Tech News: Google Bard Hits a Wrong Note

the big news throughout the beginning of this year. Now, as I mentioned earlier this week, both Microsoft and Google have launched kind of an invitation only test of their respective AI chat bought tools. Microsoft is taking advantage of chat GPT, that's the chat bought everyone has heard about. But there are some inferences with the Microsoft version. Standard chat GPT cannot give up to date responses. You wouldn't be able to get information about, say the recent earthquake

in Turkey, for example. The information that chat GPT the Vanilla brand draws from is a couple of years old. However, the Microsoft version is able to reference current events, so you could ask about stuff that's happening in the news. Now that's potentially good being able to get us accinct summary of what's going on could be a really valuable tool,

but it can also spell trouble for news organizations. You know, if they're not getting traffic because people are just getting the news from an AI chat butt, that could mean that they start seeing less and less revenue and then ultimately they could go out of business. Ironically, these are the institutions that are creating the data that the chat

bots able to pull from in the first place. However, it could also mean that summary could end up being very biased depending upon which news sources the chat bought is pulling from when it's formulating a response. If it's a far left or far right leaning news source, then you're probably not going to get an objective account of what's going on in the news, and so that could

be a really big problem down the line. Now I'm on waitlist for access to the Microsoft version, which, by the way, you can only access if you're using Microsoft's browser Edge, which is you know, obviously an attempt for Microsoft to try and get some some market share away from Google and to claim some of that for their own by keeping this being search result AI chat bot integration contained to the Edge browser. But anyway, I'm on the wait list. I have not received access to it,

so I don't have any firsthand experience. But you know, I did test out some of the example questions and being so you, as a normal user, you can go to Being through Edge and have some access to this, but only by submitting sample questions that have already been written. So I did that and what I saw was that the results came back where you have search results in the main part of the screen on the left hand side, and then the right hand column you have the AI

generated response, so they're side by side. I think that's a wise decision. It means that the search results are not outright eliminated. But whether or not folks will go beyond AI generated responses and start clicking on links remains to be seen. Anyway, Microsoft impressed a lot of folks with this announcement and the early access. Now let's contrast that with Google, which has had a much rougher week now.

As I mentioned earlier, Google had sort of rushed its own AI chat bot, which is now called Barred, into an invitation only beta test earlier this week, in anticipation of Microsoft's announcement of chat GPT. Then yesterday Google promoted the barred chat butt, but in that outing, the AI

made a boo boo. So when it was asked to list things that the James Webb Space Telescope had managed to do so far, the barred AI chat butt included a bullet point saying that the James Webb Space Telescope took the very first picture of an exo planet, so a planet that's outside of our own solar system. The problem is that's wrong. The very first image of an exo planet pre dates the James Webb Space Telescope. It

happened almost twenty years ago. But Bard presented the bullet item as if it were a fact, something that chat GPT critics have said is a huge problem with chat GPT as well. In fact, for all these AI applications, that's the problem is that they present information in such way that it seems authoritative and definitive, when in fact

it may be incorrect. Well, Google stock went tumbling, had dropped from around a hundred six dollars per share down to ninety eight dollars per share, and you might think, wow, okay, ten dollars, that's big, but it's not. You know, how much could that be collectively? Well, collectively, if you measure it across all the shares of Google that are out there, it means that the company effectively lost a hundred billion dollars in value Yawza, Microsoft Scores and early victory in

the Chat Bought search wars. Though I would caution anyone in Redwood from celebrating just yet, as chat gpt is very much capable of giving wrong answers with conviction, just as Barred has done. It is better to continue to stress that these tools are in development, they are not trustworthy by a long shot. On a related note, I also saw that Kun'lan Tech, the parent company behind the web browser Opera, plans to incorporate chat gpt into Opera

itself in the future. There are scarce details as to the extent to which chat gpt will play a role in Opera, and I also don't have really any information on a timeline for when this could be rolled out, but it is interesting that Opera, which is a niche web browser, is waiting into this competitive field. Currently, Opera ranks as the sixth most popular browser in the world and has a market share of two point four percent.

Compare that to Google Chrome. This is the dominant web browser in the world and has a sixty five point four percent market share, so more than half of all web browsers being used our Google Chrome. Now Microsoft is hoping to take a serious chunk out of Google because again it's chat GPT bing search tool is only available if you're using the Edge browser. So we'll have to see if this changes things, or if in fact chat

gpt just doesn't have the legs. Vice has a really interesting article about a very odd quirk chat gpt appears to have. The article has the title quote chat gpt can be broken by entering these strange words and nobody is sure why end quote, And yeah, that kind of summarizes the story. But the strange words in particular appear to be user names from places like Reddit and Twitch and that sort of stuff, and they include stuff like

solid gold, magic harp, and streamer bought. Researchers discovered a bank of around one hundred odd words or user names while they were probing chat gpt. They were trying to determine what sort of prompts create the most reliable results, and in the process they kind of uncovered this treasure trove of bizarre words that chat GPT just doesn't seem

to understand what to do with. Now, chat GPT can compose text that will include words that are similar to the one hundred or so odd ones that the researchers found, but they discovered if they put those actual names into a prompt, it would create an odd response from chat GPT. For example, they said that when they submitted streamer bought to chat GPT, it replied, you're a jerk. Now, that's

very funny, but obviously it's also puzzling. What would prompt chat GPT to call you a jerk just for saying streamer bought to it? Other words would either create no response at all, or chat GPT would totally misinterpret whatever it was. It would not be able to repeat these words if you told it to repeat them. Uh. In one case, it converted a user name into a number. It treated the user name as if it were a number, and if you asked questions about the user name, it

would answer as if it were a number. Now, the researchers admit they are not sure what's going on here, As the title of the piece suggests, they suspect that this could be from open AI using sites like Reddit as a data source to crawl through and gather information.

According to one of these researchers, several of the user handles were actually found to be participants in a Reddit thread where people in the thread are counting to infinity reply by reply, like each reply increases the number by one. Maybe that's why when the researchers asked about the Nitrome Fan they discovered the chat GPT interpreted that as the number two, like the Nitrome Fan meant two, So I guess you would have blink the Nitrome Fan as a

band that I don't really like very much. And sure, this might all sound trivial, but the researchers have pointed out that they want to understand why a chat bought might pick certain information over other while creating a response, what kind of information would cause a chat about to break, because that's important to understand and also for us to get a better grip on how accurate, reliable, and trustworthy the AI is at any given time, as well as

get insight into the types of sources that the AI relies upon important questions to answer. Okay, we're gonna take a quick break. When we come back, we're gonna segue away from all this AI stuff and talk about Twitter for a little bit. Okay, we're back. If you were on Twitter yesterday, you might have encountered some weirdness going on. Some users ended up seeing a message that said you are over the daily limit for sending tweets, even though

they hadn't posted anything up to that point. Twitter does in fact have a daily limit on tweets, but it is a very high limit. I'm talking two thousand, four hundred messages. There's also a limit on direct messages per day that's five hundred, and you're also limited to how many people you can follow in a day that's four hundred. These policies are largely an effort to stave off particularly active spam bots. Your average person is not likely to

send that many messages. I mean, a hundred messages per hour on average is asking a bit much, even for me. So it was certainly strange when users started to get message saying they had hit these limits, and it prompted speculation that perhaps Twitter was going to impose strict limits and then you would have to subscribe to a paid Twitter account in order to increase those limits. But honestly,

it just ended up being a glitch. And I know that Twitter has made a lot of decisions that would lead you to think, oh, they're gonna charge you if you want to post more than ten times a day, but that does not appear to be the case as far as anyone outside of Twitter is where there is no plan to do that. So Elon Musk also sent out a message to Twitter employees that urged them to kind of put projects on hold for the moment and instead prioritize making sure that Twitter is stable and that

it's performing properly. He also cited the fact that the Big Game, you know, the American football big game that has a trademark name that I have been told I cannot use, is happening this weekend and people are bound to be on Twitter talking about all the commercials and also the game, because you know, I understand there are people who really take those games very seriously. This is me not really joking about sports fans. I'm more joking about how I'm just a person who never got into sports.

I think sports fans who are passionate about the sports they love, that's awesome. Um, I just I just am not one of them. But it's a big shocking revelation. I knew you weren't expecting it anyway. I'm sure that Twitter staff are going to have their hands full keeping the ship afloat during one of the busiest days for social networks, and that's got to be hard considering the

reduced workforce that still remains at Twitter. The UK and the US have imposed sanctions against leaders in the Russian hacker group called trick Mott, which hasn't been as active recently, but in the past has been responsible for some massive

ransomware attacks on vulnerable targets, most notably healthcare facilities. Hackers will frequently concentrate on healthcare companies because their services are literally matters of life and death, so there's a strong incentive for these targets to concede to ransom demands because the consequences could be deadly. UK and US officials have also alleged that this hacker group has official ties with Russian intelligence, which would mean they are a state sponsored

hacker group. And you might wonder what good sanctions in the UK and the US are going to do against these hackers who are based out of Russia, which is a country that already has heavy sanctions against it for numerous reasons, and according to the officials, the goal is really to make it harder for the hackers to launder money from their ransomware attacks, which would potentially make them easier to catch and harder for Russian officials to ignore.

A couple of electric vehicle startup companies called Rivian and pole Star recently commissioned a climate study from the consulting firm Kearney, and the results of that study prompted the startups to announce that the automotive industry needs some serious changes beyond the electrification of the vehicles they offer in

order to make a real DNT in carbon emissions. And we've been seeing a lot of companies commit to producing more electric vehicles recently, particularly as states and countries are passing laws that are going to require all new vehicles sold within their borders to be something other than traditional

internal combustion engine vehicles. But the startups argue that to really make meaningful change, the industry as a whole has to address the entire supply chains contribution to greenhouse gas emissions. That includes things like the mining of the minerals you need in order to make batteries and further that even with the electrification of vehicles, we're moving far too slowly to meet important goals outlined by the Paris Agreement. The

report is actually pretty grim. Uh. It sounds like it's impossible to meet the Paris goals without a much more drastic change than what countries like the United States have already committed to, and it's not likely that these countries are going to uh increase that commitment. The implications for that are really sobering because it means that we will fail to limit global warming to a further one point five degrees celsius and the impact of climate change will

be greater because of that. And as the startups are stressing, every year that passes with us not doing what is necessary means that what is necessary will become even harder to do. It will grow and become more imperative for us to do something about it. We're really just making it harder for ourselves and more importantly, on future generations. According to this report, The Wall Street Journal reports that users in certain Facebook groups in India have been using

the platform to offer guns for sale. The groups in question are related to an extremist Hindu organization that the c i A alleges is connected with violent attacks against Muslims in India. Now, private gun sales between citizens is a big no no in Facebook's terms of service, and yet, when an Indian activist brought these posts to Meta's attention, he received a response that said the posts actually didn't

violate any of Facebook's policies. However, when the Wall Street Journal started to ask Meta questions about how that could possibly be the case, the posts offering guns for sale were suddenly removed. Meta essentially said, these posts violated Facebook policies, and we take that kind of stuff down when we find out about it. Except they had already found out about it because the activists had reported it. I guess

you just need to insert cricket chirping sound here. Various activist groups have long pointed out that Meta's content moderation practices have a lot to be desired in many regions around the globe, particularly in places where English is not the dominant language. And to make matters even stickier for Meta, India is home to more than a billion people, and a whole lot of them use Meta products like Facebook

and WhatsApp. So the picture of the Wall Street Journal paints is that despite this large user base, Meta really is failing to enforce its policies in India. Finally, some light news. This week, Nintendo held one of its Nintendo Direct events and showed off some upcoming titles, including the highly anticipated and delayed sequel to the legend of Zelda

game Breath of the Wild. The sequel is called Tears of the King, and Nintendo released a trailer to give players a hint of what to expect, and it looks real, moody and epic. Y'all. You get a real Lord of the Rings s ron kind of vibe going with the the villain and uh saying that the entire kingdom of High Rule is going to get wiped out. And we get some shots of Link in action doing cool stuff like writing a quad copter like device and fighting various monsters,

as well as some cut scene clips. And we also got the release date of May twelve of this year, and it looks pretty cool now. I say that as someone who hasn't played a Zelda game since Okarina of Time that was my last one because I'm old. Nintendo also announced a remaster of Metroid Prime that's available on the Shop, the digital store for the Nintendo Switch right now. Nintendo also announced that Switch Online players will be able to access a library of games that were originally designed

for the game Boy and the game Boy Advance. That really takes me back. You could finally play Tetris the way it was meant to be played a k a. The way I played it. Nintendo announced several other titles and releases, and you should definitely pop online when you can and find the full presentation. If you're a big Nintendo fan, I'm sure you'll enjoy it. That's it for the news for Thursday, February nine. I hope you are

all well. Make sure you reach out to me with anything you would like to hear in future episodes of tech Stuff. You do that on Twitter if it's up. The handel for tech Stuff is tech Stuff hs W. I'll talk to you again really soon. Tex Stuff is an I Heart Radio production. For more podcasts from I Heart Radio, visit the i Heart Radio app, Apple Podcasts, or wherever you listen to your favorite shows.