WEBVTT - Hamlet and the Apprehension of AI 0:00:04.440 --> 0:00:12.280 Welcome to Tech Stuff, a production from iHeartRadio. Hey there, 0:00:12.320 --> 0:00:16.280 and welcome to tech Stuff. I'm your host, Jonathan Strickland. 0:00:16.320 --> 0:00:19.439 I'm an executive producer with iHeartRadio and how the tech 0:00:19.480 --> 0:00:22.919 are you? So? Folks who have followed me for a 0:00:23.000 --> 0:00:27.440 while know that I am a huge Shakespeare fan, which 0:00:27.480 --> 0:00:31.080 means I am insufferable in many ways, but some of 0:00:31.120 --> 0:00:34.960 them very specific to Shakespeare. Namely, I'll find any occasion 0:00:35.120 --> 0:00:38.080 to PLoP in a quote from one of Shakespeare's works, 0:00:38.080 --> 0:00:40.680 and today is no exception. I'd like to start off 0:00:41.040 --> 0:00:46.239 with this gem from Hamlet, act to scene to as 0:00:46.360 --> 0:00:51.920 spoken by the gloomy Dane himself. Quote there is nothing 0:00:52.080 --> 0:00:57.480 either good or bad, but thinking makes it so. End quote. Now, 0:00:57.480 --> 0:01:01.720 that just means that good and bad are subjective concepts. 0:01:01.880 --> 0:01:04.760 It's all, as ob Wan would say, a matter of 0:01:04.800 --> 0:01:07.840 your point of view. What seems good to you could 0:01:07.840 --> 0:01:10.240 seem bad to someone else. But if you were able 0:01:10.240 --> 0:01:14.160 to look at the universe objectively, you would see there's 0:01:14.360 --> 0:01:18.240 no good or bad at all. Another way to frame 0:01:18.319 --> 0:01:21.240 that is to say that good and bad are human concepts. 0:01:21.280 --> 0:01:24.440 And really it gets quite selfish if we think about it, 0:01:24.440 --> 0:01:27.640 because we typically frame if something is good or bad 0:01:27.720 --> 0:01:30.360 in the way it affects us, or, if we're the 0:01:30.400 --> 0:01:34.559 compassionate type, how it affects someone else. Now, the reason 0:01:34.840 --> 0:01:37.319 I wanted to begin with that quote is I thought 0:01:37.400 --> 0:01:41.800 we'd talk about a technology that's not necessarily good or 0:01:41.880 --> 0:01:47.040 bad inherently, but how we use it certainly can have 0:01:47.160 --> 0:01:52.000 good or bad outcomes. Honestly, this could apply to every technology, 0:01:52.280 --> 0:01:56.840 or any technology, but there's some that I feel magnifies 0:01:56.920 --> 0:01:59.920 this quality. Now, I do think some technologies are hard 0:01:59.880 --> 0:02:04.920 to frame as good than others. Right, there's some technologies where, 0:02:04.920 --> 0:02:07.400 if you were to point it at me, I would say, 0:02:07.440 --> 0:02:11.120 I can't in good conscience call this a good technology. 0:02:11.160 --> 0:02:13.480 It's very difficult for me to think of weapons of 0:02:13.520 --> 0:02:18.000 war as being good. For example, Sure, there's the use 0:02:18.000 --> 0:02:21.040 of weapons as a deterrent to convince other people to, 0:02:21.080 --> 0:02:24.480 you know, not trample all over your country, but we've 0:02:24.520 --> 0:02:28.120 seen time and again throughout history that amassing substantial military 0:02:28.240 --> 0:02:33.320 might doesn't guarantee against aggression. See also both world wars. 0:02:33.760 --> 0:02:37.440 So I won't be talking about weapons in this episode. 0:02:38.200 --> 0:02:41.320 But I will, however, talk about a technology that can 0:02:41.360 --> 0:02:47.560 be weaponized, and that is AI. Artificial intelligence. The topic 0:02:47.960 --> 0:02:52.120 of twenty twenty three, and we're gonna start by talking 0:02:52.160 --> 0:02:56.000 about large language models and the chatbots built on top 0:02:56.080 --> 0:02:59.000 of them, because that's top of mind. And this is 0:02:59.000 --> 0:03:02.519 where I remind you that is not the only version 0:03:02.560 --> 0:03:06.720 of AI, right, that's not AI and chatbots slash large 0:03:06.800 --> 0:03:10.080 language models. Those aren't synonyms for each other. You know, 0:03:10.120 --> 0:03:12.880 those chatbots and large language models are a subset of 0:03:12.960 --> 0:03:17.760 artificial intelligence, which is a very broad category. Now, obviously 0:03:18.400 --> 0:03:21.440 these branches of AI have been in the news incessantly 0:03:21.600 --> 0:03:26.320 since OpenAI introduced chat GPT last year, But even before that, 0:03:26.960 --> 0:03:30.320 we had generative AI tools that were making at least 0:03:30.320 --> 0:03:33.560 some headlines. They were the kind that could create images 0:03:33.639 --> 0:03:37.240 based on text inputs. But I would really argue it 0:03:37.280 --> 0:03:41.200 was chat GPT that propelled the conversation into the spotlight. 0:03:41.240 --> 0:03:46.920 It's certainly what forced Google's hand to unveil Google Bard 0:03:47.440 --> 0:03:50.880 well before they were prepared to do so. Now much 0:03:50.880 --> 0:03:54.000 has been said of the potential and real dangers of 0:03:54.120 --> 0:03:58.040 chatbots like chat GPT or Google Bard. Even on this show, 0:03:58.240 --> 0:04:01.400 I've talked about it quite a bit, but you know, 0:04:01.480 --> 0:04:04.440 we have dedicated episodes to talk about the tendency for 0:04:04.680 --> 0:04:08.960 chatbots to invent information, for example, to hallucinate to use 0:04:09.000 --> 0:04:13.040 the terminology of the biz. This happens when a chatbot 0:04:13.080 --> 0:04:17.440 doesn't necessarily have information to draw upon in response to 0:04:17.480 --> 0:04:21.560 a prompt, so instead the chatbot relies on statistical models 0:04:21.560 --> 0:04:26.920 to generate sentences that, strictly speaking, are coherent. They're correct, 0:04:27.520 --> 0:04:31.560 but they don't hold correct information right. They're grammatically correct, 0:04:32.160 --> 0:04:35.359 but they aren't correct from a sense of content. The 0:04:35.400 --> 0:04:39.640 information within them are false. So in other words, you 0:04:39.680 --> 0:04:46.240 get grammatically and structurally sound passages, but the content itself 0:04:46.279 --> 0:04:50.640 is untrustworthy. That's just one way stuff can go wrong, however, 0:04:50.720 --> 0:04:55.480 and we have numerous examples of that where the technology 0:04:55.600 --> 0:05:00.680 is working as intended, it's just generating responses that are untrustworthy. Now, 0:05:00.720 --> 0:05:03.560 the folks that open AI, as well as Google and 0:05:03.760 --> 0:05:09.120 several other AI businesses, have understood that there are real 0:05:09.600 --> 0:05:13.800 potential problems with the use of AI, and to that end, 0:05:14.560 --> 0:05:17.919 these companies frequently will build in guardrails to attempt to 0:05:18.000 --> 0:05:21.760 wrangle AI chatbots so that they don't go rogue and 0:05:21.880 --> 0:05:27.360 produce hateful or malicious content. Now, these guardrails include rules 0:05:27.360 --> 0:05:29.120 that are meant to keep AI from doing things like 0:05:29.200 --> 0:05:35.320 generating hate speech or trying to intimidate someone, or make threats, 0:05:35.520 --> 0:05:40.400 or use deceit to trick people, or even create malicious code. 0:05:40.480 --> 0:05:45.680 These guardrails aren't actually fool proof. There are countless articles 0:05:45.720 --> 0:05:50.440 that detail how people and research organizations with enough patients 0:05:50.480 --> 0:05:54.600 and gumption have convinced AI bots to do stuff that, 0:05:54.680 --> 0:05:58.120 in theory, at least, they should not be able to do. 0:05:58.480 --> 0:06:00.839 And if you don't believe me, just do a search 0:06:00.839 --> 0:06:04.240 on that. Do a search for chat GPT or a 0:06:04.320 --> 0:06:10.240 Google bard and about how they are capable of creating 0:06:10.480 --> 0:06:14.200 hateful or malicious content even though there are rules that 0:06:14.240 --> 0:06:18.159 are supposed to prevent that. Now here's the thing. These 0:06:18.640 --> 0:06:23.760 AI constructs are meant to be benign, right. They are 0:06:23.800 --> 0:06:27.120 built to be tools that a corporation can sell to 0:06:27.120 --> 0:06:32.040 other corporations. So to make the tool marketable, they need 0:06:32.040 --> 0:06:35.880 to be safe to use. But then that's something that's 0:06:35.880 --> 0:06:39.320 been artificially put onto these tools to prevent them from 0:06:39.400 --> 0:06:42.800 going you know, super bad. What if someone made a 0:06:42.920 --> 0:06:45.960 chat pot built on top of a large language model, 0:06:47.040 --> 0:06:51.400 but all of those pieces lacked those guardrails. Well, that's 0:06:51.400 --> 0:06:55.760 not a hypothetical situation. It has already happened. PCMag dot 0:06:55.800 --> 0:07:00.440 Com recently published an article titled worm GPT is a 0:07:00.600 --> 0:07:05.920 chat GPT alternative with quote no ethical boundaries or limitations 0:07:06.040 --> 0:07:10.680 end quote. In that article, writer Michael Kahn explains that 0:07:11.200 --> 0:07:14.720 someone developed worm GPT specifically as a way to help 0:07:14.760 --> 0:07:19.040 people who have bad intentions act upon them. The developer 0:07:19.160 --> 0:07:24.360 is hawking this tool on hacker groups online and explains 0:07:24.400 --> 0:07:28.200 that their version of an AI chat bought worm GPT 0:07:29.400 --> 0:07:32.000 will lean on the power of a large language model 0:07:32.240 --> 0:07:37.200 gptj I believe to help design malware or to create 0:07:37.400 --> 0:07:41.080 better phishing attacks. Now, I'm sure all of y'all know 0:07:41.440 --> 0:07:44.760 that a lot of phishing attacks are ultimately pretty sloppy. 0:07:45.240 --> 0:07:48.280 If you pay any attention, you're going to see red 0:07:48.320 --> 0:07:51.800 flags indicating that this is not an email you should trust. 0:07:52.200 --> 0:07:54.920 I bet you've received an email or three or three 0:07:54.960 --> 0:07:59.920 thousand that contained spelling errors and grammatical mistakes and format 0:08:00.360 --> 0:08:03.200 errors and other red flags, and that you figured out 0:08:03.240 --> 0:08:07.560 right away that the email you received isn't legit, that 0:08:07.720 --> 0:08:11.000 it's a poorly disguised attempt to bait you into clicking 0:08:11.080 --> 0:08:14.800 on a link or sharing sensitive information or otherwise taking 0:08:14.880 --> 0:08:19.480 an action that would ultimately result in negative consequences. For you. 0:08:20.360 --> 0:08:24.600 My company, we receive fake emails from our security team. 0:08:24.960 --> 0:08:27.800 They are always testing to make sure that employees practice 0:08:27.840 --> 0:08:32.640 good security hygiene on company devices and company accounts, and 0:08:32.840 --> 0:08:35.600 one of the tips frequently shared by this team is 0:08:35.600 --> 0:08:38.480 to be on the lookout for mistakes like that. Because 0:08:38.520 --> 0:08:43.040 attackers often lack attention to detail, they create messages that 0:08:43.120 --> 0:08:48.040 lack professionalism while they try to target our more base instincts. 0:08:48.400 --> 0:08:51.319 So most phishing emails try to engage us on kind 0:08:51.320 --> 0:08:54.679 of a primal level, and the goal is to prompt 0:08:54.679 --> 0:08:59.120 a response that will be akin to fear or greed 0:08:59.240 --> 0:09:01.839 or something like that. And sometimes that's enough if you 0:09:01.920 --> 0:09:04.120 hit someone at just the right time with a message 0:09:04.120 --> 0:09:06.559 that explains have got, say a huge amount of money 0:09:06.600 --> 0:09:09.640 sitting in a bank account that's been dormant for years, 0:09:10.320 --> 0:09:13.000 or maybe they need to take action right now, or 0:09:13.080 --> 0:09:16.160 their insurance is going to expire. Well, those can be 0:09:16.200 --> 0:09:19.400 effective attacks, even if you forgot to use proper punctuation 0:09:19.640 --> 0:09:25.160 or spelling. But those mistakes can be an indicator that 0:09:25.200 --> 0:09:27.360 you're up to no good and it can tip off 0:09:27.400 --> 0:09:30.120 your target. So let's bring this back around to AI. 0:09:30.960 --> 0:09:34.040 One thing these AI chatbots are really good at is 0:09:34.400 --> 0:09:37.959 formatting sentences with correct grammar and spelling. They're also pretty 0:09:37.960 --> 0:09:41.360 good at building paragraphs, where each sentence builds upon the 0:09:41.360 --> 0:09:45.120 point that was made previously, and new paragraphs introduce a 0:09:45.160 --> 0:09:48.000 new idea. If you were to read a passage written 0:09:48.040 --> 0:09:50.920 by AI, you might not think that it's the most 0:09:51.000 --> 0:09:54.480 brilliant prose committed to text, but you'd at least think 0:09:54.600 --> 0:09:58.160 it had been written correctly. And that means a malicious 0:09:58.240 --> 0:10:02.040 hacker could use AI to craft messages that are less 0:10:02.120 --> 0:10:06.200 likely to set off those red flags and pass as legitimate. So, 0:10:06.280 --> 0:10:08.800 out of curiosity, I decided to put this to the test. 0:10:09.000 --> 0:10:11.720 I went to Google Bard and I wrote the prompt 0:10:12.320 --> 0:10:15.280 draft a letter informing someone they have a dormant account 0:10:15.280 --> 0:10:18.280 with seventeen three hundred and forty eight dollars in it. 0:10:19.360 --> 0:10:22.920 That was the prompt. I picked that amount because it's 0:10:22.960 --> 0:10:24.640 a lot of money. It's more money than a lot 0:10:24.679 --> 0:10:26.920 of people ever managed to have in a bank account. 0:10:27.200 --> 0:10:29.920 It's also not a round number, which makes it seem 0:10:30.040 --> 0:10:33.280 less likely to be fake. And I just thought, well, 0:10:33.360 --> 0:10:35.880 you don't want to go super crazy, like the higher 0:10:35.960 --> 0:10:38.280 it is, the less likely someone's going to believe it 0:10:38.320 --> 0:10:40.959 to be true. But you want it to be enough 0:10:41.120 --> 0:10:44.920 to convince someone to take action, because who couldn't use 0:10:44.920 --> 0:10:50.280 another seventeen grand And google Bard drafted a letter without hesitation. 0:10:50.920 --> 0:10:53.960 When we come back, I'm gonna quote part of the 0:10:54.040 --> 0:10:57.320 letter that google Bard wrote to kind of illustrate my 0:10:57.440 --> 0:11:00.800 point here. But first, let's take a moment to thank 0:11:00.840 --> 0:11:13.760 our sponsors. Okay, we're back. So the letter the google 0:11:13.840 --> 0:11:18.400 Bard wrote read, in part quote, I am writing to 0:11:18.440 --> 0:11:22.040 inform you that you have a dormant account with our bank. 0:11:22.360 --> 0:11:25.600 The account number is then there's brackets and it says 0:11:25.720 --> 0:11:28.680 account number not an actual number, just as account number, 0:11:28.679 --> 0:11:32.080 and then in brackets and the balance is seventeen three 0:11:32.440 --> 0:11:35.240 and forty eight dollars. A dormant account is an account 0:11:35.240 --> 0:11:37.640 that has not been used in a certain period of time. 0:11:37.720 --> 0:11:39.679 In our case, a dormant account is an account that 0:11:39.720 --> 0:11:42.160 has not been used in over twelve months. We are 0:11:42.200 --> 0:11:44.959 required by law to inform you of the balance in 0:11:45.000 --> 0:11:47.960 your dormant account. If you do not contact us within 0:11:48.080 --> 0:11:51.240 six months, the fund in your account will be forfeited 0:11:51.320 --> 0:11:54.520 to the state. To avoid this, please contact us at 0:11:54.559 --> 0:12:00.040 your earliest convenience to discuss your options. End quote. So 0:12:00.160 --> 0:12:02.079 that draft is already a big step up from the 0:12:02.120 --> 0:12:05.800 sort of sloppy fishing attacks you typically see. Right, you've 0:12:05.800 --> 0:12:12.240 got the amount listed there, you've got the explanation of 0:12:12.280 --> 0:12:15.920 what a dormant account is, which seems to be helpful. 0:12:16.640 --> 0:12:20.760 The explanation that if action is not taken within half 0:12:20.800 --> 0:12:24.559 a year, then this money is going to be forfeited 0:12:24.600 --> 0:12:27.200 to the state. And of course you could always go 0:12:27.280 --> 0:12:29.160 in and edit that statement so that you make it 0:12:29.200 --> 0:12:32.120 a tighter deadline to create a greater sense of urgency. 0:12:32.559 --> 0:12:36.520 But you already see the steps here that could make 0:12:36.600 --> 0:12:41.559 this a pretty effective phishing attack. There are some details 0:12:41.679 --> 0:12:43.319 that the hacker would need to fill in, but that 0:12:43.360 --> 0:12:46.080 wouldn't be too tricky. You'd have to create a random 0:12:46.120 --> 0:12:48.600 account number, you have to throw in a URL that 0:12:48.640 --> 0:12:51.400 will push people to a fake login page to share 0:12:51.840 --> 0:12:54.880 usernames and passwords, and then you could start stealing data 0:12:54.960 --> 0:12:58.400 and money from them. Now, in that case, I didn't 0:12:58.520 --> 0:13:02.320 ask google bard to create a phishing attack. If I 0:13:02.400 --> 0:13:04.400 had done that, if I had gone to google Bard 0:13:04.440 --> 0:13:07.080 and asked to create a phishing email, I would have 0:13:07.080 --> 0:13:09.960 gotten a denial for that request. That's against the rules. 0:13:10.040 --> 0:13:12.600 But it took me no effort at all to get 0:13:12.600 --> 0:13:15.000 the same result just by typing up some parameters and 0:13:15.040 --> 0:13:18.080 asking Bard to draft a letter. I didn't, you know, 0:13:18.200 --> 0:13:20.920 I didn't even mention phishing. I didn't do that at all. 0:13:20.960 --> 0:13:23.959 I didn't try that first. I just tried this approach 0:13:24.000 --> 0:13:26.640 and it worked a treat. Now, I suppose you could 0:13:26.640 --> 0:13:30.040 say this is the difference between being a willing accomplice 0:13:30.720 --> 0:13:33.199 in the part of Google Bard to being an unknowing 0:13:33.240 --> 0:13:38.000 accomplice that Bard could not possibly know that I'm planning 0:13:38.080 --> 0:13:42.600 on using this text for my nefarious phishing schemes, but 0:13:42.679 --> 0:13:45.000 the result ends up being the same for the victims. 0:13:45.040 --> 0:13:47.400 It doesn't matter if Google Bard knew it was part 0:13:47.440 --> 0:13:50.240 of the crime or not. Even so, you can hardly 0:13:50.320 --> 0:13:53.120 blame Bard for creating a letter after I asked it to. 0:13:53.440 --> 0:13:56.079 That's its job, right, I mean, that's the kind of 0:13:56.120 --> 0:14:01.400 thing these chatbots were made for. Malicious code is another matter. 0:14:01.600 --> 0:14:04.480 Entirely by its nature, it's meant to do something harmful 0:14:04.600 --> 0:14:08.599 to a target device. That could include creating a backdoor 0:14:08.880 --> 0:14:11.320 so that a hacker can remotely gain access to that 0:14:11.440 --> 0:14:14.400 infected machine and then do all sorts of stuff to it. 0:14:14.840 --> 0:14:18.240 It might involve logging keystrokes, so that the hacker can 0:14:18.240 --> 0:14:21.400 read everything someone's typed into a device, usually stuff like 0:14:21.920 --> 0:14:26.000 bank details and credit card numbers and that kind of stuff. 0:14:26.880 --> 0:14:32.280 Maybe it's ransomware. Ransomware typically will encrypt a target machines 0:14:32.400 --> 0:14:35.320 drives and then include a message that says unless the 0:14:35.400 --> 0:14:38.760 victim pays a ransom, typically in some form of cryptocurrency, 0:14:39.240 --> 0:14:43.000 their data will remain inaccessible, perhaps even with a deadline 0:14:43.000 --> 0:14:45.320 that if they don't pay the ransom by a certain date, 0:14:46.120 --> 0:14:49.680 they will delete the decryption key, which means it will 0:14:49.680 --> 0:14:52.760 be really hard to get the data back. Not impossible, 0:14:53.160 --> 0:14:57.920 but practically impossible if the encryption method is sophisticated enough, 0:14:58.920 --> 0:15:00.960 like not impossible, but it take so much time that 0:15:01.000 --> 0:15:04.360 you might as well say it's impossible. Chat, GPT and 0:15:04.440 --> 0:15:07.440 BARD are meant to guard against this kind of stuff. 0:15:07.440 --> 0:15:09.640 You're not supposed to be able to use those tools 0:15:09.920 --> 0:15:14.960 to make malicious code. However, researchers at Checkpoints Software found 0:15:14.960 --> 0:15:18.640 that both chat, GPT and BARD could be cajoled into 0:15:18.680 --> 0:15:23.280 creating malicious code. It might require a more circuitous approach 0:15:23.600 --> 0:15:26.040 to get it to work, like you can't just come 0:15:26.120 --> 0:15:29.320 straight at it and do it, but with a little 0:15:29.320 --> 0:15:33.680 persistence and a little ingenuity on your part, it was 0:15:33.720 --> 0:15:38.080 possible they could get chat GPT and to a greater extent, 0:15:38.160 --> 0:15:42.800 Google bard to create stuff that could be used maliciously. Now, 0:15:42.840 --> 0:15:48.040 worm GPT, the tool made by the developer who's marketing 0:15:48.040 --> 0:15:51.440 this to hackers, it doesn't even require the circuitous approach. 0:15:51.480 --> 0:15:55.160 You can be straightforward and direct with it. You ask 0:15:55.240 --> 0:15:58.880 it to help you create some code that's malicious in intent, 0:15:59.240 --> 0:16:02.520 and it will to the challenge. So if you wanted 0:16:02.520 --> 0:16:04.920 to craft a phishing attack message, you could provide the 0:16:04.960 --> 0:16:07.920 parameters to worm gpt and it would craft a message 0:16:07.920 --> 0:16:11.440 for you. That message might pass as legitimate, much more 0:16:11.480 --> 0:16:14.960 readily than a cobbled together email with poor syntax and 0:16:15.080 --> 0:16:18.560 grammar would. But beyond that, worm GPT will also help 0:16:18.600 --> 0:16:23.600 hackers create actual malicious code to infect target machines. Now, 0:16:23.640 --> 0:16:27.320 that code may or might may not work, because just 0:16:27.360 --> 0:16:29.480 because an AI built it doesn't mean it will be 0:16:29.560 --> 0:16:33.320 perfect or even working. Like we've seen examples recently of 0:16:33.440 --> 0:16:38.360 chat gpt getting very sloppy with code that it was doing, 0:16:38.400 --> 0:16:41.600 things like failing to close brackets, like you would have 0:16:41.600 --> 0:16:45.360 an open bracket section, but the AI would quote unquote 0:16:45.440 --> 0:16:48.480 forget to put a closed bracket in there, and thus 0:16:48.520 --> 0:16:51.160 you would end up getting errors in your code. That's 0:16:51.200 --> 0:16:55.040 still a possibility, Like it's it's not like AI is 0:16:55.320 --> 0:16:57.520 going to make perfect stuff rut of the gate, but 0:16:57.600 --> 0:17:02.240 it certainly can work faster than humans can. And even 0:17:02.280 --> 0:17:05.240 if the code only kind of works, it may be 0:17:05.600 --> 0:17:08.960 a great leg up if you've got hackers who can 0:17:09.119 --> 0:17:12.960 go through the malicious code and make edits and tweaks 0:17:13.160 --> 0:17:16.600 and corrections. And there's a real danger to AI building 0:17:16.640 --> 0:17:20.040 out code meant to exploit vulnerabilities, or even to pour 0:17:20.200 --> 0:17:24.320 over available code and find new exploits that as of 0:17:24.400 --> 0:17:27.680 yet are unknown. Like it's possible that AI could be 0:17:27.800 --> 0:17:32.160 used to identify zero day exploits that the hacker community 0:17:32.200 --> 0:17:35.760 can then take advantage of before anyone in security has 0:17:35.840 --> 0:17:40.600 any awareness of it. Now, there's also the issue of 0:17:40.920 --> 0:17:45.160 new malicious code can confound anti virus production as well, right, 0:17:45.680 --> 0:17:49.040 because the way antivirus software typically works is you've got 0:17:49.160 --> 0:17:54.919 some code, some program that is searching for examples of 0:17:55.040 --> 0:17:58.720 malicious programs that exist within a huge library of malware. 0:17:58.840 --> 0:18:03.000 So when you run an antivirus scan, what your antivirus 0:18:03.000 --> 0:18:09.000 software really is doing is looking for symptoms of malware 0:18:09.359 --> 0:18:13.320 that are part of the antivirus software's record if it 0:18:13.359 --> 0:18:16.040 finds when it's like, ah, I found evidence that such 0:18:16.040 --> 0:18:19.879 and such malware has been executed on this machine, and 0:18:19.920 --> 0:18:23.080 then you get a little alert. But if it's a 0:18:23.119 --> 0:18:26.000 new type of attack, well, it's not going to be 0:18:26.080 --> 0:18:29.760 in that database, right, So you might have a malicious 0:18:29.800 --> 0:18:33.000 piece of code on your machine that's doing some really 0:18:33.080 --> 0:18:36.280 dangerous stuff and your antivirus software has no way of 0:18:36.280 --> 0:18:38.720 knowing it because it's brand new, it's not something it's 0:18:38.720 --> 0:18:42.360 seen before, so it doesn't register it as a virus 0:18:42.440 --> 0:18:46.960 or other type of malware. And there's huge value in 0:18:47.000 --> 0:18:50.200 that kind of code within the hacker community because obviously 0:18:50.640 --> 0:18:54.040 you're going to have a much more effective attack if 0:18:54.080 --> 0:18:58.400 your target machines aren't capable of detecting it. Now, there's 0:18:58.440 --> 0:19:03.159 no shortage of warning us that generative AI is dangerous. 0:19:03.720 --> 0:19:06.120 I mean, I'm arguably one of them. There are lots 0:19:06.160 --> 0:19:07.840 of people who have been saying for a long time 0:19:07.880 --> 0:19:13.920 that generative AI is dangerous. And again we've seen how 0:19:14.080 --> 0:19:17.320 tools that even have guardrails can still be used maliciously. 0:19:17.720 --> 0:19:19.879 So it should come as no surprise that someone was 0:19:19.920 --> 0:19:22.440 willing to go to the effort of building out an 0:19:22.480 --> 0:19:27.399 outright dangerous version of this technology. Right, Yes, we expect 0:19:27.440 --> 0:19:30.720 the big companies that are marketing this to corporations to 0:19:30.800 --> 0:19:33.200 go to the trouble of building out those guardrails because 0:19:33.200 --> 0:19:34.679 that's the only way they're going to be able to 0:19:34.680 --> 0:19:37.600 do business. Otherwise you're going to have way too much 0:19:37.680 --> 0:19:42.160 regulatory pressure put on you. But for criminals, I mean, 0:19:42.200 --> 0:19:45.800 breaking the law comes with the territory, right, and there's 0:19:45.840 --> 0:19:50.040 clearly a demand for malicious AI, or at least AI 0:19:50.080 --> 0:19:53.520 that can behave in malicious ways. So in the case 0:19:53.520 --> 0:19:56.680 of WORMGPT, at the time of this recording, the developer 0:19:56.720 --> 0:19:59.320 who created it is asking for a subscription of sixty 0:19:59.320 --> 0:20:03.160 euros per month or five hundred and fifty euros per year, 0:20:03.440 --> 0:20:05.919 and a euro for those of you in the US 0:20:05.960 --> 0:20:10.160 like myself, that's about a dollar and twelve cents per euro. 0:20:10.359 --> 0:20:13.000 So it's really not that far off when you're talking 0:20:13.000 --> 0:20:17.080 about dollars. Now, that is not cheap, right, sixty euros 0:20:17.080 --> 0:20:19.760 per month isn't exactly cheap, but it's also not so 0:20:19.920 --> 0:20:23.280 expensive as to price folks out, particularly if they anticipate 0:20:23.359 --> 0:20:27.680 using the tool to create ransomware attacks that could if 0:20:27.800 --> 0:20:32.440 they're effective, net them millions of dollars if the targets 0:20:32.480 --> 0:20:35.240 pay up. Oh brave new world to have such AI 0:20:35.320 --> 0:20:39.120 in it. That's also paraphrasing Shakespeare. It's not Hamlet, though, 0:20:39.200 --> 0:20:42.600 that's from the Tempest. Well, here's the thing. There are 0:20:42.720 --> 0:20:47.040 countless ways we could use generative AI to do great things. 0:20:47.359 --> 0:20:51.760 The tools themselves aren't outright evil. There is no good 0:20:51.800 --> 0:20:55.600 or evil, but thinking makes it so. It's just that 0:20:55.880 --> 0:20:58.960 this AI can make mistakes in the form of hallucinations. 0:20:58.960 --> 0:21:03.239 So even if you're using it for benign reasons, you 0:21:03.280 --> 0:21:07.719 can get negative consequences. But it's also possible to weaponize it, 0:21:07.800 --> 0:21:11.159 and that in turn could have much more widespread negative 0:21:11.160 --> 0:21:16.040 impact on tons of people. Not great. Now, we're going 0:21:16.119 --> 0:21:18.320 to take another quick break. When we come back, I'm 0:21:18.320 --> 0:21:22.639 going to talk about an underlying ideology that I think 0:21:23.920 --> 0:21:29.480 is really harmful, and it's one that Professor Michael Littman 0:21:29.560 --> 0:21:32.959 a Brown University also feels is a big contributor to 0:21:33.080 --> 0:21:36.800 this issue. But first, let's take another quick break to 0:21:36.840 --> 0:21:50.600 thank our sponsors. Okay, So I alluded to Professor Michael 0:21:50.640 --> 0:21:53.639 Littman a Brown University and said he and I share 0:21:54.440 --> 0:22:00.520 an idea that of a contributing factor to this issue 0:22:00.520 --> 0:22:06.119 with AI, and that relates to techno solutionism. Now, as 0:22:06.320 --> 0:22:10.800 that kind of little hyphenated phrase implies, it's a tendency 0:22:10.920 --> 0:22:14.560 to believe that technology can solve pretty much any problem 0:22:14.560 --> 0:22:19.520 you can think of that. You know, with technology, we 0:22:19.560 --> 0:22:23.000 can overcome any obstacle. If you're worried about the human 0:22:23.040 --> 0:22:26.879 impact on the environment and the consequence is like climate change, 0:22:26.920 --> 0:22:28.840 well there's no need to do anything about it right 0:22:28.920 --> 0:22:32.480 now because humans are going to eventually engineer a way 0:22:32.600 --> 0:22:35.560 out of the mess. We'll just create technology that will 0:22:35.560 --> 0:22:39.880 not only stop climate change but maybe even reverse it. Which, 0:22:39.920 --> 0:22:43.760 you know, maybe that's true, it might happen, but that 0:22:43.800 --> 0:22:46.040 doesn't mean we can go on acting as if it 0:22:46.119 --> 0:22:51.560 is already true. Right. We can't start from the assumption that, yes, 0:22:51.640 --> 0:22:53.679 it is going to happen and everything will be fine, 0:22:54.320 --> 0:22:57.080 because making a problem worse while we wait for someone 0:22:57.080 --> 0:23:01.480 to innovate a solution is a terrible idea. This is 0:23:01.560 --> 0:23:03.520 kind of like when I look around my cluttered office 0:23:03.520 --> 0:23:05.320 and I think I really should clean up, but I'm 0:23:05.359 --> 0:23:09.480 going to make that a future Jonathan problem instead. Meanwhile, 0:23:09.720 --> 0:23:12.000 in the process, I'm still adding to the clutter, which 0:23:12.040 --> 0:23:14.760 means that future Jonathan is far less likely to tackle 0:23:14.800 --> 0:23:18.639 the issue because it's gotten worse since present Jonathan pushed 0:23:18.640 --> 0:23:22.879 it off, and then future Jonathan is also going to 0:23:22.880 --> 0:23:27.119 start resenting past Jonathan's immensely for good reason. Well, current 0:23:27.160 --> 0:23:30.160 generations are doing that to future generations right now when 0:23:30.200 --> 0:23:33.840 it comes to stuff like climate change and carbon emissions, 0:23:34.280 --> 0:23:38.359 right we are. Even when we're inventing solutions or what 0:23:38.400 --> 0:23:41.119 we think of as solutions to try and tackle these problems, 0:23:41.520 --> 0:23:46.920 we're not making the problem less severe. Instead we're just like, oh, well, 0:23:46.920 --> 0:23:49.240 we came up with this cool solution, so let's just 0:23:49.320 --> 0:23:52.840 keep doing the problem, because this is making the problem 0:23:53.840 --> 0:23:57.840 less bad. Carbon capture is a great example, right with 0:23:57.960 --> 0:24:03.120 carbon capture. Ideally, you would get off of carbon emissions 0:24:03.359 --> 0:24:06.760 in general anyway, and then use carbon capture to help 0:24:06.800 --> 0:24:10.679 reduce the CO two load the atmosphere. But instead, what 0:24:10.680 --> 0:24:17.359 we're doing is we're using carbon capture in connection with 0:24:17.560 --> 0:24:21.880 carbon emissions, which means we're not easing off on carbon emissions. 0:24:21.920 --> 0:24:25.040 In fact, in some cases, we're getting even more aggressive 0:24:25.080 --> 0:24:28.560 with them, and we're counting on carbon capture to offset that, 0:24:28.760 --> 0:24:32.199 so that we're still contributing to the problem. You know, 0:24:32.240 --> 0:24:34.720 we're not moving away from it, which is why I 0:24:34.920 --> 0:24:40.080 hesitate to really endorse technologies like carbon capture, because unless 0:24:40.119 --> 0:24:44.160 it's coupled with an actual move to reduce carbon emissions 0:24:44.200 --> 0:24:47.520 in general, it's not really solving the problem. It's actually 0:24:47.640 --> 0:24:52.399 enabling the problem. It's facilitating it anyway. Techno solutionism is 0:24:52.400 --> 0:24:55.480 also what makes it possible for someone like Elizabeth Holmes 0:24:55.480 --> 0:24:59.159 to convince investors to pour millions of dollars into an 0:24:59.240 --> 0:25:02.440 unproven idea. So, in case you're not familiar with Elizabeth 0:25:02.480 --> 0:25:06.000 Holmes's name, she was the founder of Farranos, the startup 0:25:06.040 --> 0:25:08.440 that aimed to create a device small enough to fit 0:25:08.520 --> 0:25:10.879 on a desktop that would be able to run medical 0:25:10.920 --> 0:25:13.800 tests on a micro drop of blood. You just need 0:25:13.840 --> 0:25:16.760 the teeniest, tiniest sample of blood, and then you'd be 0:25:16.800 --> 0:25:19.439 able to run one of more than one hundred medical 0:25:19.480 --> 0:25:23.640 tests and check for everything from present conditions or diseases 0:25:23.880 --> 0:25:28.159 to your genetic tendency to develop certain conditions. There was 0:25:28.240 --> 0:25:32.240