Copy Transcript
Copied
WEBVTT 00:00:01.419 --> 00:00:06.780 I use AI. I find it useful. I think there are 00:00:06.780 --> 00:00:11.580 a lot of legitimate uses for it. Just like anything, 00:00:12.160 --> 00:00:16.059 it's a tool. And I think a lot of the knee -jerk 00:00:16.059 --> 00:00:19.320 reactions to anything involving artificial intelligence 00:00:19.320 --> 00:00:21.440 at this point is just freaking ridiculous and 00:00:21.440 --> 00:00:26.550 I'm tired of hearing about it. But there is a 00:00:26.550 --> 00:00:30.489 massive difference between you know using AI 00:00:30.489 --> 00:00:33.649 as a tool and giving an AI agent the keys to 00:00:33.649 --> 00:00:38.189 your entire digital life and Meta's new Muse 00:00:38.189 --> 00:00:42.030 AI tool is a really good example of where I am 00:00:42.030 --> 00:00:46.579 drawing the line and if you didn't catch that 00:00:46.579 --> 00:00:49.359 Muse AI made by Meta you should already know 00:00:49.359 --> 00:00:52.520 where I'm going with this because an AI assistant 00:00:52.520 --> 00:00:58.119 recently gave a stranger a man's home address 00:00:58.119 --> 00:01:03.149 through Facebook Marketplace. It's real. It happened. 00:01:03.390 --> 00:01:06.829 Not theoretically. A guy actually showed up at 00:01:06.829 --> 00:01:11.079 his house. Tech creator Matt Robb had listed 00:01:11.079 --> 00:01:13.840 a keyboard for sale on Facebook Marketplace and 00:01:13.840 --> 00:01:19.359 let Meta's Muse AI handle some of the Marketplace 00:01:19.359 --> 00:01:22.519 interaction for him. I mean, that sounds like 00:01:22.519 --> 00:01:25.840 something you might want an agent to do for you. 00:01:25.959 --> 00:01:30.280 But according to Robb, Muse accepted an offer, 00:01:30.280 --> 00:01:33.900 communicated with the buyer, gave the buyer Rob's 00:01:33.900 --> 00:01:37.659 home address for pickup and arranged a transaction. 00:01:39.150 --> 00:01:41.489 Rob didn't realize any of this had happened until 00:01:41.489 --> 00:01:45.980 the buyer showed up. Imagine that for a second. 00:01:47.299 --> 00:01:50.280 Just picture it. You're sitting at home, the 00:01:50.280 --> 00:01:52.760 doorbell rings, there's a stranger standing there 00:01:52.760 --> 00:01:55.519 expecting to buy something from you and apparently 00:01:55.519 --> 00:02:00.459 your AI agent, your AI assistant, arranged a 00:02:00.459 --> 00:02:09.250 whole damn thing. According to Dexerto's reporting, 00:02:10.030 --> 00:02:14.669 Rob later discovered that he had selected an 00:02:14.669 --> 00:02:18.270 allow always permission when Muse asked about 00:02:18.270 --> 00:02:23.460 handling marketplace messages. Rob thought Muse 00:02:23.460 --> 00:02:26.759 would still ask for approval before accepting 00:02:26.759 --> 00:02:30.259 any offers or doing something as sensitive as 00:02:30.300 --> 00:02:34.569 Maybe giving his home address out? No, of course 00:02:34.569 --> 00:02:37.930 it didn't work that way. Metta's David Singleton 00:02:38.120 --> 00:02:42.300 responded publicly and said previous cases the 00:02:42.300 --> 00:02:46.080 company investigated involved Muse following 00:02:46.080 --> 00:02:50.639 instructions and correctly requesting permission. 00:02:51.590 --> 00:02:54.629 Metta offered to investigate what happened to, 00:02:54.629 --> 00:02:58.189 you know, in Rob's case. So yeah, there's a little 00:02:58.189 --> 00:03:05.020 bit of nuance here. Now, That wasn't necessarily 00:03:05.020 --> 00:03:07.780 Muse breaking into a database and stealing his 00:03:07.780 --> 00:03:12.819 address. Rob gave Muse the information he had. 00:03:13.240 --> 00:03:16.479 Also, I guess, given it permission to perform 00:03:16.479 --> 00:03:20.050 certain actions. But that's exactly why this 00:03:20.050 --> 00:03:24.780 bothers me, though. Most people, they're not 00:03:24.780 --> 00:03:28.319 going to read every permission screen. I mean, 00:03:29.219 --> 00:03:32.939 right? You're not reading those those terms of 00:03:32.939 --> 00:03:37.280 services or privacy policies, I know you're not. 00:03:37.520 --> 00:03:40.620 Nobody's reading that stuff because it's written 00:03:40.620 --> 00:03:43.979 out in such legalese that none of us normie people 00:03:43.979 --> 00:03:46.759 could even understand it. And that is a problem 00:03:46.759 --> 00:03:52.300 and that should be addressed. Most people don't 00:03:52.300 --> 00:03:56.759 understand what the allow always means when they're 00:03:56.759 --> 00:04:01.020 using a completely new category of technology. 00:04:04.150 --> 00:04:07.889 primarily dominant on facebook right now right 00:04:07.889 --> 00:04:14.180 which generation it's the boomers i'm not saying 00:04:14.180 --> 00:04:16.399 that to be derogatory or anything like that but 00:04:16.399 --> 00:04:19.220 that's what it is it is the older folks now you 00:04:19.220 --> 00:04:21.819 know maybe elder millennials but boomers your 00:04:21.819 --> 00:04:25.439 your mom and dad your aunt uncle they're on facebook 00:04:25.439 --> 00:04:28.000 and they don't understand how these tools work 00:04:29.209 --> 00:04:32.689 so you've got your mom your grandmother somebody's 00:04:32.689 --> 00:04:38.220 on there giving their permission away to Meta's 00:04:38.220 --> 00:04:45.029 AI agent. And we know Meta's track record with 00:04:45.029 --> 00:04:47.930 permission issues, right? They're selling off 00:04:47.930 --> 00:04:50.350 all your personal data. They've got everything 00:04:50.350 --> 00:04:55.439 on you. To me, that's not a company I want to 00:04:55.439 --> 00:05:01.649 give any more information to. If it wasn't for 00:05:02.199 --> 00:05:06.980 certain contacts that I have in the broadcast 00:05:06.980 --> 00:05:09.589 engineering community, I wouldn't be on Facebook 00:05:09.589 --> 00:05:13.769 anymore. It's a dreadful place to be. But that's 00:05:13.769 --> 00:05:15.949 just me getting off on a tangent. So I'm going 00:05:15.949 --> 00:05:18.430 to reel it back in. Hang on. I'm getting there. 00:05:20.629 --> 00:05:25.300 You know, we've spent years, you know, we've 00:05:25.300 --> 00:05:28.399 spent years training people to click through 00:05:28.399 --> 00:05:31.199 all these permission prompts, allow notifications, 00:05:31.720 --> 00:05:34.079 allow location, allow contacts, allow Bluetooth, 00:05:34.240 --> 00:05:38.360 allow the microphone, allow photos. Now. we're 00:05:38.360 --> 00:05:41.779 moving into something significantly more powerful. 00:05:42.790 --> 00:05:45.670 We're not just giving software permission to 00:05:46.149 --> 00:05:50.490 access information, we're giving software permission 00:05:50.490 --> 00:05:55.050 to make decisions with that software. That's 00:05:55.050 --> 00:05:58.550 entirely different. An AI agent can read something, 00:05:58.790 --> 00:06:01.089 interpret it, combine it with something else 00:06:01.089 --> 00:06:04.529 it knows about you, and then take an action on 00:06:04.529 --> 00:06:08.920 that. And apparently that action can include 00:06:08.920 --> 00:06:13.180 telling a complete stranger on the internet where 00:06:13.180 --> 00:06:18.079 you live. I don't know about you, but that should 00:06:18.079 --> 00:06:22.259 make everybody really stop and think. When I 00:06:22.259 --> 00:06:26.060 heard this story on the latest episode of Syntax, 00:06:27.139 --> 00:06:29.620 I had to go look it up. I couldn't believe it. 00:06:29.959 --> 00:06:36.660 Again, meta being meta, doing meta things. Now, 00:06:36.759 --> 00:06:39.160 of course, meta describes Muse as a personal 00:06:39.160 --> 00:06:42.920 AI agent that can work across apps. Remember 00:06:42.920 --> 00:06:45.699 information about you. learn from conversations 00:06:45.699 --> 00:06:50.639 and perform tasks on your behalf. Meta says users 00:06:50.639 --> 00:06:54.420 control what Muse can access. Meta also says 00:06:54.420 --> 00:06:57.620 Muse doesn't share conversations or data inside 00:06:57.620 --> 00:07:02.420 its virtual machine with Meta's advertising systems 00:07:02.420 --> 00:07:06.459 and users can opt out of having interactions 00:07:06.459 --> 00:07:16.670 used to train Meta's AI models. Colorful language, 00:07:17.050 --> 00:07:20.990 I call bullshit. Alright, Meta doesn't have a 00:07:20.990 --> 00:07:24.470 good track record with this. I'm not going to 00:07:24.470 --> 00:07:29.750 claim, you know, take it with a grain of salt. 00:07:30.149 --> 00:07:33.430 I shouldn't claim Meta is secretly sending every 00:07:33.430 --> 00:07:35.949 conversation directly into its advertising database 00:07:35.949 --> 00:07:42.220 when Meta explicitly says otherwise, but I still 00:07:42.220 --> 00:07:44.439 wouldn't give this thing unrestricted access 00:07:44.439 --> 00:07:47.540 to your life. Because think about what these 00:07:47.540 --> 00:07:52.699 assistants Think about what makes these assistants 00:07:52.699 --> 00:07:57.129 useful. Your email, your calendar, your messages, 00:07:57.329 --> 00:08:00.689 your contacts, where you live, where you work, 00:08:01.209 --> 00:08:04.029 what you're buying, what you're selling, when 00:08:04.029 --> 00:08:06.910 you're available, who your family members are, 00:08:07.370 --> 00:08:11.029 what you're planning. Potentially your financial 00:08:11.029 --> 00:08:13.170 information, potentially your health information, 00:08:13.490 --> 00:08:16.170 potentially your travel plans, potentially access. 00:08:16.850 --> 00:08:19.930 to websites and services that you use every day, 00:08:20.290 --> 00:08:21.769 especially if you're using the Facebook login 00:08:21.769 --> 00:08:26.069 thing. The better the assistant understands you, 00:08:26.129 --> 00:08:29.410 the more useful it becomes. And that's also precisely 00:08:29.410 --> 00:08:33.230 what makes this thing incredibly dangerous. Especially 00:08:33.230 --> 00:08:38.509 for folks who don't understand how this technology 00:08:38.509 --> 00:08:45.279 works. The demographic that is on Facebook primarily 00:08:45.279 --> 00:08:48.399 at this point are the ones that are looking at 00:08:48.399 --> 00:08:51.940 the ai slop cat videos and saying this looks 00:08:51.940 --> 00:08:54.379 so real look at this and you try to tell them 00:08:54.379 --> 00:08:59.529 it's not real and they don't believe you so you're 00:08:59.529 --> 00:09:01.830 you know this the whole thing is creating one 00:09:01.830 --> 00:09:05.740 incredibly viable concentration of information 00:09:05.740 --> 00:09:10.330 on you meta already probably knows a quite quite 00:09:10.330 --> 00:09:13.269 a deal about you and they're trying to get more 00:09:13.750 --> 00:09:17.190 i don't want one system knowing everything anymore 00:09:22.659 --> 00:09:27.029 Wired recently reported that Muse can build detailed 00:09:27.029 --> 00:09:29.730 profiles of people in your life based on what 00:09:29.730 --> 00:09:33.500 it learns through your interactions. Friends, 00:09:33.960 --> 00:09:36.139 family members, coworkers, relationships between 00:09:36.139 --> 00:09:38.740 people, shared history, information that helps 00:09:38.870 --> 00:09:43.350 the AI understand who these people are and why 00:09:43.350 --> 00:09:48.799 they matter to you. Meta argues that this improves 00:09:48.799 --> 00:09:51.899 personalization and that users have controls 00:09:51.899 --> 00:09:58.700 over the information Muse remembers. I understand 00:09:58.700 --> 00:10:01.340 why that capability exists. From an engineering 00:10:01.340 --> 00:10:04.039 perspective, it makes perfect sense. If I tell 00:10:04.039 --> 00:10:08.980 some AI assistant, hey, remind me to ask Mike 00:10:08.980 --> 00:10:12.610 about that transmitter problem next week. The 00:10:12.610 --> 00:10:14.470 assistant needs some context about who the hell 00:10:14.470 --> 00:10:19.950 Mike is and what I'm talking about. But follow 00:10:19.950 --> 00:10:24.779 that concept far enough, eventually, You've created 00:10:24.779 --> 00:10:29.600 a database containing a great, surprisingly detailed 00:10:29.600 --> 00:10:35.039 log, model, whatever, of your life. Not just 00:10:35.039 --> 00:10:38.539 your life either. Other people's lives. People 00:10:38.539 --> 00:10:44.639 who never agreed to use this service. It starts 00:10:44.639 --> 00:10:46.080 getting a little uncomfortable when you think 00:10:46.080 --> 00:10:52.029 about it. This is how we keep getting here. Every 00:10:52.029 --> 00:10:54.350 new generation of technology offers convenience 00:10:54.350 --> 00:10:57.779 in exchange for another little piece of control. 00:10:58.399 --> 00:11:00.940 Put your photos here. put your contacts here 00:11:00.940 --> 00:11:03.259 put your documents over there connect your bank 00:11:03.259 --> 00:11:05.179 account connect your email connect your calendar 00:11:05.179 --> 00:11:08.179 connect your location let the assistant read 00:11:08.179 --> 00:11:10.779 your messages let it answer your messages let 00:11:10.779 --> 00:11:14.100 it negotiate facebook marketplace sales let it 00:11:14.100 --> 00:11:16.539 make the appointment let it buy the thing let 00:11:16.539 --> 00:11:22.100 it handle everything every individual permission 00:11:22.100 --> 00:11:28.500 sounds reasonable then one day you look at the 00:11:28.500 --> 00:11:30.539 complete picture and realize you've created a 00:11:30.539 --> 00:11:34.710 system that knows where you live, where you work, 00:11:35.289 --> 00:11:38.370 who you talk to, what you're buying, what you're 00:11:38.370 --> 00:11:42.620 selling, and when you're going to be home. And 00:11:42.620 --> 00:11:45.379 it has permission to communicate with strangers 00:11:45.379 --> 00:11:55.629 on your behalf. Yeah, no thanks. Now, let me 00:11:55.629 --> 00:12:01.389 be clear. I don't think AI is the problem. And 00:12:01.389 --> 00:12:05.549 this is an important distinction for me. This 00:12:05.549 --> 00:12:08.509 isn't an anti... AI episode. I'm not interested 00:12:08.509 --> 00:12:11.210 in becoming one of those people who shouts on 00:12:11.210 --> 00:12:13.830 threads that, you know, who, you know, these 00:12:13.830 --> 00:12:16.570 people who shout on, on threads about, you know, 00:12:16.870 --> 00:12:20.610 who they think uses chat GPT to rewrite an email 00:12:20.610 --> 00:12:25.179 means civilization is fully collapsing. AI can 00:12:25.179 --> 00:12:28.700 be incredibly useful. There also is ways you 00:12:28.700 --> 00:12:33.090 can use AI wrong, like a hammer. There's a correct 00:12:33.090 --> 00:12:35.610 way to use the hammer, but if you use a hammer 00:12:35.610 --> 00:12:40.809 incorrectly, You could put your eye out. Chainsaw. 00:12:40.990 --> 00:12:43.029 Use a chainsaw incorrectly, cut your damn leg 00:12:43.029 --> 00:12:46.980 off. I mean, I guess AI is going to kill us. 00:12:47.019 --> 00:12:48.899 A lot of people are saying it will. So maybe 00:12:48.899 --> 00:12:51.419 eventually it will do that too. I don't know. 00:12:53.519 --> 00:12:59.120 Again, but back to my point. I think AI can be 00:12:59.120 --> 00:13:02.159 useful. I use it for certain things. I build 00:13:02.159 --> 00:13:06.470 things with it. I'll continue using it. But I 00:13:06.470 --> 00:13:11.090 want AI to work for me inside boundaries. That 00:13:11.090 --> 00:13:14.750 I set and that I can understand there's a huge 00:13:14.750 --> 00:13:17.370 difference between help me analyze this information 00:13:17.370 --> 00:13:20.950 and hey Here's my email calendar contacts location 00:13:20.950 --> 00:13:24.029 personal history and permissions to act as me 00:13:24.259 --> 00:13:28.700 Go figure it out Those are not the same thing 00:13:28.700 --> 00:13:32.919 those you know, those are completely two different 00:13:32.919 --> 00:13:37.470 risk profiles one again is being that tool the 00:13:37.470 --> 00:13:41.820 other is well an autonomous representation of 00:13:41.820 --> 00:13:45.940 you do we want that do i need that i don't think 00:13:45.940 --> 00:13:53.940 so and we're and we're deploying that second 00:13:53.940 --> 00:13:57.799 way that second model that second one to millions 00:13:57.799 --> 00:14:00.740 of ordinary people again in that certain demographic 00:14:01.580 --> 00:14:04.539 who have spent 20 years clicking allow without 00:14:04.539 --> 00:14:11.210 ever thinking about it that's the problem don't 00:14:11.210 --> 00:14:17.460 treat ai agents like magic If you're going to 00:14:17.460 --> 00:14:23.889 use something like Muse or Gemini, ChatGPT, Cloud, 00:14:24.730 --> 00:14:29.940 or whatever agentic system comes next, understand 00:14:29.940 --> 00:14:35.899 exactly what you're giving it. Don't connect 00:14:35.899 --> 00:14:38.279 something just because the setup wizard says 00:14:38.279 --> 00:14:42.320 it'll make your experience better. If it wants 00:14:42.830 --> 00:14:46.210 access to your microphone, your photos, your 00:14:46.210 --> 00:14:52.970 location, it's okay to ask why. It's definitely 00:14:52.970 --> 00:14:59.500 okay to weigh the risk of giving it that information. 00:15:03.759 --> 00:15:05.740 Don't connect something just because the setup 00:15:05.740 --> 00:15:09.059 wizard says it's going to help or make things 00:15:09.059 --> 00:15:13.139 easier. Think about it. Use your brain. Ask yourself 00:15:13.139 --> 00:15:19.320 a few questions. Does the assistant actually 00:15:19.320 --> 00:15:24.730 need my email? Does it need permission to send 00:15:24.730 --> 00:15:28.899 email? Does it need my contacts? Does it need 00:15:28.899 --> 00:15:34.139 my exact home address? Does it need access to 00:15:34.139 --> 00:15:37.559 my calendar? Does it need permission to make 00:15:37.559 --> 00:15:41.860 purchases? Does it need to send messages without 00:15:41.860 --> 00:15:47.549 asking me? Does it need permanent access? If 00:15:47.549 --> 00:15:53.600 the answer is no, don't give it access. Use the 00:15:53.600 --> 00:15:58.350 minimum permissions that are necessary for what 00:15:58.350 --> 00:16:02.639 you're trying to do. And for sensitive actions, 00:16:03.179 --> 00:16:07.799 I want a human in the mix to make the confirmation 00:16:07.799 --> 00:16:12.080 every time. Sending a message, accepting an offer, 00:16:12.399 --> 00:16:15.340 making a purchase, sharing an address, changing 00:16:15.340 --> 00:16:19.320 an appointment, deleting something, especially 00:16:19.320 --> 00:16:21.879 deleting something. How many times have we heard 00:16:21.879 --> 00:16:23.980 the stories about AI agents going through and 00:16:23.980 --> 00:16:28.789 deleting Just a huge amount of files on someone's 00:16:28.789 --> 00:16:31.570 computer and erasing a project someone's been 00:16:31.570 --> 00:16:36.830 working on for years Think about it. Don't give 00:16:36.830 --> 00:16:45.480 it just free rate access those should not You 00:16:45.480 --> 00:16:47.419 know all that list I just gave you those should 00:16:47.419 --> 00:16:49.419 not suddenly happen because I clicked always 00:16:49.419 --> 00:16:55.809 allow three weeks ago and forget about it remember 00:16:56.129 --> 00:16:59.750 Your data is the most valuable thing that you're 00:16:59.750 --> 00:17:03.409 handing over people tend to think about privacy 00:17:04.299 --> 00:17:07.559 as if somebody discovering your favorite pizza 00:17:07.559 --> 00:17:10.079 topping is the worst thing that could happen. 00:17:11.309 --> 00:17:13.349 Now, that's not what concerns me. The dangerous 00:17:13.349 --> 00:17:19.640 part is aggregation. One piece of information 00:17:19.640 --> 00:17:24.130 isn't particularly interesting to anybody. Your 00:17:24.130 --> 00:17:27.049 address, your calendar, your contacts, your shopping 00:17:27.049 --> 00:17:31.500 history, your emails, your location, your browsing, 00:17:32.519 --> 00:17:36.859 your conversations separately. They're just pieces 00:17:36.859 --> 00:17:42.390 of information, but put enough of them together 00:17:42.390 --> 00:17:47.990 and you've built a model of a person. That's 00:17:47.990 --> 00:17:52.599 incredibly powerful. And when we're talking about 00:17:52.599 --> 00:17:59.650 AI agents, collecting that context isn't some 00:17:59.650 --> 00:18:05.190 weird side effect. Context is the product. The 00:18:05.190 --> 00:18:09.849 system becomes more useful because it knows more 00:18:09.849 --> 00:18:16.420 about you. So I would be extremely cautious about 00:18:16.420 --> 00:18:19.079 how much information I voluntarily feed into 00:18:19.079 --> 00:18:23.609 any of these systems, especially one designed 00:18:23.609 --> 00:18:28.569 to operate across multiple services. Assume anything 00:18:28.569 --> 00:18:33.130 you give an AI agent could become part of the 00:18:33.130 --> 00:18:35.910 context it uses to make future decisions unless 00:18:35.910 --> 00:18:40.599 the product clearly tells you otherwise. It's 00:18:40.599 --> 00:18:42.440 always better to err on the side of caution. 00:18:43.819 --> 00:18:46.660 Don't put sensitive information into it simply 00:18:46.660 --> 00:18:52.519 because the interface feels conversational. You're 00:18:52.519 --> 00:18:56.420 not talking to a friend. You're interacting with 00:18:56.420 --> 00:18:58.940 a piece of software connected to infrastructure 00:18:58.940 --> 00:19:03.200 operated by a mega billion dollar corporation. 00:19:04.259 --> 00:19:10.329 Treat your information accordingly. And I've 00:19:10.329 --> 00:19:12.970 been stressing this this whole episode. Think 00:19:12.970 --> 00:19:18.269 about The failure mode. This is something engineers 00:19:18.430 --> 00:19:27.839 do all the time. Don't just ask how something 00:19:27.839 --> 00:19:33.420 works when everything goes correctly. Ask how 00:19:33.420 --> 00:19:38.910 it fails. What's the failure mode? With an AI 00:19:38.910 --> 00:19:42.049 assistant that summarizes my notes, maybe the 00:19:42.049 --> 00:19:44.369 failure mode is that it summarizes something 00:19:44.369 --> 00:19:50.400 incorrectly. Annoying. Yeah. With an AI agent, 00:19:50.529 --> 00:19:53.029 connected to my personal accounts, the failure 00:19:53.029 --> 00:19:59.980 mode can be much more serious. It sends something 00:19:59.980 --> 00:20:03.099 I didn't intend to send. It deletes something. 00:20:03.119 --> 00:20:05.819 It accepts something. It buys something. It reveals 00:20:05.819 --> 00:20:10.380 something. It misunderstands a certain instruction. 00:20:13.369 --> 00:20:16.690 It gives information to the wrong person. Someone 00:20:16.690 --> 00:20:19.309 manipulates it. Someone compromises the account. 00:20:19.890 --> 00:20:22.829 If permission behaves differently, then maybe 00:20:22.829 --> 00:20:27.279 I thought it would. That's another category of 00:20:27.279 --> 00:20:32.440 risk. And the Matt Robb situation is nearly a 00:20:32.440 --> 00:20:38.380 perfect demonstration of why they need some safety 00:20:38.380 --> 00:20:42.240 rails, something on these systems. The system 00:20:42.240 --> 00:20:45.119 apparently operated within permissions that I 00:20:45.119 --> 00:20:49.250 guess existed. The human thought those permissions 00:20:49.250 --> 00:20:53.380 meant something much more limited. The computer 00:20:53.380 --> 00:20:56.380 did what it believed that it was allowed to do. 00:20:57.170 --> 00:21:01.980 and a stranger arrived at somebody's home. Now 00:21:01.980 --> 00:21:04.980 that's not a minor UX problem, man. That's a 00:21:04.980 --> 00:21:11.210 physical world consequence. If you want, if you 00:21:11.210 --> 00:21:17.960 want my advice, here's what I'll tell you. Use 00:21:17.960 --> 00:21:21.259 AI, experiment with it, build things, use it 00:21:21.259 --> 00:21:23.579 for research, use it to code, use it to summarize 00:21:23.579 --> 00:21:27.569 documents, use it to brainstorm if you got half 00:21:27.569 --> 00:21:31.099 of an idea. It can really help fill in some gaps 00:21:31.099 --> 00:21:33.200 or point you in some directions you maybe didn't 00:21:33.200 --> 00:21:37.670 think of before That's how a lot of these episodes 00:21:37.670 --> 00:21:41.410 come about I get kind of the root the base layer 00:21:41.410 --> 00:21:46.170 of an idea and I throw it, you know against Claude 00:21:46.170 --> 00:21:48.890 and I ask it, you know, hey, I'm thinking about 00:21:48.890 --> 00:21:53.750 this What's some other angles that we can? include 00:21:53.750 --> 00:21:57.089 and it gives me it gives me list now some of 00:21:57.059 --> 00:21:59.460 It's suggestions I don't like and I ignore them 00:21:59.460 --> 00:22:01.500 or tell them, Hey, let's try again. You know, 00:22:01.519 --> 00:22:04.240 maybe go this direction. It's a useful tool. 00:22:07.279 --> 00:22:12.309 I, so using it to brainstorm is one I like. Use 00:22:12.309 --> 00:22:16.029 it to automate some boring work, but be extremely 00:22:16.029 --> 00:22:19.890 cautious about AI agents that want deeper access 00:22:19.890 --> 00:22:25.069 into your personal life. We need boundaries. 00:22:26.410 --> 00:22:30.230 I would not, you know, yeah, I wouldn't give 00:22:30.319 --> 00:22:33.519 one unrestricted access to everything. I definitely 00:22:33.519 --> 00:22:35.700 wouldn't give Claude that, I wouldn't give OpenAI 00:22:35.700 --> 00:22:40.069 that, Gemini, any of these tools. Definitely 00:22:40.069 --> 00:22:46.210 not giving it to Muse. I wouldn't casually connect 00:22:46.210 --> 00:22:50.440 every account because the onboarding screen makes 00:22:50.440 --> 00:22:54.119 it convenient and you know, maybe I don't need 00:22:54.119 --> 00:23:02.170 it. And absolutely, I would not assume that because 00:23:02.170 --> 00:23:05.670 something has an AI logo on it for whatever company 00:23:05.670 --> 00:23:09.890 and a friendly little chat interface or This 00:23:09.890 --> 00:23:12.470 new thing that chat GPT is doing with dots. They 00:23:12.470 --> 00:23:18.009 look all nice and cute. I get it, but always 00:23:18.009 --> 00:23:24.150 keep your privacy in mind. Think before you hit 00:23:24.150 --> 00:23:28.589 allow. The more these systems can do, the more 00:23:28.589 --> 00:23:35.109 useful they might become. But remember, where's 00:23:35.109 --> 00:23:39.700 all that data going? I should also say, you know, 00:23:39.720 --> 00:23:43.539 the more these systems can do, the more careful 00:23:43.539 --> 00:23:46.019 we need to become about what we're allowing them 00:23:46.019 --> 00:23:52.039 to do. Every new piece of technology, it makes 00:23:52.039 --> 00:23:54.740 things a little more convenient, right? Convenience 00:23:54.740 --> 00:24:00.730 is great, but I don't need an AI assistant badly 00:24:00.730 --> 00:24:02.930 enough to have a stranger standing at my front 00:24:02.930 --> 00:24:05.109 door because the computer decided, hey, we're 00:24:05.109 --> 00:24:08.599 going to make a deal. I'll answer my own Facebook 00:24:08.599 --> 00:24:11.299 marketplace messages. That's what I got for you 00:24:11.299 --> 00:24:14.690 today. Hit up tylerwoodward .com. Well, it's 00:24:14.690 --> 00:24:17.130 tylerwoodward .me, but .com also redirects to 00:24:17.130 --> 00:24:19.329 that. So you're good there if you go to the wrong 00:24:19.329 --> 00:24:22.849 one. If you like what I'm doing here, like and 00:24:22.849 --> 00:24:25.630 subscribe, give a rating and review there on 00:24:25.630 --> 00:24:29.109 Spotify, Apple, podcast, tune in, iHeart, wherever 00:24:29.109 --> 00:24:31.170 you're listening. And if you want to buy me a 00:24:31.170 --> 00:24:33.950 cup of coffee, just to show you, you know, your 00:24:33.950 --> 00:24:36.130 appreciation, hey, you can do that too at tylerwoodward 00:24:36.130 --> 00:24:38.390 .me. And if you want to leave a voicemail for 00:24:38.390 --> 00:24:43.200 the show, 608 -668 -4702. I'll catch you next 00:24:43.200 --> 00:24:43.420 time.