WEBVTT - Tech News: Samsung Vulnerabilities and Drastic Measures 0:00:04.400 --> 0:00:12.200 Welcome to tech Stuff, a production from iHeartRadio. Be there 0:00:12.240 --> 0:00:15.800 and welcome to tech Stuff. I'm your host, Jonathan Strickland. 0:00:15.840 --> 0:00:19.239 I'm an executive producer with iHeartRadio. And how the tech 0:00:19.320 --> 0:00:21.759 are you? It's time for the tech news for Tuesday, 0:00:21.840 --> 0:00:27.360 March twenty first, twenty twenty three. First up, Google's bug 0:00:27.480 --> 0:00:34.760 hunting team found eighteen zero day vulnerabilities in Samsung hardware, 0:00:35.040 --> 0:00:39.440 specifically the xin US chip sets. That's e x Y 0:00:39.920 --> 0:00:44.199 n os SO. Zero day vulnerabilities, as the name suggests, 0:00:44.200 --> 0:00:49.600 are vulnerabilities that are in various technologies from day one, 0:00:50.159 --> 0:00:55.120 that have been there since the technology came out, and 0:00:55.280 --> 0:01:02.959 potentially are actively being exploited by hackers. And the xosed 0:01:03.000 --> 0:01:08.319 chip sets are arm based system on a chip chip sets, 0:01:08.520 --> 0:01:11.480 and they serve as the brains of stuff like certain 0:01:11.560 --> 0:01:15.760 smartphones and wearables and even some car systems. The team 0:01:15.760 --> 0:01:19.880 has been uncovering these vulnerabilities since late twenty twenty two, 0:01:20.319 --> 0:01:23.399 so it's not like this is brand new discoveries in 0:01:23.400 --> 0:01:26.640 all cases, but it has been ongoing and at least 0:01:26.680 --> 0:01:31.960 four of them are considered severe vulnerabilities, potentially allowing an 0:01:31.959 --> 0:01:35.880 attacker to target any affected device as long as that 0:01:35.959 --> 0:01:40.480 attacker knows the associated phone number with the XANOS modem 0:01:40.520 --> 0:01:43.600 in that device. So if someone happened to own a 0:01:43.680 --> 0:01:48.000 smartphone with one of these chips as its processor, an 0:01:48.040 --> 0:01:53.400 attacker could potentially perform a remote code execution or RCE 0:01:53.800 --> 0:01:56.880 on that phone just by knowing that phone number. So 0:01:56.960 --> 0:01:59.480 you wouldn't have to even be tricked into doing anything, right. 0:01:59.520 --> 0:02:04.000 A lot of vulnerabilities we hear about involve someone taking 0:02:04.000 --> 0:02:06.760 an action that they shouldn't have. They get a prompt, 0:02:06.960 --> 0:02:10.800 they respond to it, and that in turn creates the 0:02:10.880 --> 0:02:15.200 opportunity for an attacker to compromise a device. In this case, 0:02:15.560 --> 0:02:17.720 it literally just means that the attacker has to have 0:02:17.760 --> 0:02:21.320 the phone number of the target device. This is a 0:02:21.520 --> 0:02:25.440 massive security flaw obviously, because then the attacker could force 0:02:25.520 --> 0:02:28.919 the victim's phone to run all sorts of different types 0:02:28.960 --> 0:02:34.400 of code. They could have the phone execute code that 0:02:34.480 --> 0:02:39.600 turns it into an espionage device similar to the Pegasus 0:02:39.639 --> 0:02:42.320 stuff we talked about with iPhones a couple of years ago, 0:02:43.320 --> 0:02:46.760 or maybe they could lock the phone in a sort 0:02:46.800 --> 0:02:50.680 of ransomware attack. Now, the other fourteen flaws, because there 0:02:50.680 --> 0:02:53.440 were eighteen total, four of them were super extreme. The 0:02:53.480 --> 0:02:56.200 other fourteen are also bad news, but they don't pose 0:02:56.320 --> 0:03:01.440 the same critical risk as the four that I'm alluding to. Now, 0:03:01.520 --> 0:03:06.160 patches are going out in security updates, but patches are 0:03:06.200 --> 0:03:12.000 things that different companies roll out at different scales and 0:03:12.040 --> 0:03:18.080 at different timetables, so not all of the effective devices 0:03:18.120 --> 0:03:23.040 are necessarily able to be patched right now. That's a 0:03:23.120 --> 0:03:26.640 real serious issue that you know, companies like Google have 0:03:26.760 --> 0:03:30.280 rolled out patches and security updates already, so as long 0:03:30.320 --> 0:03:32.639 as you're up to day in your security patches, you 0:03:32.680 --> 0:03:39.320 have presumably blocked off this potential attack vector. But the 0:03:39.360 --> 0:03:42.880 proposed workaround for those who are still waiting to get 0:03:42.880 --> 0:03:46.640 a patch is pretty extreme. It essentially involves shutting down 0:03:46.680 --> 0:03:50.720 both the call over Wi Fi and voice over LTE 0:03:50.760 --> 0:03:54.840 functions of your phone, So essentially it means removing the 0:03:54.880 --> 0:03:57.640 phone from your phone, and I get it a lot 0:03:57.640 --> 0:04:00.120 of y'all never use a phone as a phone anyway, 0:04:00.600 --> 0:04:03.720 right Like a voice call is probably the last thing 0:04:03.720 --> 0:04:06.320 you would use your phone for. But it does strike 0:04:06.360 --> 0:04:09.440 me as particularly weird to remove the phone part from 0:04:09.440 --> 0:04:12.960 a phone anyway. Some of the devices affected include the 0:04:13.040 --> 0:04:16.640 Pixel six and Pixel seven from Google, a whole bunch 0:04:16.680 --> 0:04:20.400 of different Samsung Mobile devices are affected, which isn't a surprise. 0:04:20.480 --> 0:04:23.800 They include the S twenty two, the M twelve, the 0:04:23.960 --> 0:04:26.880 M thirteen, the M thirty three, and a whole bunch 0:04:26.880 --> 0:04:30.000 of devices that are in the A series of Samsung 0:04:30.040 --> 0:04:33.159 Mobile devices, not all a series, but a lot of them. 0:04:33.520 --> 0:04:37.839 Plus any vehicles that have an Xinos Auto T five 0:04:38.000 --> 0:04:41.240 one two three chip set or any wearables that have 0:04:41.320 --> 0:04:45.600 the Xinos W nine twenty chip set are affected by 0:04:45.640 --> 0:04:49.360 this vulnerability, which is a big old yikes. By the way, 0:04:49.400 --> 0:04:52.919 this is a good reminder to activate real security updates 0:04:52.960 --> 0:04:57.159 when you get them, because security updates can address critical 0:04:57.240 --> 0:05:00.240 vulnerabilities like this one and prevent you from the coming 0:05:00.279 --> 0:05:03.039 a victim in the future. The only thing is that 0:05:03.080 --> 0:05:06.680 you should verify that a security update you receive is 0:05:07.000 --> 0:05:10.760 real first, so like do just bear minimum research to 0:05:10.839 --> 0:05:14.200 make sure that yes, you know, whatever company has issued 0:05:14.320 --> 0:05:18.120 a security update. So if you are using say a 0:05:18.160 --> 0:05:20.919 Google Pixel device, just make sure that Google did in 0:05:21.000 --> 0:05:24.000 fact issue a security update. Just do a quick search 0:05:24.040 --> 0:05:26.440 to make sure that is a thing that has happened, 0:05:26.760 --> 0:05:29.440 and then yeah, install that as soon as you can 0:05:29.680 --> 0:05:35.960 because it can potentially cut off attack vectors. Sometimes these 0:05:36.080 --> 0:05:39.600 updates can also affect other things on a device, which 0:05:40.279 --> 0:05:43.920 is a pain in the butt, but it's still better 0:05:44.040 --> 0:05:48.080 than having your phone compromised. So you know, just a 0:05:48.080 --> 0:05:52.400 word out there. Google has recently banned a Chinese based 0:05:52.440 --> 0:05:55.760 app from the Google Play Store. It is not TikTok, 0:05:56.120 --> 0:05:59.719 though we will talk about the besieged video based social 0:06:00.040 --> 0:06:02.960 work a little later in this episode. No, it's Pen 0:06:03.160 --> 0:06:08.279 Duo Duo. Now. Essentially, this app matches consumers with merchants 0:06:08.760 --> 0:06:12.400 for various products. It started out as an agricultural thing 0:06:12.600 --> 0:06:15.920 and went on to stuff like groceries, and now it's 0:06:15.920 --> 0:06:20.040 pretty much everything. So why is Google removing the app? Well, 0:06:20.080 --> 0:06:22.559 according to Google, the most recent version of the app 0:06:22.600 --> 0:06:27.000 contains malware of some sort, so Google suspended the app 0:06:27.120 --> 0:06:30.599 out of a concern for security. Reuters reached out to 0:06:30.680 --> 0:06:33.440 Penn Duo Duo and received a response that claims that 0:06:33.480 --> 0:06:36.760 Google has alerted the company that the app has been suspended, 0:06:36.800 --> 0:06:39.479 but that Google did not give any details as to 0:06:39.640 --> 0:06:43.000 why the app was suspended anyway. This isn't the first 0:06:43.000 --> 0:06:45.960 time that Penn Duo Duo has been in legal hot 0:06:46.000 --> 0:06:48.760 water here in the United States. One of the major 0:06:48.920 --> 0:06:53.120 issues that is often connected to it involves the counterfeiting 0:06:53.200 --> 0:06:56.960 of various goods. So essentially, it's saying that pen Duo 0:06:57.080 --> 0:07:01.000 Duo tends to match folks up with merchants who sell 0:07:01.080 --> 0:07:06.039 knockoff versions of designer or luxury goods. That is a 0:07:06.120 --> 0:07:09.800 big part of China's business as well, is just creating 0:07:09.880 --> 0:07:15.560 counterfeit goods that mimic high end designer goods. It's a 0:07:15.600 --> 0:07:19.160 big problem in the designer good industry. I can't pretend 0:07:19.160 --> 0:07:24.640 like I fully understand it because I'm basic. I don't 0:07:24.680 --> 0:07:27.840 do like the whole designer goods thing because I have 0:07:27.880 --> 0:07:32.120 no class. Routers also reports that Google reps have claimed 0:07:32.120 --> 0:07:35.680 that it was just an accident that certain communication records 0:07:35.720 --> 0:07:38.640 that could have been used as evidence against it in 0:07:38.680 --> 0:07:42.160 an upcoming trial in which the US Justice Department is 0:07:42.200 --> 0:07:46.880 bringing antitrust charges against Google is scheduled to happen. So essentially, 0:07:47.200 --> 0:07:50.920 the Department of Justice is saying, hey, there's all these 0:07:50.960 --> 0:07:54.360 records that we potentially should have had access to, and 0:07:54.440 --> 0:07:58.440 you're deleting them. It's essentially like your shredding documents because 0:07:58.480 --> 0:08:01.720 you anticipate a raid. And Google is saying, no, we 0:08:01.760 --> 0:08:05.280 didn't mean to delete anything. We were actually taking reasonable 0:08:05.320 --> 0:08:09.320 steps to preserve everything, but some stuff slipped through the cracks, 0:08:09.880 --> 0:08:12.440 not the heart of the lawsuit or claims from most 0:08:12.480 --> 0:08:15.120 of the states here in the US that Google has 0:08:15.240 --> 0:08:18.640 been using anti competitive strategies to prevent any other company 0:08:18.680 --> 0:08:21.800 from getting a foothold in the search business. The d 0:08:21.960 --> 0:08:25.400 o J suspects that Google has denied access to or 0:08:25.480 --> 0:08:30.200 deleted documents that would lend credence to their case against 0:08:30.240 --> 0:08:33.520 the company, and Google says, ah, that's ridiculous. We've already 0:08:33.520 --> 0:08:36.480 shared millions of documents with you. Now. I don't have 0:08:36.520 --> 0:08:39.599 eyes on this one. I have no idea if Google's 0:08:39.640 --> 0:08:43.400 claims are with or without merit. I don't know if Google, 0:08:44.080 --> 0:08:49.520 you know, intentionally deleted files that would have been damaging 0:08:49.559 --> 0:08:53.200 to the company's case, or if this truly was just 0:08:53.720 --> 0:08:57.839 something that happened without any malice involved. I will say 0:08:57.840 --> 0:09:00.920 it's not a good look. I mean, anytime documents go missing, 0:09:01.800 --> 0:09:06.320 that isn't exactly a ringing endorsement of your claims of 0:09:06.360 --> 0:09:09.920 being innocent. But yeah, I don't know. Well way or 0:09:09.920 --> 0:09:14.079 the other. Tech Crunch has a piece titled Facebook political 0:09:14.200 --> 0:09:18.760 micro targeting at center of GDPR complaints in Germany. So 0:09:19.040 --> 0:09:21.960 GDPR if you are not in the know. Is a 0:09:22.000 --> 0:09:26.200 set of regulations that are centered around protecting EU citizen 0:09:26.320 --> 0:09:30.560 personal information. And if you want to do business in 0:09:30.600 --> 0:09:33.280 the EU, if you're a digital company, then you want 0:09:33.280 --> 0:09:37.559 to have any sort of business practices within the European Union. 0:09:38.200 --> 0:09:41.520 Complying with GDPR is a really big deal. In fact, 0:09:41.520 --> 0:09:45.839 their entire consulting companies out there designed to help other 0:09:45.920 --> 0:09:49.679 companies make sure that their business practices are in alignment 0:09:49.720 --> 0:09:53.840 with GDPR, because if you don't comply with it, then 0:09:53.960 --> 0:09:58.160 your company can get hit with massive fines or even 0:09:58.240 --> 0:10:03.120 being denied access to doing business within the European Union, 0:10:03.559 --> 0:10:06.880 which would be a huge deal. Right. So in this case, 0:10:07.360 --> 0:10:11.280 what has happened is that Facebook has used microtargeting to 0:10:11.440 --> 0:10:16.000 direct political advertisements to German citizens without first getting their 0:10:16.040 --> 0:10:21.000 express consent to make such use of their personal data. 0:10:21.040 --> 0:10:24.200 And the EU assigns different levels of importance to different 0:10:24.320 --> 0:10:28.280 kinds of personal information, so things like your political leanings, 0:10:28.679 --> 0:10:32.559 that's way way way up there. That's highly protected information. 0:10:33.080 --> 0:10:35.600 So there is a high bar to clear if you 0:10:35.640 --> 0:10:39.120 are going to make use of that personal information in 0:10:39.160 --> 0:10:44.040 any way. And according to the charges meta Slash, Facebook 0:10:44.320 --> 0:10:47.199 totally failed to do that like they should have gotten 0:10:47.200 --> 0:10:51.640 the express consent of each and every person who's information 0:10:51.679 --> 0:10:56.719 they used to sell microtargeting ads. You know that they did, 0:10:56.760 --> 0:11:00.679 and they didn't do that. And arguably just as bad, 0:11:00.720 --> 0:11:04.360 if not worse, is that every single political party in 0:11:04.520 --> 0:11:10.559 Germany took advantage of meta's disregard for the GDPR regulations. 0:11:11.240 --> 0:11:15.120 So whether it was a far right political party, far left, 0:11:15.160 --> 0:11:19.720 and everything in between, all of them used microtargeting to 0:11:19.760 --> 0:11:25.719 deliver advertising messages, including messages that contradicted each other from 0:11:25.720 --> 0:11:29.880 the same party, just based on microtargeting of individuals. So, 0:11:30.000 --> 0:11:34.199 for example, some one party might have sent out a 0:11:34.240 --> 0:11:38.640 message to someone who is identified as being very pro 0:11:39.559 --> 0:11:45.120 climate change regulations as saying, we're the party that is 0:11:45.160 --> 0:11:50.280 pursuing climate protection over everything else, so believe in us. 0:11:50.760 --> 0:11:54.000 But that same party might have identified someone else as 0:11:54.040 --> 0:12:00.679 being more in tune with personal freedom over time regulation, 0:12:00.720 --> 0:12:03.319 as saying, we believe in the climate just like you do, 0:12:03.559 --> 0:12:05.960 but we also believe you should be able to travel 0:12:06.000 --> 0:12:09.320 wherever you want, whenever you want, so we're about protecting 0:12:09.360 --> 0:12:13.360 the climate without infringing on personal liberty. These two messages 0:12:13.840 --> 0:12:17.120 cannot coexist from the same party, right, you cannot stay 0:12:17.120 --> 0:12:20.120 true to both simultaneously. If you're telling one person we're 0:12:20.160 --> 0:12:23.080 going to protect the climate at all costs, and you're 0:12:23.080 --> 0:12:25.520 telling someone else will protect the climate, but we're not 0:12:25.600 --> 0:12:28.320 gonna you know, we're not going to cramp your style. 0:12:28.760 --> 0:12:32.680 Those are two messages that cannot be true at the 0:12:32.720 --> 0:12:36.559 same time. And the point is that if the same 0:12:36.600 --> 0:12:41.760 party is delivering these two contradictory messages, they are lying 0:12:41.840 --> 0:12:45.680 to someone or they're lying to everyone, at least one person, 0:12:45.720 --> 0:12:49.840 but possibly everybody, And ultimately it undermines the democratic process 0:12:49.880 --> 0:12:53.040 because you cannot trust the messages coming out of a 0:12:53.040 --> 0:12:57.440 political party. If those messages are contradictory of one another, 0:12:58.080 --> 0:13:03.480 it destroys democracy itself. Fun times anyway, an advocacy group 0:13:03.840 --> 0:13:07.600 in OYB has really cranked up the pressure here, and 0:13:07.920 --> 0:13:10.200 I would argue for a very good reason, and that 0:13:10.520 --> 0:13:14.760 in OYB is putting pressure on EU officials to enforce 0:13:14.840 --> 0:13:18.480 the rules under GDPR like they've created these rules, and 0:13:18.559 --> 0:13:21.800 in OYB is saying clearly these rules are not being followed, 0:13:21.840 --> 0:13:24.560 so it's your job to get in here and to 0:13:25.200 --> 0:13:29.320 actually enforce them and to get things like explicit consent 0:13:29.760 --> 0:13:32.040 before any of this kind of data can be used 0:13:32.080 --> 0:13:36.080 for microtargeting advertising as well as you know, putting some 0:13:36.080 --> 0:13:40.160 pressure on the political party system within Germany to obey 0:13:40.280 --> 0:13:44.720 the laws that these political parties are actually, you know, 0:13:44.800 --> 0:13:48.200 supposed to uphold in the first place. So yeah, pretty 0:13:48.440 --> 0:13:52.200 big expose when you look at it in that term, 0:13:52.679 --> 0:13:56.400 and I'm very curious where it goes from here. Okay, 0:13:56.520 --> 0:13:59.480 we've got some more news to cover before we get 0:13:59.480 --> 0:14:12.280 to that. Let's take a quick break. Russia is gearing 0:14:12.360 --> 0:14:16.160 up for its own presidential election that will happen in 0:14:16.200 --> 0:14:20.080 twenty twenty four, and now election officials in Russia have 0:14:20.400 --> 0:14:25.560 a new directive ditch the iPhone. So Reuters has reported 0:14:25.560 --> 0:14:28.280 that the Kremlin has ordered all officials connected with the 0:14:28.320 --> 0:14:32.320 twenty twenty four Russian presidential election, which I am sure 0:14:32.360 --> 0:14:36.240 will be absolutely straightforward and fair, we'll have to stop 0:14:36.440 --> 0:14:40.680 using Apple iPhones, reportedly due to concern that Western intelligence 0:14:40.720 --> 0:14:44.720 agencies could be working with Apple, or maybe even without 0:14:44.720 --> 0:14:49.960 Apple's knowledge and participation, to leverage iPhones in order to 0:14:50.000 --> 0:14:54.680 gather intelligence. I mean, this could sound familiar because we 0:14:54.760 --> 0:14:57.800 certainly have worried about Russia doing the exact same sort 0:14:57.840 --> 0:15:01.200 of thing, not through the iPhone, but through various means 0:15:01.240 --> 0:15:08.200 to inject misinformation and to affect elections across the United States, 0:15:08.200 --> 0:15:12.040 to undermine the democratic process. So Russian officials are kind 0:15:12.080 --> 0:15:14.920 of worried that the same thing's going to happen in Russia. Now, 0:15:15.320 --> 0:15:17.400 I don't have a lot of faith in the fairness 0:15:17.520 --> 0:15:20.360 of the Russian election system, if I'm being totally honest, 0:15:20.680 --> 0:15:23.800 But I also think that perhaps the officials have a point. 0:15:23.880 --> 0:15:27.320 Year I don't think that Apple would work so closely 0:15:27.320 --> 0:15:30.240 with US authorities on this. I don't think Apple would 0:15:30.280 --> 0:15:35.880 comply and allow their phones to be turned into surveillance devices. 0:15:36.960 --> 0:15:40.160 Apple has a history of resisting US government efforts to 0:15:42.080 --> 0:15:46.000 leverage Apple's abilities to do this kind of thing. Apple 0:15:46.080 --> 0:15:50.040 famously resisted the FBI's efforts to create a backdoor entry 0:15:50.080 --> 0:15:54.360 point for Apple's systems, saying that no, that compromises user 0:15:54.440 --> 0:16:00.000 security and privacy, and it compromises our reputation with our customers. 0:16:00.600 --> 0:16:03.960 So I don't think Apple would necessarily be on board 0:16:04.240 --> 0:16:11.160 for any official attempt to surveil a foreign election process. However, 0:16:11.680 --> 0:16:16.920 we have seen companies, surveillance companies creating apps that exploit 0:16:17.000 --> 0:16:20.200 vulnerabilities and iOS devices in the past. You know, we 0:16:20.240 --> 0:16:24.240 saw that with an Israeli company creating the Pegasus app 0:16:24.640 --> 0:16:27.600 that would turn an iPhone into a spy device just 0:16:27.640 --> 0:16:32.680 by sending a message through I Message. Apple has since 0:16:32.800 --> 0:16:36.280 patched out that vulnerability, But the point is we have 0:16:36.400 --> 0:16:40.400 seen that in the past, and really anytime you're talking 0:16:40.440 --> 0:16:43.560 about relying upon a device that's going to send data 0:16:43.920 --> 0:16:47.160 out of the country you're in to another country that 0:16:47.280 --> 0:16:50.880 might be, if not outright hostile, at least wary of 0:16:50.920 --> 0:16:53.760 the one you live in, there are concerns, right because 0:16:54.000 --> 0:16:57.120 if you don't have full control of the information chain, 0:16:57.800 --> 0:17:01.120 then there are worries about how that could hack national security. 0:17:01.200 --> 0:17:03.600 So while I don't think everything's on the up and 0:17:03.640 --> 0:17:07.080 up and Russia necessarily, I do also understand the concern 0:17:07.640 --> 0:17:13.679 about officials relying upon technology that is created in a 0:17:13.720 --> 0:17:18.119 country that at least is wary of Russia, if not 0:17:18.160 --> 0:17:23.520 outright hostile at times. So yeah, I get the security concern. 0:17:23.600 --> 0:17:25.680 It's not really a surprise. I don't even think it's 0:17:25.680 --> 0:17:28.439 really an indictment against Apple, because again, Apple does not 0:17:28.560 --> 0:17:33.240 have a reputation of working alongside US government to that degree. 0:17:33.760 --> 0:17:37.720 But allegedly all election officials will need to be free 0:17:37.760 --> 0:17:43.280 of iPhones by April first, Stanford has introduced Alpaca AI, 0:17:43.480 --> 0:17:46.720 which according to researchers, performs on a level similar to 0:17:47.160 --> 0:17:52.720 GPT three point five that's the previous generation of open 0:17:52.840 --> 0:17:57.240 aiyes language model. But the surprising thing is that these 0:17:57.280 --> 0:18:02.200 researchers spend about six hundred dollars developing their Alpaca AI. 0:18:02.840 --> 0:18:06.040 You know, open ai has had billions of dollars of 0:18:06.119 --> 0:18:10.080 investments poured into it, and apparently these Stanford researchers were 0:18:10.119 --> 0:18:13.920 able to achieve a similar outcome to the previous generation 0:18:13.960 --> 0:18:17.880 of GPT for just six hundred bucks. So they started 0:18:17.920 --> 0:18:22.760 with Meta's open source LAMA seven B language model. So 0:18:23.000 --> 0:18:27.040 this is Meta's own large language model. It is not 0:18:27.119 --> 0:18:30.640 the same one that open ai uses, and Meta allows 0:18:30.680 --> 0:18:35.320 academics to access these models for a price. The seven 0:18:35.440 --> 0:18:38.760 B model is the cheapest and smallest of the large 0:18:38.840 --> 0:18:41.399 language models, and that's the one the researchers went with, 0:18:41.840 --> 0:18:45.240 so they saved some money by going with that. Then 0:18:45.280 --> 0:18:50.560 they used open AI's GPT to generate sample conversations around 0:18:50.560 --> 0:18:56.960 fifty two thousand sample conversations to train to post train 0:18:57.080 --> 0:19:01.200 their own AI model. So they made use of open 0:19:01.240 --> 0:19:07.160 aiyes official GPT research tools to do that that cost 0:19:07.200 --> 0:19:10.199 a few hundred bucks. Then they made some tweaks and 0:19:10.240 --> 0:19:13.800 some adjustments, and they ended up with an AI language 0:19:13.800 --> 0:19:17.679 model that can perform really well against GPT itself, like 0:19:17.720 --> 0:19:21.640 almost at the same level. I think in something like 0:19:23.000 --> 0:19:28.120 almost one hundred and eighty tests, ninety showed their AI 0:19:28.160 --> 0:19:33.159 model outperforming GPT, and an eighty nine cases GPT outperformed 0:19:33.160 --> 0:19:36.359 the AI model, so they're essentially neck and neck. As 0:19:36.600 --> 0:19:39.760 New Atlas the website New Atlas points out, the research 0:19:39.800 --> 0:19:43.159 shows that if you have the right tools at your disposal, 0:19:44.080 --> 0:19:47.119 you can potentially create your own large language model for 0:19:47.200 --> 0:19:50.320 AI and then build stuff on top of that, like 0:19:50.840 --> 0:19:55.600 chatbots and whatnot. And while Meta limits access to LAMA 0:19:55.680 --> 0:20:00.080 to academics and such, the problem is someone leaked a 0:20:00.200 --> 0:20:05.479 code online. So while you officially have to be an 0:20:05.480 --> 0:20:08.880 academic to get access to LAMA, if you're the unethical type, 0:20:08.880 --> 0:20:12.200 you can just grab the code online. It is available 0:20:12.240 --> 0:20:15.440 out there, and you can start to develop your own 0:20:15.600 --> 0:20:18.280 language model and your own work based on top of that. 0:20:19.440 --> 0:20:21.760 This means we could be at the beginning of an 0:20:21.800 --> 0:20:27.679 era of chat bought explosions and AI activations, all powered 0:20:28.160 --> 0:20:34.919 by similar but independent language models. It's entirely possible that 0:20:35.040 --> 0:20:38.440 some or even many of these will lack the restrictions 0:20:39.040 --> 0:20:41.840 that we see open aie putting on top of its 0:20:41.880 --> 0:20:45.679 own products like chat GPT. You know, open ai works 0:20:45.760 --> 0:20:51.119 hard to try and limit things like you know, racist 0:20:51.240 --> 0:20:58.440 or misogynyistic or transphobic language from being generated through its tools, 0:20:58.920 --> 0:21:03.280 and you know, obviously we've also seen issues with the 0:21:03.400 --> 0:21:08.240 reliability of information generated through tools like chat GPT. So 0:21:08.280 --> 0:21:11.600 the fear is that these new tools, which could be 0:21:11.960 --> 0:21:15.880 pretty easy to make yourself based upon this research project, 0:21:17.119 --> 0:21:21.440 won't have those safeguards in place, and that we could 0:21:21.440 --> 0:21:25.080 see a lot more misinformation or abuse of such systems, 0:21:25.440 --> 0:21:28.800 maybe using systems to try and trick people with phishing attacks, 0:21:28.840 --> 0:21:33.200 all sorts of different tactics that could be harmful could 0:21:33.800 --> 0:21:36.399 be coming down the road. Because we've already seen the 0:21:36.440 --> 0:21:39.400 barrier to entry, it's not that high, and if you're 0:21:39.440 --> 0:21:43.200 able to figure out a way to exploit that and 0:21:43.280 --> 0:21:45.879 make a whole lot of money, maybe you'd be willing 0:21:46.000 --> 0:21:50.240 to take that unethical route and create a whole new 0:21:50.280 --> 0:21:54.199 generation of victims out there. Oh brave new world to 0:21:54.359 --> 0:21:57.119 have such chat bots in it. And on the topic 0:21:57.119 --> 0:22:02.080 of chat GPT recently above started showing users the title 0:22:02.200 --> 0:22:06.600 and short description of chat GPT sessions that were not 0:22:06.760 --> 0:22:08.840 their own. You would look at your history and start 0:22:08.920 --> 0:22:12.479 seeing stuff that you definitely did not do. So users 0:22:12.480 --> 0:22:15.160 were having the titles and short descriptions of their chat 0:22:15.200 --> 0:22:18.199 sessions shared with other users without their consent, and this 0:22:18.359 --> 0:22:21.560 was not by design. This is a good time to 0:22:21.600 --> 0:22:26.600 remind everyone that if you are using chat GPT, you 0:22:26.640 --> 0:22:31.440 should not share sensitive information in chat GPT conversations because 0:22:31.480 --> 0:22:35.880 you never know when those conversations could become public. So, 0:22:35.920 --> 0:22:38.360 if you were using chat GPT to I don't know, 0:22:38.880 --> 0:22:43.040 maybe draft a resignation letter, but maybe you aren't seriously 0:22:43.040 --> 0:22:45.200 considering leaving your job, you were just kind of toying 0:22:45.200 --> 0:22:47.960 with the idea, and you have chat GPT draft a 0:22:48.040 --> 0:22:51.640 resignation letter for your behalf. Well, that could end up 0:22:51.640 --> 0:22:55.000 being awfully awkward. If your chat session title and description 0:22:55.400 --> 0:22:58.800 then gets shared with somebody else but without your permission, 0:22:59.280 --> 0:23:04.280 and now people have the impression that you are leaving 0:23:04.280 --> 0:23:07.719 your job, that could create all sorts of complications in 0:23:07.720 --> 0:23:09.800 your life that you probably don't want to have to 0:23:09.800 --> 0:23:13.440 deal with. That's just one example. What if you were drafting. 0:23:13.480 --> 0:23:16.000 I don't know, a love letter to your crush. I 0:23:16.040 --> 0:23:18.440 mean like, there's all sorts of things that you could 0:23:18.440 --> 0:23:21.720 have been using chat GPT to do, even just for fun, 0:23:22.440 --> 0:23:25.919 that you probably don't want people to see. So in 0:23:25.960 --> 0:23:28.280 my case, if people were to see my history with 0:23:28.400 --> 0:23:31.880 chat GPT, they would probably see how I was trying 0:23:31.880 --> 0:23:35.520 to get chat gpt to compose a Shakespearean sonnet about AI. 0:23:36.240 --> 0:23:39.600 Or there was one time when I asked it to 0:23:39.640 --> 0:23:42.080 create an episode of tech stuff. I was just curious 0:23:42.119 --> 0:23:44.639 what would happen, and I told it to create an 0:23:44.640 --> 0:23:47.680 episode of tech stuff about chat GPT, which I had 0:23:47.880 --> 0:23:50.760 already done an episode about chat GPT. So it's not 0:23:50.800 --> 0:23:52.960 like I was trying to get out of doing my work, y'all. 0:23:53.000 --> 0:23:54.840 That's not what this was about. I was just curious 0:23:54.840 --> 0:23:58.240 what would happen, Like would it actually sound like one 0:23:58.240 --> 0:24:00.480 of my episodes? The answer is no, it did not 0:24:00.600 --> 0:24:05.040 sound anything like one of my episodes. It created an 0:24:05.080 --> 0:24:09.760 interview with a fake chat GPT representative. So if I 0:24:09.760 --> 0:24:12.199 had actually recorded the episode, I would have had to 0:24:12.280 --> 0:24:16.560 invent a person who supposedly worked at open AI who 0:24:16.560 --> 0:24:19.600 would be talking about chat GPT in an authoritative way 0:24:19.840 --> 0:24:24.480 which seems unethical at the very slightest, but anyway, Open 0:24:24.520 --> 0:24:28.880 Ai shut down chat GPT for a short while yesterday 0:24:28.880 --> 0:24:33.720 evening and they took the chat history offline for a while. 0:24:33.760 --> 0:24:36.840 They are reportedly starting to roll that back into service, 0:24:36.920 --> 0:24:40.080 but it might be a while before all users are 0:24:40.119 --> 0:24:44.280 actually able to access their chat histories, so there may 0:24:44.320 --> 0:24:46.960 be a while before you're able to see back on, 0:24:47.560 --> 0:24:49.399 you know, the stuff you have been talking about with 0:24:49.480 --> 0:24:53.160 chat GPT in the past. The verges. Tom Warren has 0:24:53.160 --> 0:24:58.159 a great article titled Microsoft's new Copilot will change Office 0:24:58.200 --> 0:25:02.119 documents Forever. This kind of again falls in line with 0:25:02.200 --> 0:25:07.200 AI and chatbots. So Tom Warren was in a Microsoft 0:25:07.200 --> 0:25:11.399 Teams meeting with Microsoft reps to talk about Copilot. That's 0:25:11.520 --> 0:25:17.200 Microsoft's AI feature that's going to be incorporated into all 0:25:17.400 --> 0:25:21.600 office products moving forward. During the meeting, the reps turned 0:25:21.680 --> 0:25:25.480 Copilot on to do stuff like take notes about the 0:25:25.840 --> 0:25:28.919 Microsoft Teams meeting. At the conclusion of the meeting, Warren 0:25:28.960 --> 0:25:31.600 actually received these notes, which included a recap of all 0:25:31.600 --> 0:25:34.560 the questions he had asked during the interview, and it 0:25:34.600 --> 0:25:37.800 sounds like Copilot can help users take advantage of office 0:25:37.840 --> 0:25:42.080 products as if the user was an elite user. You 0:25:42.200 --> 0:25:44.440 probably maybe you are an elite user in one or 0:25:44.480 --> 0:25:47.040 two office products, or maybe all of them, or maybe 0:25:47.040 --> 0:25:49.720 you know someone who is, like someone who just knows 0:25:49.800 --> 0:25:54.840 all the different tricks to create the most snazzy and 0:25:54.960 --> 0:26:00.359 effective presentation, or they're able to create really help full 0:26:00.760 --> 0:26:03.919 graphs and charts out of Excel sheets, like there's an 0:26:04.040 --> 0:26:06.119 art to it. You know, you can learn the basic 0:26:06.200 --> 0:26:10.040 steps and do a fine job, but to do something 0:26:10.119 --> 0:26:14.280 really effective requires a little more work. And you know, 0:26:14.520 --> 0:26:17.920 it sounds to me like Copilot can actually help you 0:26:17.960 --> 0:26:20.400 achieve those things and not only help you do them, 0:26:20.400 --> 0:26:24.160 but also teach you how to do them yourself. So 0:26:24.200 --> 0:26:27.920 it's not just like a magic wand that makes things better, 0:26:28.359 --> 0:26:31.520 but a tutorial tool that can help you learn how 0:26:31.560 --> 0:26:34.080 to take advantage of features that maybe you never even 0:26:34.160 --> 0:26:38.880 knew existed. It's kind of like the old Microsoft Clippy, 0:26:38.920 --> 0:26:43.440