> 00:29:30,320 [John] Right 00:29:30,320 --> 00:29:31,360 [Eric] ... you know, but for different reasons. 00:29:31,360 --> 00:29:32,740 [John] Like a elevation of the average. 00:29:32,740 --> 00:29:34,310 [Eric] An elevation of the average, right? 00:29:34,310 --> 00:29:34,320 [John] Yeah. 00:29:34,320 --> 00:29:38,660 [Eric] Because, uh, no, I mean, there's a lot of junk that's gonna be produced as well. 00:29:38,660 --> 00:29:40,090 [John] Mm-hmm. 00:29:40,090 --> 00:29:48,060 [Eric] Um, but the people who are wielding it within their craft as a soft... I think, uh, the, the conversation about people who are not software developers- 00:29:48,060 --> 00:29:49,140 [John] Yeah. I think that's a different conversation 00:29:49,140 --> 00:29:50,230 [Eric] ... generating code is different. 00:29:50,230 --> 00:29:50,260 [John] Yeah. 00:29:50,260 --> 00:29:53,080 [Eric] So I'm talking within the realm of, like, these are professionals- 00:29:53,080 --> 00:29:53,510 [John] Mm. Yeah 00:29:53,510 --> 00:29:54,880 [Eric] ... right, who can do this without AI. 00:29:54,880 --> 00:29:55,710 [John] Right. 00:29:55,710 --> 00:30:02,430 [Eric] And I think the, the general quality will go up because you can just run through the cycles so, so, so much faster, right? 00:30:02,430 --> 00:30:02,440 [John] Yeah. 00:30:02,440 --> 00:30:11,300 [Eric] And so there may be 10 ways to do something, and in the old world, it would've taken you months, right? And so- 00:30:11,300 --> 00:30:11,800 [John] Yeah 00:30:11,800 --> 00:30:14,190 [Eric] ... you make an educated guess on the best way to do it. 00:30:14,190 --> 00:30:14,920 [John] Yep. Uh, oops. 00:30:16,360 --> 00:30:18,370 [Eric] You make an educated be- guess on the best way to do it, 00:30:19,900 --> 00:30:24,350 [Eric] and you choose to spend your time as wisely as possible. And so you execute- 00:30:24,350 --> 00:30:24,350 [John] Yep 00:30:24,350 --> 00:30:26,440 [Eric] ... the one or two best ways and maybe test them, right? 00:30:26,440 --> 00:30:28,480 [John] Yeah. You would never test the other seven or eight or however many. 00:30:28,480 --> 00:30:33,620 [Eric] Right. A- and so now imagine you can do 50, and you can do it in one day. 00:30:33,620 --> 00:30:33,740 [John] Right. 00:30:33,740 --> 00:30:34,900 [Eric] And that is possible now. 00:30:34,900 --> 00:30:35,250 [John] Yeah. 00:30:35,250 --> 00:30:38,840 [Eric] And so these software engineers are, are, you know- 00:30:38,840 --> 00:30:38,860 [John] Yeah. 00:30:38,860 --> 00:30:40,940 [Eric] There are consequences to that. 00:30:40,940 --> 00:30:40,990 [John] Right. 00:30:42,936 --> 00:30:48,216 [Eric] I say seven out of 10 because, um, AI still makes a lot of mistakes, and you still actually- 00:30:48,216 --> 00:30:48,226 [John] Right 00:30:48,226 --> 00:30:53,466 [Eric] ... have to review code that's going to go in production, especially if it impacts real users and has real- 00:30:53,466 --> 00:30:53,466 [John] Yeah 00:30:53,466 --> 00:30:54,576 [Eric] ... security- 00:30:54,576 --> 00:30:54,676 [John] Right 00:30:54,676 --> 00:30:55,486 [Eric] ... implications. 00:30:55,486 --> 00:30:55,496 [John] Right. 00:30:55,496 --> 00:30:55,916 [Eric] And so, 00:30:57,036 --> 00:31:01,376 [Eric] um, and we're-- I-- the industry is still figuring out what that looks like. 00:31:01,376 --> 00:31:01,536 [John] Right. 00:31:02,626 --> 00:31:04,866 [John] Right. Cool. Yeah, I mean, coding, 00:31:05,996 --> 00:31:08,756 [John] coding I think is up near research for me. 00:31:08,756 --> 00:31:09,586 [Eric] Mm-hmm. 00:31:09,586 --> 00:31:22,776 [John] Um, but I also think that... And this probably, I'm just not a designer, so I can't say this authoritatively, but I, I bet designers would feel the same way, where it's essentially like, okay, who's u-using it? Is it somebody that has n-no idea about design- 00:31:22,776 --> 00:31:22,826 [Eric] Yeah. Yep 00:31:22,826 --> 00:31:24,336 [John] ... and, like, is just trying to, like, generate images? 00:31:24,336 --> 00:31:24,696 [Eric] Totally. 00:31:24,696 --> 00:31:24,776 [John] Like, 00:31:25,816 --> 00:31:28,836 [John] sure. They, they, they're better than average than maybe they would've been. 00:31:28,836 --> 00:31:29,296 [Eric] Yep. 00:31:29,296 --> 00:31:33,196 [John] Um, but for coding, yeah, I, I think similar to what you're saying, 00:31:34,276 --> 00:31:40,376 [John] it, um, drastically increases the average, and it also increases the, like, standardization. Like, uh- 00:31:40,376 --> 00:31:40,406 [Eric] Mm-hmm 00:31:40,406 --> 00:31:48,835 [John] ... like code, code, you know, even, even now, like you're trying to implement, um, user authentication or- 00:31:48,836 --> 00:31:48,896 [Eric] Yep 00:31:48,896 --> 00:31:56,136 [John] ... whatever, like it's gonna start looking pretty uniform if we're all using basically the same one or two tools to do the thing. 00:31:56,136 --> 00:31:56,146 [Eric] Mm-hmm. 00:31:56,146 --> 00:31:57,296 [John] Um, which is interesting, right? 00:31:57,296 --> 00:31:57,916 [Eric] Yeah. 00:31:57,916 --> 00:32:03,056 [John] And, and I think over time that elevates security and standards and stuff that helps everyone. 00:32:03,056 --> 00:32:03,976 [Eric] Yep. 00:32:03,976 --> 00:32:04,136 [John] Um, 00:32:05,176 --> 00:32:10,596 [John] I think the gaps there is a gap that's always been there, um, 'cause I've always been more on the operations side than the development side. 00:32:10,596 --> 00:32:10,956 [Eric] Mm-hmm. 00:32:10,956 --> 00:32:17,166 [John] Data and operations. Um, is it works for me locally on my computer is not the same as production grade software- 00:32:17,166 --> 00:32:18,006 [Eric] Yes, exactly. Yep 00:32:18,006 --> 00:32:31,366 [John] ... which is kinda like what you're saying. And now we have an even larger base of people in the like it works for me category that, like, makes the jump of like, "Yeah, like cool, let's just do it for- 00:32:31,366 --> 00:32:31,516 [Eric] Mm-hmm 00:32:31,516 --> 00:32:32,936 [John] ... the 10,000 people that work here now." 00:32:32,936 --> 00:32:33,346 [Eric] Mm-hmm. 00:32:33,346 --> 00:32:34,846 [John] And just feels hand-wavy about it. 00:32:34,846 --> 00:32:35,016 [Eric] Yeah. Right. Exactly. 00:32:35,016 --> 00:32:35,436 [John] You know? 00:32:35,436 --> 00:32:35,516 [Eric] Yeah. Yeah. 00:32:35,516 --> 00:32:36,776 [John] And I still think there's a gap there. 00:32:36,776 --> 00:32:37,666 [Eric] Yeah, there is. Yeah. 00:32:37,666 --> 00:32:37,686 [John] Um- 00:32:37,686 --> 00:32:39,156 [Eric] That's why I did seven out of 10, not- 00:32:39,156 --> 00:32:39,336 [John] Yeah 00:32:39,336 --> 00:32:42,056 [Eric] ... eight out of 10. But I think that gap will be closed over time. 00:32:42,056 --> 00:32:47,416 [John] Yeah. Yeah. Yep, for sure. Okay, last one. Um, this is one I've been into recently. 00:32:47,416 --> 00:32:47,636 [Eric] Okay. 00:32:47,636 --> 00:32:50,696 [John] We haven't talked too much about it, but AI voice. 00:32:52,376 --> 00:32:55,556 [Eric] Okay. Uh, in what regard? 00:32:55,556 --> 00:32:55,616 [John] Yep. 00:32:55,616 --> 00:32:56,616 [Eric] Just in general- 00:32:56,616 --> 00:32:56,825 [John] So I've got three- 00:32:56,825 --> 00:33:01,076 [Eric] ... or like as input or output or...? 00:33:01,136 --> 00:33:01,716 [John] Think about it 00:33:02,736 --> 00:33:06,496 [John] if chat is a product and voice is a product. Think about it that way. 00:33:06,496 --> 00:33:07,156 [Eric] Yep. 00:33:07,156 --> 00:33:12,156 [John] And then applications would be like sales, customer support, whatever. 00:33:12,156 --> 00:33:12,866 [Eric] Hmm. 00:33:12,866 --> 00:33:15,576 [John] I think there's that. Like, that's the practical of how- 00:33:15,576 --> 00:33:15,716 [Eric] Yep 00:33:15,716 --> 00:33:17,426 [John] ... people are using voice. And then, 00:33:18,616 --> 00:33:20,496 [John] and personal assistant. So personal assistant- 00:33:20,496 --> 00:33:20,516 [Eric] Mm-hmm. Mm-hmm 00:33:20,516 --> 00:33:24,545 [John] ... sales, customer service. Let's do that. Personal assistant and then like sales, customer service- 00:33:24,545 --> 00:33:24,556 [Eric] Yep 00:33:24,556 --> 00:33:25,476 [John] ... like classic- 00:33:25,476 --> 00:33:25,676 [Eric] Yep 00:33:25,676 --> 00:33:26,856 [John] ... phone tree use case. 00:33:28,876 --> 00:33:32,616 [Eric] Yep. Uh, okay, uh, I'm going nine out of 10 on this- 00:33:32,616 --> 00:33:32,836 [John] Whoa 00:33:32,836 --> 00:33:38,576 [Eric] ... in terms of utility, but I also think that it's the most underutilized... 00:33:39,896 --> 00:33:42,136 [Eric] I think it's very, very underutilized. 00:33:42,136 --> 00:33:44,265 [John] Mm-hmm. 00:33:44,265 --> 00:33:44,876 [Eric] Um, and 00:33:47,016 --> 00:33:51,656 [Eric] I-- the, the reason I say that is because I'm increasingly 00:33:52,896 --> 00:33:55,426 [Eric] using voice as input. 00:33:55,426 --> 00:33:55,426 [John] Yeah. 00:33:55,426 --> 00:34:04,876 [Eric] And so I used to... I mean, I've spent decades typing, and I don't even, I don't even wanna know how many words I type a day- 00:34:04,876 --> 00:34:05,056 [John] Right 00:34:05,056 --> 00:34:06,325 [Eric] ... you know, into- 00:34:06,325 --> 00:34:06,325 [John] Right 00:34:06,325 --> 00:34:07,356 [Eric] ... AI chatbot inputs. 00:34:07,356 --> 00:34:07,696 [John] Yeah. 00:34:08,816 --> 00:34:10,256 [Eric] But the 00:34:11,876 --> 00:34:16,666 [Eric] power of voice as input is significant. Um, 00:34:17,896 --> 00:34:24,756 [Eric] and the technology is good enough now to where you can essentially stream of consciousness- 00:34:24,756 --> 00:34:24,846 [John] Right 00:34:24,846 --> 00:34:26,616 [Eric] ... and it will clean it, like- 00:34:26,616 --> 00:34:26,736 [John] Yeah. 00:34:26,736 --> 00:34:28,496 [Eric] It will live clean up- 00:34:28,496 --> 00:34:29,616 [John] Yeah 00:34:29,616 --> 00:34:30,985 [Eric] ... everything into- 00:34:30,985 --> 00:34:30,985 [John] Right 00:34:30,985 --> 00:34:33,636 [Eric] ... like a pretty, a pretty clean input- 00:34:33,636 --> 00:34:33,646 [John] Yep 00:34:33,646 --> 00:34:35,176 [Eric] ... for an AI model as a prompt. 00:34:35,176 --> 00:34:35,436 [John] Yep. 00:34:35,436 --> 00:34:40,536 [Eric] Um, and then voice output is, is really good. 00:34:40,536 --> 00:34:40,796 [John] Mm-hmm. 00:34:40,796 --> 00:34:45,786 [Eric] And I don't know what this is like on the lower plans because I generally use it with an enterprise plan- 00:34:45,786 --> 00:34:45,786 [John] Right 00:34:45,786 --> 00:34:46,496 [Eric] ... on, you know- 00:34:46,496 --> 00:34:46,526 [John] Right 00:34:46,526 --> 00:34:48,316 [Eric] ... Claude or GPT. 00:34:48,316 --> 00:34:48,356 [John] Right. 00:34:48,356 --> 00:34:52,456 [Eric] But it's very common for me to be in the car alone. You know, I drop- 00:34:52,456 --> 00:34:52,466 [John] Mm-hmm 00:34:52,466 --> 00:34:54,416 [Eric] ... the kids off at school, and I'm driving to the office. 00:34:54,416 --> 00:34:55,496 [John] Yep. 00:34:55,496 --> 00:35:03,316 [Eric] And I will just turn on conversation mode in GPT or Claude and, you know, talk through a project that I'm working on or an idea- 00:35:03,316 --> 00:35:03,326 [John] Yeah 00:35:03,326 --> 00:35:06,366 [Eric] ... or whatever. And it is, it's unbelievable. 00:35:06,366 --> 00:35:07,156 [John] Which do you like more? 00:35:10,736 --> 00:35:13,176 [Eric] Uh, GPT's form factor is better. 00:35:13,176 --> 00:35:13,436 [John] Okay. Great. Yeah. 00:35:13,436 --> 00:35:14,856 [Eric] It's way better on voice. 00:35:14,856 --> 00:35:15,556 [John] Yep. 00:35:15,556 --> 00:35:15,856 [Eric] Um, 00:35:17,276 --> 00:35:22,036 [Eric] I mean, the, the output is... I, I don't know. GPT is a better experience. 00:35:22,036 --> 00:35:31,676 [John] Here. Okay. So this is, this is, I think s- I'm glad we're, like ending on this one 'cause I think it's so interesting. General public experience with voice, like- 00:35:31,676 --> 00:35:32,596 [Eric] Probably horrible 00:35:32,596 --> 00:35:34,596 [John] ... you call your, call your bank or call- 00:35:34,596 --> 00:35:34,836 [Eric] Yes 00:35:34,836 --> 00:35:37,956 [John] ... you know, customer service at some, like retail store or something. 00:35:37,956 --> 00:35:37,966 [Eric] Mm-hmm. 00:35:37,966 --> 00:35:39,816 [John] And it's like awful, and you're like- 00:35:39,816 --> 00:35:39,906 [Eric] Yep 00:35:39,906 --> 00:35:40,416 [John] ... yelling at it- 00:35:40,416 --> 00:35:40,626 [Eric] And you're- 00:35:40,626 --> 00:35:43,086 [John] ... and you're like dialing zero a bunch of times, like, "Can I talk- 00:35:43,086 --> 00:35:43,096 [Eric] Yep 00:35:43,096 --> 00:35:43,536 [John] ... to a human?" 00:35:43,536 --> 00:35:44,056 [Eric] Yep. 00:35:44,056 --> 00:35:47,066 [John] So I think that's like a common perception of like voice. 00:35:47,066 --> 00:35:47,076 [Eric] Mm-hmm. 00:35:47,076 --> 00:35:49,736 [John] But in actuality... So we, couple things. One, 00:35:50,776 --> 00:35:52,916 [John] um, Grok came out with a new voice model this week. 00:35:52,916 --> 00:35:53,076 [Eric] Mm-hmm. 00:35:53,076 --> 00:35:53,896 [John] It's really good. 00:35:53,896 --> 00:35:54,076 [Eric] Mm-hmm. 00:35:54,076 --> 00:35:56,376 [John] Check it out. GPT is still my favorite. 00:35:56,376 --> 00:35:58,126 [Eric] Yep. It's awesome. 00:35:58,126 --> 00:36:11,876 [John] Um, it's awesome. I use it a lot. The um, the edges that are frustrating for me right now, and I don't know if it's security or just, like they just haven't done it, um, is it does not integrate it with anything. So when you're in voice mode- 00:36:11,876 --> 00:36:11,886 [Eric] Yep. Yep 00:36:11,886 --> 00:36:13,216 [John] ... you can't like tell it like- 00:36:13,216 --> 00:36:13,446 [Eric] Mm-hmm 00:36:13,446 --> 00:36:18,286 [John] ... "Hey, what's new in my inbox?" Or what's whatever. There, there's like, I think calendar integrations- 00:36:18,286 --> 00:36:18,296 [Eric] Yep 00:36:18,296 --> 00:36:19,916 [John] ... works, but like, basically nothing works. 00:36:19,916 --> 00:36:20,776 [Eric] Yep. 00:36:20,776 --> 00:36:23,695 [John] Um, but I think that's coming. I think people will love it. 00:36:23,696 --> 00:36:24,016 [Eric] Yep. 00:36:24,016 --> 00:36:28,376 [John] Um, and the other like axis here is, um, 00:36:29,496 --> 00:36:34,526 [John] the... And this is a little, this is a silly thing, but if you think about Siri, if that's like your- 00:36:34,526 --> 00:36:34,526 [Eric] Mm-hmm. Mm-hmm 00:36:34,526 --> 00:36:46,944 [John] ... your, um, point of comparison, voice-to-text, long form, like talk to it for 20 minutes, the like latest and best models areWay better- 00:36:46,944 --> 00:36:47,054 [Eric] They're- 00:36:47,054 --> 00:36:49,144 [John] ... than Siri or anything that you've ever used. 00:36:49,144 --> 00:36:51,384 [Eric] It's, it is, it is astounding. 00:36:51,384 --> 00:36:51,624 [John] Yeah. 00:36:51,624 --> 00:36:53,004 [Eric] I, I cannot overstate- 00:36:53,004 --> 00:36:53,154 [John] Yeah 00:36:53,154 --> 00:36:53,763 [Eric] ... how good it is. 00:36:53,764 --> 00:36:57,244 [John] They're remarkably good, even for the podcast. So- 00:36:57,244 --> 00:36:57,424 [Eric] Yes 00:36:57,424 --> 00:37:05,324 [John] ... we re- record this, and then we don't use... Like, we have software that we use to record the podcast. We'll use a separate external model to process- 00:37:05,324 --> 00:37:05,404 [Eric] Yep 00:37:05,404 --> 00:37:10,284 [John] ... this to get a transcript, and, um, it is awesome. 00:37:10,284 --> 00:37:10,484 [Eric] Yep. 00:37:10,484 --> 00:37:12,614 [John] Like, super, super clean. 00:37:12,614 --> 00:37:18,153 [Eric] Yeah. I think that's the reason I'm 9 out of 10, because based [laughs] on everything you just said, I should be lower, but- 00:37:18,153 --> 00:37:18,153 [John] Right 00:37:18,153 --> 00:37:20,153 [Eric] ... I'm so impressed, 00:37:21,424 --> 00:37:32,244 [Eric] e- even though you're right, like the utility is not quite there. Actually, I mean, there is... You can still... Again, I'm, I'm probably skewed on this, 'cause a lot of the things I use are on enterprise plans- 00:37:32,244 --> 00:37:33,304 [John] Sure, right 00:37:33,304 --> 00:37:36,164 [Eric] ... which, you know, have more functionality. 00:37:36,164 --> 00:37:36,884 [John] Right, right. 00:37:36,884 --> 00:37:42,304 [Eric] Uh, you can, you know, sort of, um... Well, even WhisperFlow can abstract some of this out. 00:37:42,304 --> 00:37:42,534 [John] True. 00:37:42,534 --> 00:37:43,174 [Eric] But, you know, sort of- 00:37:43,174 --> 00:37:43,174 [John] Yeah 00:37:43,174 --> 00:37:43,584 [Eric] ... like- 00:37:43,584 --> 00:37:43,804 [John] Right 00:37:43,804 --> 00:37:47,284 [Eric] ... cleaning up your stuff and then putting it into, directly into the tool as a prompt. 00:37:47,284 --> 00:37:48,004 [John] Right. 00:37:48,004 --> 00:37:48,664 [Eric] But, um, 00:37:49,964 --> 00:37:51,524 [Eric] it's so useful- 00:37:51,524 --> 00:37:51,644 [John] Right 00:37:51,644 --> 00:37:52,844 [Eric] ... in the current state- 00:37:52,844 --> 00:37:52,934 [John] It is 00:37:52,934 --> 00:37:59,103 [Eric] ... that it's hard to see a world in which it's not the dominant way- 00:37:59,104 --> 00:37:59,124 [John] Yeah 00:37:59,124 --> 00:38:01,424 [Eric] ... that you communicate with AI, you know, 00:38:02,944 --> 00:38:03,724 [Eric] in the future. 00:38:03,724 --> 00:38:05,444 [John] And, and we've always thought it would be. 00:38:05,444 --> 00:38:06,384 [Eric] Yes. Yeah. 00:38:06,384 --> 00:38:06,984 [John] Any, like- 00:38:06,984 --> 00:38:07,984 [Eric] Isn't that interesting? 00:38:07,984 --> 00:38:09,404 [John] ... s- like, Star Trek- 00:38:09,404 --> 00:38:09,544 [Eric] Yeah 00:38:09,544 --> 00:38:09,963 [John] ... or like- 00:38:09,964 --> 00:38:10,844 [Eric] Yep, science fiction 00:38:10,844 --> 00:38:11,984 [John] ... any science fiction. 00:38:11,984 --> 00:38:12,014 [Eric] Totally. 00:38:12,014 --> 00:38:13,144 [John] That's how we thought it would be. 00:38:13,144 --> 00:38:13,694 [Eric] Yeah, yeah. Exactly. 00:38:13,694 --> 00:38:15,054 [John] So that, that is the part of science fiction- 00:38:15,054 --> 00:38:15,624 [Eric] Isn't that interesting? Yeah 00:38:15,624 --> 00:38:17,444 [John] ... that I think is coming true right now. 00:38:17,444 --> 00:38:17,864 [Eric] Okay. 00:38:17,864 --> 00:38:18,524 [John] All right. 00:38:18,524 --> 00:38:22,704 [Eric] That's what you can do and what you can't do with AI. So let's just run through it again really quickly. 00:38:22,704 --> 00:38:23,204 [John] All right. 00:38:23,204 --> 00:38:27,204 [Eric] So, uh, web, web search and deep research. 00:38:27,204 --> 00:38:27,264 [John] Yep. 00:38:27,264 --> 00:38:29,164 [Eric] We landed on an 8 out of 10. 00:38:29,164 --> 00:38:34,084 [John] Yep. Okay. Running a fully autonomous one-person company, average our scores were about a two. [laughs] 00:38:34,084 --> 00:38:35,724 [Eric] Average of two. 00:38:35,724 --> 00:38:39,174 [John] Um, creating professional-grade creative work, I think we're s- 00:38:39,174 --> 00:38:39,964 [Eric] We were split 00:38:39,964 --> 00:38:41,074 [John] ... split on that one. 00:38:41,074 --> 00:38:41,984 [Eric] So it's 50/50. Yep. 00:38:41,984 --> 00:38:45,964 [John] Yep. Um, and then coding, seven. 00:38:45,964 --> 00:38:48,034 [Eric] Seven-ish out of 10. 00:38:48,034 --> 00:38:49,044 [John] Yeah. And then voice. 00:38:49,044 --> 00:38:49,584 [Eric] Yeah. 00:38:49,584 --> 00:38:51,204 [John] Uh, Eric's a nine. I didn't give a score on this one. 00:38:51,204 --> 00:38:51,944 [Eric] Yeah. 00:38:51,944 --> 00:38:55,574 [John] I'm, I'm like a, like a seven and a half, just because of the edge of like- 00:38:55,574 --> 00:38:56,084 [Eric] Yeah, yeah. Okay 00:38:56,084 --> 00:38:57,874 [John] ... can't you just... Can't I just ask it about my calendar? 00:38:57,874 --> 00:39:00,094 [Eric] That's probably a better... That's probably a more accurate description, yeah. 00:39:00,094 --> 00:39:02,003 [John] Like, just tell me what's on my calendar, please. 00:39:02,004 --> 00:39:02,604 [Eric] Yep. Yeah. It's coming. 00:39:02,604 --> 00:39:03,544 [John] So it's coming, I'm sure. 00:39:03,544 --> 00:39:08,604 [Eric] Yeah. All right. Well, thanks for joining Token Intelligence, and we will catch you on the next one. 00:39:12,144 --> 00:39:15,844 [Eric] [outro music]