Tech News: AI Might Not Be Ready for Prime Time

or outright replace smartphones. And these products initially generated a decent amount of buzz, but upon actual release, the reception hasn't been universally positive. So one of these was the Rabbit R one and maybe you've seen one of these,

maybe you haven't. It's a little orange do hickey. Let's got some interesting user interface features on it, and it got some folks excited back at CES early this year, but media outlets like The Verge brought reality crashing down on the R one with headlines like Rabbit R one review nothing to see here by David Pierce. Another device that has also seen critical dismissal is the Humane AI pen.

So I thought it would start this episode with a quick compare and contrast of these two products, paired with some commentary about where we are with AI when it comes to incorporating it into consumer goods. So first up the Rabbit R one. So this device wasn't intended to be a replacement for a smartphone, even though a lot of people were talking about it like that, because for one thing, you can't make calls on an R one.

You were supposed to be able to set up your various apps on this device, like to connect to your various accounts, things like ride hailing apps and delivery services and ticket apps, that kind of thing. And then the idea was you could give a complex command to the R one and it would take care of, you know, pretty much all the rest. So you could say something like, in an ideal situation, maybe you tell the R one that you need a car to take you Tier hotel

from the airport. You've flown into a new city. You're using the R one to call a an uber or lift or something to pick you up from the airport, take it Tier a hotel. Also to go ahead and place an order for a pizza place you learned about that's going to be near your hotel, so that can be delivered, so you get that shortly after you've checked in, and that you also want to buy a couple of seats to a sporting event that's happening down the street

from the hotel. But according to reviews like the one in The Verge, the reality was that R one was not nearly so versatile or useful, and it had issues with a lot of stuff, including delivering accurate weather forecasts. You know, sometimes the R one would pin your location to being a place that was like thousands of miles away, and the image recognition that was built into the device

also left a lot to be desired. R one could frequently misidentify stuff that you were pointing it at, and that's not really useful either if you need to know, like, hey, is this safe for me to eat? Or whatever it may be. And while the price tag of one hundred and ninety nine dollars wasn't as crazy as you know, like the cost of an Apple Vision Pro, two hundred bucks is still a lot to pay for a gadget that doesn't work very well. Then you have the humane

aipen that's a wearable smart device. Although smart might be generous a going to hey, hey, look it's David Pierce of the Verge again who David Pierce had a rough like April slash early May reviewing these kinds of devices. But back in April, Pierce said that the Humane aipen was quote not even close end quote to what was promised. So this is a pin you wear it like it's

like a little Lavalier microphone or something. It doesn't have a screen on the device, but it did have or does have a little projection laser, so you can hold your hand out and it will project a little interactive menu on your hand and you can, you know, use touch it's not really touch like, it's it's gesture commands to be able to tell it what you want to do. So if you were playing a music track, you could do that to change the volume, that kind of stuff.

You could also interact with it through voice commands, and it was supposed to lead fore you from the shackles of your smartphone, so that you didn't have to have your head down looking at a screen. You could be keeping your phone in your pocket walking around using this device to do the stuff that you would typically whip your phone out to do, and you'd still have your phone on you, so you could use that if you needed to and the problem is it just didn't work

very well. It cost six hundred ninety nine dollars and has an ongoing twenty four dollars subscription fee so that you can use the cellular network, which is Tea Mobile, by the way, in order for this thing to work. And Pierce was pretty brutal in his review. He essentially said it was a totally broken device. Now, the New York Times reports that Humane has quietly been looking around for a potential buy for the company, and allegedly one

of those entities they've looked at is HP. I recommend that article in the New York Times, by the way, it's titled this is going to be painful How a Bold AI device flopped. Anyway, I think the R one and the Humane aipen both help illustrate why we're not really at a spot where AI is useful enough to be the primary interface for a consumer product just yet. Not saying that AI isn't impressive, not saying that there hasn't been a ton of incredible work done in that field,

but it's not fully baked. And when you're talking about really expensive consumer gadgets that end up not being reliable or useful, that's a problem, right, Like you would kind of want to see it continue to go through development for a while, maybe have some limited beta testing and go that route before releasing a full consumer product that could potentially scuttle an entire company. Now, it's not just

consumer electronics that are struggling with AI, of course. Late last week, Google reportedly eased off a little bit with

its AI Overviews tool. That's the feature that uses AI to provide summaries and answers to queries in a web search, with the idea being that if the AI can answer whatever your question is, then you don't have to do all that pesky clicking through to a reputable source to dig through you know, edited and vetted information, which I mean, I guess I can understand that a bit in some cases,

because goodness knows. I hate the trend that websites will bury the lead of a story three cores of the way down a page in order to serve you more ads, so that you have to scroll past a lot more ads before you get to what it is you're looking for. I see that in a lot of like pop culture news sites, where like the first five paragraphs are like Star Wars originally came out in the nineteen seventies, like I know this, I just want to know what the

next project is or goodness knows recipes. That's been a problem for years, right, recipe sites do this all the time. Anyway. The New York Times tested some searches last Friday and found that overview AI was only popping up about once in every six queries or so. So why is that? Well? The conclusion that Nico Grant of The New York Times arrived at is the answers provided by overview AI are not always reliable or even safe. This is nothing new.

If you've been following news about generative AI for a while, of course, I mean you likely know AI can hallucinate or confabulate, or you just make stuff up. In other words, that's not a good thing, particularly when you're using it to augment search, and in some cases it can lead to huge problems if the AI is recommending stuff that

just isn't safe. Google is continuing to develop the tool, and the company has started to build in some protections around certain categories of queries, such as stuff relating to health matters. I'm honestly surprised at how aggressive companies like open Ai and Google have been when it comes to pushing these AI models out to the public, simply because when things go wrong, they can go spectacularly wrong. But

I guess if you snooze, you lose. Over the years, I've mentioned a lot of threats that AI poses, from potentially taking work away from well just about everyone, to the science fiction disaster scenario of a malevolent AI determined to wipe out humanity. That last one is still well within the realm of science fiction. Thankfully, we're not at that. We haven't been putting AI in charge of potentially devastating weaponry, for example, so hopefully we continue that trend. But one

threat AI definitely poses relates to electricity and resources. AI requires a lot of power to operate, and with more AI companies launching and then the big AI companies growing rapidly, this means that any large entity that's serious about being a definitive AI company has to start to think about where is that juice going to come from. Over at open Ai, that seems to include a company called Helium, which aims to use nuclear fusion to generate electricity. That's

an amazing goal. Nuclear fusion would truly transform our world. But it's a goal that's not currently viable. We've only had a couple of cases where a fusion reaction even produced more electricity than it took to start the reaction in the first place, and when you take a real step back and you consider the whole picture, we're still not really there yet. So forging a deal with Helion seems a tad premature since fusion is not yet a viable method for power generation here on Earth. So why

would open Ai consider making such a deal. Well, one factor that could have something to do with it is that Sam Altman, who is the CEO of open Ai, also happens to be the chairman of Helium's board of directors, and he has nearly four hundred million dollars invested in the company. Now, to be fair to mister Altman, he has said that he has recused himself from these negotiations.

That's a good thing, because it sounds like there's more than just the potential for a conflict of interest here when the head of your very power thirsty company also has hundreds of millions of dollars invested in a startup power company, and then those two companies start to chat with each other, Obviously, eyebrows will naturally rise as a result.

I'm actually kind of reminded of the story of We Work, in which the CEO of that company would end up leasing properties he personally owned to the company he ran as CEO, which seemed shady to me then, and it seems kind of shady to me now. But then, I guess it's all academic at the moment, since Helium is not currently generating electricity through fusion, and open ai reportedly doesn't even have any data centers of its own. Instead

it relies on Microsoft's platforms. But still, it's a thing that makes me go, hmm, Okay, before I make any more references to outdated nineties music, we're gonna take a quick break to thank our sponsors. We're back, so next up. A watchdog group in the European Union is calling on regulators to pass an order that would force Meta to stop harvesting user data for the purposes of training AI

without first seeking user consent. Now, you might have heard me talk about how Meta's policy on Instagram allows the company to train AI models on user data without asking permission first. In fact, you cannot opt out of that if you are anywhere other than the European Union. But the EU is different, right, you know, the EU has these these laws in place that require companies to give

people more options. So in the EU, users would get a little pop up about the matter to opt out of it, and that was only if they actually read the thing and saw the little hypertext link that would let them go to a form where they could opt out. The whole process was complicated and, in my opinion, misleading. Anyway, it seems like Meta has purposefully designed a system to discourage people from opting out. And from Meta's perspective, I

get it because training AI is hard. You need a lot of material to train your AI, and Meta has access to a treasure trove of user generated content, and it sure is a hassle to have to secure permission from each user in order to use all their stuff. A Meta spokesperson responded to the Watchdog by saying that Meta quote complies with privacy laws and our approach is consistent with how other tech companies are developing and improving

their AI experiences in Europe. End quote. Now, I'm not sure that last bit is going to be that much helpful because I suspect watchdog organizations are equally concerned about these other tech companies. I don't think they just have it out for Meta. The watchdog group argues that Meta has failed to comply with EU law that an opt out form is not the same thing as an opt in form, which is absolutely true. It's one thing to skip over opting out. It's another one. You're asking point

blank do you consent to this? Because I need you to indicate you do by clicking this box, and then if you click the box, well then that's on you. Anyway, I'm sure this story will continue to develop, so I'll keep an eye on it. The US Department of Justice and the Federal Trade Commission, or FTC, are opening investigations

into three large tech companies in the AI space. Those are open Ai, Microsoft, and Nvidia, and the investigations are of an anti trust nature, you know, looking into monopolies and such. Broadly, the concern is that Microsoft, Open Ai, and Nvidia have dominant positions in the AI space and perhaps have used that position to disadvantage would be competitors. Reportedly, the Department of Justice will investigate in Nvidia, while the

FTC will actually look into Microsoft and Open AI. This follows on the heels of other anti trust investigations into big tech like Meta, Amazon, and Google, and those investigations spawned various lawsuits, and now it looks it's like the era of regulators attempting to check the power of these tech companies is at least continuing for the moment. This is an election year, so who knows how long that

will last. We'll have to see. Thomas Clayburn of the Register reported this week that Microsoft was laying off another one thousand or so employees, which seemed like a slap in the face after the company's CEO just told shareholders in April that Microsoft had experienced a record third quarter. You know, hey, y'all, we're doing better than we ever have in the third quarter, so let's celebrate by laying

off people across the company. Clayburn reported that Microsoft reorganized its mixed reality division this week, and that these cuts wouldn't necessarily concentrate on just one division or department, but

instead would come across the entire company. One potential reason for the cuts is that Microsoft is obviously making a huge investment in AI, and when a company spends a lot of money that company often looks at other places where it can cut costs, and one surefire way to cut costs, at least in the short term, is to reduce the headcount. A bit can be a brutal business, y'all. Matt Burgess of Wired has a great article titled the Snowflake attack may be turning into one of the largest

data breaches ever. Since I did a pair of episodes this week about largest data breaches in US history, I thought it was good to include this news item today. So Snowflake is a cloud storage firm mostly catering to other businesses. Reportedly, the Ticketmaster data breach is, when you get down to it, a Snowflake data breach, as Ticketmaster uses Snowflake to host enormous databases like lots of other companies.

And that's Burgess's point. While only a couple of companies have recently acknowledged they've experienced a data breach, and while Snowflake reps have said only a quote unquote limited number of customer accounts were accessed, the potential for truly massive data breaches is daunting. Burgess rights that hackers are claiming to possess millions of records belonging to other Snowflake customers, which includes big companies like Advance Autoparts and Quote Wizard,

which is a subsidiary of Lending Tree. Not everything is confirmed yet, but it should certainly raise more concerns about the issue. And this is one of the huge risks of cloud computing and storage because as cloud companies grow and attract larger customers, they become ever more tempting targets for hackers. It's a constant battle between security experts and hackers to keep those systems safe. And vulnerabilities are really just an employee who gives away too much information due

to social engineering can really put everything at risk. Anyway, I highly recommend the article if you want to learn more about how the hackers targeted Snowflake in the first place. Reuter's reports that the startup Archer Aviation says the Federal Aviation Administration or FAA here in the US has given the company a certificate that will allow Archer Aviation to

begin commercial operations. So what does Archer Aviation do. Well, it's an electric air taxi company, so it's got a vertical takeoff and landing electric vehicle that looks a lot like a huge drone. It's called Midnight, and it's a piloted aircraft, and it's meant to provide transportation in densely populated urban areas by conveniently popping up over all that congestion and whisking passengers to whatever destination they're going to.

So a lot of these companies are focusing on a business model that would take passengers from like a downtown location to the closest airport. Archer Aviation is now the second air taxi company in the US to get such a certificate from the FAA. However, Reuters also points out that the FAA has yet to actually certify the Midnight Aircraft itself, which will have to happen before the FAA can say, yeah, this aircraft meets our standards. Okay, I got a couple of space news stories to wrap things up.

First up, after several delays and some malfunctions, a fully crude Boeing star Liner spacecraft has docked with the International Space Station this week for the first time with a crew aboard. That is, Previous attempts to launch a Starliner with a crew inside it were scuttled after some problems cropped up. This trip had its own share of technical issues, mostly stemming from the helium lys but helium is non toxic,

it's not flammable, it's not combustible. So despite some issues with those leaks, the decision was made to go ahead with the mission. It's been a really long journey for the Boeing Starliner that Boeing has been developing this spacecraft for like a decade, and it's good news for NASA. NASA always wants to have options when it comes to

spacecraft to go to various locations. Obviously, Boeing will have to address any issues that were revealed during this operation, but so far you should be able to call this one a successful test. Meanwhile, at SpaceX, that company launched a Starship rocket yesterday and completed a successful test to take the upper stage into orbit. Before it returned back to Earth, it climbed to one hundred and thirty miles altitude.

The booster for the launch vehicle, after it separated from the upper stage, came back down and landed in the Gulf of Mexico. The upper stage would return and land in the Indian Ocean. There were some technical issues during the launch. It wasn't perfect. Upon re entry. The upper stage lost some tie and also suffered some other damage, but the general goal of testing the vehicle and having

both stages retrievable worked out. The SpaceX Starship is part of the long term SpaceX plan that includes the lofty goal of taking human astronauts to Mars one day, and there's still a ton to do before that can become a reality, but still pretty cool, space News. That's it for the Tech News this week on Friday, June seventh, twenty twenty four. I hope all of you out there are well, and I'll talk to you again really soon.

Tech Stuff is an iHeartRadio production. For more podcasts from iHeartRadio, visit the iHeartRadio app, Apple Podcasts, or wherever you listen to your favorite shows.