When Elon Musk announced his new AI chatbot “Grok” last month, almost everything about it looked like a publicity stunt or a vanity project or both. The product was seemingly built over a matter of months, it was only available to a small group of X users and its chief selling point was Musk’s promise it wouldn’t be constrained by the “woke mind virus.” He boasted that Grok’s chatbot was “based”—a slang term meaning unapologetically right wing—and posted fratty memes meant to emphasize Grok’s dominance over market-leading competitor ChatGPT.
None of this inspired confidence. But with Grok now available to paying users of Musk’s social media platform, the reviews are coming in. And they’re surprisingly positive. This week we are joined by Bloomberg reporter Shirin Ghaffary, who has been testing out the new chatbot and comparing it ChatGPT and Google’s Bard chatbot.
Well, Elon Musk gives now the richest person on the planet.
More than half the satellites in space are owned and controlled by one man.
Starting his own artificial intelligence company.
Well, he's a legitimate super genius. I mean legitimate. He says. He's always voted for Democrats, but this year it will be different.
He'll vote Republican.
There is a reason the US government is so reliant on him.
Elon Musk is a scam artist and he's done nothing.
Anything he does is fascinating the people. Welcome to Elon Inc. Where we discuss Elon Musk's vast corporate empire, his latest gambits and antics, and how to make sense of it all. I'm your host, David Papadopolis. We were surprised this week to learn from our colleague Sharen Gafari that GROC, the AI chatbot from Elon Musk, this very new AI company, is actually giving programs like Chad GBT a serious run for their money. We were less surprised when we saw the EU as investigating X for illicit content and disinformation on the platform, and we did raise our eyebrows when a recall alert for Tesla's autopilot software went out last week. Now, this wasn't the kind of recall we're all used to. The company called for users to update their cars software remotely. It felt like a peek into the future of cars. To discuss all of this and more, we've beamed in Sharene, who reports on AI for US in San Francisco. Hello, hither and Max Schafkin, senior reporter and editor at Bloomberg BusinessWeek.
Hello David, Hello Max.
Later on, Dana Hall, our longtime Tesla reporter, will join us as well. Okay, so now Max, we know, okay that recently you had a little cameo and last week tonight with John Oliver, they found some old tape that they put on.
So in that way, he's kind of the Henry Ford of space.
But you know, Max, we also pulled some old tape of yours on our very first podcast, our debut episode, in which you said this about groc.
Until people start actually using this, I think it's worth like thinking about this as vaporware or not even taking it seriously as a real product.
All right, Well, stand okay, so Max, we'll give you your time. Just pipe down, you'll get your chance to respond. But Sharen, you've tried groc Out for a few days recently. What did you find?
It exceeded my expectations? I like max given first of all, how how much other companies had leg up over Elon with this one right open a eyes chat SPT came out over a year ago. Google's been working on AI forever. I didn't think in a matter of a couple months and a small team could spin up something that would significantly be better than any of the other AI chatbots out there. However, I did find that, especially with one important use case for me, which is summarizing real time events, Rock performed really well.
And when you say really well, a tell us what that looked like? What really well looked like? And is really well better than the competition?
Yes, I should say really well comparative to other chatbots that are out there right now, I think none of these tools are anywhere near perfect. They're still wrong a good percentage of the time. I will caveat that my test was not a scientific one. You know, I used it for three days and asked a bunch of news questions basically, and you know how it did better? Was that a lot of times? And I've written about this in the past. The mainstream chatbots out there, the leading moost cutting edge ones can be really smart if you ask them to solve, you know, a math problem or to write you a poem in the style of Shakespeare. But when you have some about current events, they can just totally get stumped and flub the answer because even as they're even if they're being updated with real time information, it takes training to be able to tell you know that the chat bought hey, this is right, this is wrong, and you can't constantly be training something in real time at least they haven't figured that out quite yet. And so where Grock excelled was that when I asked it about the news, it actually was able to give me, in my experience, a better answer, more or less than the other ones out there, like chat, Scheept, your Google's.
Bard and so you're talking about right, you asked the things about the Israel Hamas war, you asked the things about the cop twenty eight climate summit, and it did write better often than Chat, chapt and Bart. So Max Chafkin, what.
Say you, I agree with everything Shreena's saying. I mean it's a really.
You retract your statements from Wakwa.
No, because I do.
First of all, it's not clear how quickly they put this thing together, like it might have been true, you know, four weeks ago when I said it, and maybe they actually moved really quickly. I also think, you know, it's sort of hard to know how to think about this because on one hand, you know, it's like, Wow, what an impressive showing by Elon Musk. He's still got it right, Like he can small team, he's able to, you know, despite all of the crazy things swirling around him, he's able to recruit competent engineers and who are able to then build a product.
I mean, the other way looking.
At is I guess maybe chat GPT wasn't quite as good as we thought. Maybe like that thirteen billion dollars that open Ai spent, or you know, this enormous amount of money, many years spent. The fact that one guy could replicate it in a matter of I don't know, we don't know exactly how long they built this, but probably a few months.
What does that say about the state of the art.
Maybe it's maybe these maybe these chatbots just aren't as great as they as we thought they were. Or maybe a better way to say it would be maybe these chatbots aren't as like sort of defensible as we thought they were. Maybe, like there's not that much unique in anything.
There's not much of a n buffet terms of a moat around them as as one would have thought. I mean, now, sure, and you're you're saying you peppered it with news questions when and to to help you be up to date on what's going on in the in the open AI space, in the artificial intelligence space. How important is that functionality? Is that something that people would actually legitimately use it a lot for?
Look, I think there's there's you know, you don't always need to know what's happening in right away, like right now. For example, if you're just interested in using these bots to code, to help you code, it doesn't really matter. Code doesn't you know, change minute by minute. But if you care, and when you think about when you use Google every day, right, how many times does it matter if information is fresh when you Google a lot of times it matters, like if you want to know when the super Bowl is or if you want to know like when to go to the movies, Like, so I asked it you know about Israel Amas war and I did that because in the past when I had asked Google's chatbot and chat Shept, it was flat out wrong. It was telling me, you know, in the p the heat of the war, it was telling me there's been a ceasefire, which was just not true. And when I asked GROC, it gave a pretty nuanced and correct answer.
Well, exactly was that answer?
It was pretty concise. It was about a paragraph. It said, you know, the war is continuing, there's no end in sight. It listed, you know, an estimate of the casualties and gave sources for those casualties. And you know, the reason I think that that GROC may be better with this and some of its competitors is because, first of all, it's willing to engage on some of the stuff. So when you asked Bart about the Israel mas war, for example, after Bard a couple of months ago, embarrassing.
Or declined to comment.
Basically Bard declined to comment, right, And then this is mind you after my reporting that it flunked on this question a few months ago, so now it seems like it's shut down the answer entirely to what's happening with Israel Amas war. Chech Pt did give a directionally correct answer actually performed better than I thought it would. But Groc's answer was more concise. It listed references better I thought, And I think a big reason for that is because it's pulling from from tweets or I should call them exposts. And you know, what was really surprising to me was that I thought, Okay, even if it's pulling from people's exposts, that a lot of them could be junkie. I mean, just looking at my own, for you feet as degraded in the past couple of months or a year, I should say. However, when I asked Rock, it seemed like it was filtering where a majority of vast majority of sources it was pulling from. The in these tweets or exposts were you know, mainstream news sources like Bloomberg, like the Financial Times, Sky News.
Wait, which, so they're siting Bloomberg, Oh we love Grock, then Croc's great.
Now, of course they're going to be right if they're signing Bloomberg.
Now, Max, in.
That opening episode, when you were expressing those concerns and and and your your you know the fact that you were a bit dubious about it, that was specifically one of the things you brought up that hey, if you're pulling from X, which is what GROC is going to largely be doing. And indeed, as as Sharen was saying that the content on X is taking a turn for the worst, how relyible can it be? So I guess maybe they've managed to come up with the formula to make this work.
I mean to me, what's most from a sort of if you were an investor in one of Elon Musk's companies, or you were his friend or something, what will be most encouraging about this is that you have the rhetoric, right, you have Elon Musk talking twenty four to seven about the woke mind virus, and then you look at what grog is actually doing. I was reading Sharen's story and I saw Financial Times, Bloomberg, Wall Street Journal like Where's cat Turn? And you realize, like, this thing has the woke mind virus at least as Elon Musk has articulated it, and so clearly he is sick. He's able to you know, this would make you think, okay, maybe what's going on in his own personal you know, Twitter feed X feed is a performance and when it when it really matters, when it gets down to business, he's making better decisions or people underneath him are able to say, like, look like, let's not like exclusively compose these answers of like the right wing influencers that are you know, friends with Elon.
Let's let's useables. So GROC is so far better than expected. It's doing well, It's passed some surprisingly some early tests. Scharene does does it matter though, like in terms of making this some way scalable monetizable in a way that matters for XAI and for X and for Musk. Yeah, is there is Does this have any actual import it?
Could you Look, if X really builds a reputation for being the chatbot that will give you the most accurate information about the world as it's happening, I can definitely see that being monetizable or being a competitive asset. But it's quite you know, a paradox because at the same time, as you're pointing out, Musk is under fire for quite the opposite, right for you know, how unreliable some people are finding X to be. So it's kind of a hard sell too, even if Grok is exceeding expectations in this in this one area, it's the area of exact weakness for him right now in terms of X's perception.
Now, Max, indeed you are in general you're dubious about AI, not just as it applies to Grock, but across the board. What's your take on this?
I mean, I think, look in certain ways, it makes a lot of sense, just because Elon Musk needs to create some reason for people to use this site and to subscribe to X premium and Groc would be a way to do that.
So who does indeed have access to GROC? Who can use Grok?
Right now?
I believe it's it's I hope I'm getting the branding right. I think it's ex Premium plus users. This is the higher tier from the.
Twitter blue and apparently that group includes Scharene from what we can tell.
Sharen, is it like twenty bucks a month or how much? How much are you show up?
That's right, I'm a VIP users.
I believe it's around sixteen dollars a month, and it is the highest package you can get on the pagement. And I got it specifically for this reason, mind you, not for my own vanity for the check mark. I got it just because I wanted to test out there likely story.
I mean, there's some really interesting issues around corporate governance because it seems like Grock is a separate company, and yet clearly it's very closely tied in with X. It seems to be part of Elon Musk's plan to like, you know, save the value of his investment.
So so I think that that's one thing.
The other thing is, you know, for a long time, even going back to the to the old Twitter, everyone knew there was a lot of value in the data contained in tweets, and and Twitter had had a really reliable revenue stream. I think to some extent still does selling that data called the fire hose to companies that would essentially pay for it and analyze it and so on and and so you know, in certain ways, Grock is like a logical the logical next step. On the other hand, in other ways, the kind of Elon Musk media troll way he has undermined that he's undermined Twitter X as a sort of reliable record of sentiment, whereas now it's it's more like a reliable reliable record for sentiment in certain domains like sports and so on. I bet you Grock will be great on you know, the question of like which who to start in your fantasy league or whatever, and less reliable on some of these more controversial issues where we've seen really seen an impact of Elon Musk's you know, leadership, or and and sort of the the right wingification of the site, so.
Max and and in that vein. Then the EU investigation of X which was announced i believe earlier this week for disinformation. They're investigating on disinformation and illicit content. Does this does the fact that GROC is a more reliable source, does this in any way, shape or form help them fend off that EU investigation? Are these totally unrelated things.
You know, I'm not an expert on European technology regulation. I'm not sure that anybody is, but I know I don't think it does. I mean, this is very much a niche product. If what you're worried about is misinformation hate speech. By the way, in parts of Europe, of course, in Germany, certain kinds of hate speech that is you know, allowed in the United States is not allowed there. So if you're worried about like bad content broadcasting widely, this doesn't do anything to address that. Right, You're talking about a very narrow segment that has access to this thing using it in a narrow way. I don't think Sharene was the kind of person who was, like, you know, liable to like retweet, you know, the latest cat turd or whatever. And that's the kind of thing that European regulators are worried about, although I should correct myself and say I don't think they're specifically concerned about cat turn I'm using that right stand in for the overall issue of disinformation. When I was in college, we had a cat called Claws, and so it's kind of like the same thing.
It's like a small part.
I got to describe the whole cat, the whole understood, by the way, And I don't know if anybody saw this or anybody can make heads or tails of it. So there's there's another groc out there is this?
Right?
Is anybody?
Yes?
This is so confusing?
This may help us, Serene.
Okay.
So Elon Musk's former partner and you know, mother to his children. So she came out with a rocket ship AI plushy from Curio named Grock, same name, and this came out. This was announced, you know, I believe let's see, just like was it last week? She announced on Instagram. It's this really cute looking you know, great, I saw it. It's not basically stufty animal, that's aipower.
What kind of odd so Grock custody battle. Yeah, I don't know exactly, so GROC. So we're they're fighting over the kids and now they're fighting over Groc And yeah, it's odd. We went from having zero Gros in our lives the now multiple in the span of weeks. So do we think that this was deliberate?
Like is this like? I mean, it would have to be right.
There's no way that they could have just independently arrived at the same So I.
Saw a techcrun headlines today that she had the trademark before he did or something. I don't know if that means you came up with it first.
But we're going to say goodbye to Sharen for now. That was great train that we're going to have you back on both later in the show, and we're going to have you back on once you you know, test other elements of Groc will bring you back. Okay, So welcome back. Max and I are still here, and we're bringing in Dana Hall, our longtime Tesla reporter, to help us talk about the recall. Hello Dana, Hello Hello. Okay, So Tesla indeed recalled a few cars the other day, and you know, it's attracted a lot of attention because even my mother, who doesn't happen to own a Tesla, asked me what was going on, So Dana help her and everyone else out what exactly did Tesla do and how big a deal is this.
So the first thing to know is that the word recall is a little bit of an anachronism when it comes to the auto industry these days. You hear recall and you think, oh my god, like all of these cars, like two million cars are going to be yanked off the roads. It's like, you know, like a baby's toy is recalled because there's a choke hazard. Like the idea of a recall makes you think that everything is off the market. But you know, now, like a lot of automakers, it's Tesla in particular, they can improve the product via over the air software updates. So, yes, this was a recall, but the remedy is that, you know, Tesla, free to consumers, is going to beam and over the air software update to everyone and the consumer does not need to do anything you don't need to bring it into a shop. You can still drive it. This is Tesla basically admitting, voluntarily, with Nitza's guidance, that autopilot is a problem and that drivers tend to get lulled into kind of like a sense of complacency, and that in certain conditions drivers may not pay attention. So the remedy is that Tesla's going to nag you a little bit more to keep your hands on the wheel and your eyes on the road. And I think the significance here is that, you know, for years, like Nitza has been working with Tesla to kind of improve the company's safety culture. They invest gated all these accidents where Tesla drivers on autopilot or with autopilot engaged seem to keep smashing into stationary fire trucks, and it was like, hmm, what's going on with that? And then you know, this kind of recall that was announced last week was basically the result of those two years of negotiations back and forth between the regulatory agency and the company.
It's Dan's right that, like, you know, there you could read this in a very literal way and think, oh my god, they're going to take all these cars and you're gonna have to give them back and they're gonna be whatever rebil.
A literal sense in the old in the old sense of the world.
Well maybe, but I mean a lot of recalls work like this, Like if your Ikea bookshelf was recalled, they don't take the bookshelf back. They just send you like the new pins to like hold the shelves in better. But it is bad, and it's it's bad, especially in relation to what Elon Musk has been saying for years and years and years. We are so far away from the swaggering you know, we're gonna have robotaxis immediately. You barely have to look at this thing. You know, I before before going on, like I went back and you know, there was that that video that that they ran, the Painted Black video. I think Dana is that the right song from sixteen?
Yeah.
Yeah, it's just like awesome video where the cars like driving around and there's a little thing on the bottom of the screen says the driver is not doing anything, he's just there for legal reasons.
Right as seven years ago, yeah exactly.
And and then you had you had like you know, I think shortly after that, an adult film actress you know, tweeted a video, you know, having sex in a Tesla, and Elon Musk was like, ha ha, ha isn't it great? Another use and like this was all kind of an exaggeration, like they this thing doesn't work the way Elon Musk has said it works for years and years and years and years. And not only that, it doesn't work the way many many Tesla investors have sort of been assuming it would work. So in that sense, like I think this is damaging.
But max to that point, right, because you're saying investor, Tesla investors have been and assuming that the technology works are going to work like that soon ergo. That is part of the very very lofty valuation that Tesla has. It's worth more than all the other automakers combined, and by a lot, but the stock price didn't move it all off this really well.
The thing is that even though Tesla markets it as autopilot, the truth is that it is a level two system, which means that the driver is always in control. And there's like sort of daylight between how it's marketed and how it's actually used legally, Like if you get a Tesla and you read all the documents and even look at their website, they make it very clear that the driver is supposed to be in control at all times.
But then presumably Knitz's concern though, is that no, you know, reasonable or typical user is actually going to read all the fine print, is that right? And so they are going to perhaps think that like, oh, whatever autopilot's got it is that right.
I mean, there may be an increased risk of a collision if you are if auto steer is engaged and you're not fully paying attention.
It says the new warning that you see what where does that pop up? By the way, that pops up one on your on your screen as you as you sit down in the car.
So the remedy is that there will be increased visual alerts when you're driving, and like, if you ignore all the alerts completely, like the car could basically ground to a halt and pull you over.
It.
Tesla already kind of like nags drivers to pay attention. Now the nagging is more pronounced and more severe. But I mean, what's interesting is that you know, Tesla initiated this recall. This wasn't nitsa coming in and like forcing a recall, Tesla basically begrudgingly acknowledged it and said they would do it. They said that they didn't agree with Nitza's findings, but they were they were going along with it.
I mean one reason for that is that there are worse things that NITSA could do, I mean, and and so like what well, I mean, they could force them to change the name. I mean, they could, they could like you know, they could they could more heavily regulate or try to stop the way Tesla has talked about these products, both autopilot and it's you know, quote unquote quote full self driving product, you know. And I think you could make an argument that both of those names are a little bit misleading. And you know, I think there's an assumption, right that this is going to sort of settle the matter for NITSA, although NITSA you know, released the stament saying that they are not necessarily like they have not closed this investigation.
Yeah.
And it's interesting too because there's a lot of litigation coming up on behalf of you know, drivers and their families who were injured or killed in accidents where autopilot was or was not engaged. And you have all these plaintiffs, attorneys who are very keen to kind of look at this recall as part of their strategy going forward. There's some big trials that are coming up or slated for twenty twenty four involving some of the more high profile autopilot cases.
What do those lawsuits allege.
Well, they basically alleged that the product is defective and that Tesla knowingly marketed this defective product that like lulled their these drivers into this false sense of complacency. But it's tricky because you know, from Tesla's perspective, like all of the fine prints says that you've got to pay attention and if you didn't, then then it's on you. And just you know, when you're in a car and there's something like autopilot engaged, like you know, as a driver, it can be confusing, like in terms of how much is it actually capable of when do you take control? It's like that murky middle it's not a full robotaxi, but it's it's helping you like keep lanes and you know, like keep speed, and you know it's it's it's a driver assistance project. But to Max's point, a big part of the valuation of Tesla is the promise of robotaxis, and Musk raised billions of dollars from Wall Street with that idea, and like a full robotaxis always kind of right around the corner. And that's been going on like year after year.
Now.
Yeah, I was gonna say, that's a regulatory problem, like getting we like, they are so far away from getting the federal government to say, okay, you can use your Tesla as a robotaxi. I mean we've seen you know, Cruise essentially like layoff a huge sex segment of the company basically slammed the brakes on its robotaxi business. It really just seems like, you know, good for Elon to solve this problem, but again, you are so far away from what that ultimate goal is, which is you know, the driver's not looking at all.
Yeah, and I mean to go back to twenty sixteen, which was your reference to that ad that you remembered from back then, I remember being involved in a story at the time absolutely wasn't just a Tesla thing, but across the board, there was this sense out there in the industry that like, you know, we're days, we're hours away from from everybody, just you know, sitting in the backseat and sleeping while their car drives. Yeah, that was a myth.
By the way, that was caused.
You know, I've done some reporting on this that was caused in part by Elon because because Elon went out there and said that that this is like definitely gonna happen, then you had other CEOs promising to buy certain numbers of Tesla. I believe it was a CEO of a big ride sharing company promised to buy you know, big, big quantity of of Tesla's and that caused everyone to make models right showing enormous sales for you know, like like it caused this whole like Wall Street apparatus to like kick into gear and develop all these justification.
Wall Street military industrial costs, and a.
Lot of that has kind of has gone away, like people have sort of woken up to the fact that, like this stuff hasn't happening as quickly as it has.
Accepted.
One important exception, which is test that'sl stock brash. So it's probably worth saying that. You know, there's this other case going on, which is the case of the Nicola founder who has been found guilty of defrauding investors around you know, basically the capabilities of this what was essentially a vapor ware truck now much much more serious case.
I mean, and look, investors love who investors who have autopilot, I mean a lot of them love it. They see it, they see it getting better, they experience it getting better. They feel like, Okay, Elon might be wrong on his timelines, but if anyone's going to figure out this problem, it will be him. And Tesla has a lot of data. I mean, that's the thing. The other thing, like you know, Weimo and all these other companies that kind of tried to do full robotoxis are you know, training small numbers of vehicles in these kind of geofence locations, whereas Tesla is learning from all the autopilot data that it is just unleashed on the universe. And you know, they've got a really big fleet now, so their model is based on a lot more real world data.
All right, So last question on this topic that I have for both of you is, and this is kind of brings us back to the beginning of the conversation on it is so is recall then in the end the right word and is it the word we're going to use going forward on something like this, Dana.
You could say a software recall, because ultimately the remedy is to update the software. You could say a software recall.
It's a recall. It's what I mean this whole conversation about it. Is it a recall?
Yeah, there are lots of recalls that that are satisfied without taking the product in. But like if you're if you're having to like make a change to your product, that is because it's unsafe.
And that's what it's just saying.
So sharene welcome back. Now. You need to know that. One if you hear that is a source of never ending raging debate on the show is the looming death match cage match between Elon Musk and his rival Mark Zuckerberg. Max has very strong opinions, I have very strong opinions, Dana and Sarah Friar. Who's not with us? I think we're crazy and should move on and do other things with our lives. Sharen, who do you think wins this match when it happens?
It's tough because I mean, look, Elon's a bigger guy, right, he's six to one.
Sucks he's seven. But I've been saying let her answer, So who wins.
But I think Mark Zuckerberg and look, Mark Zerberg also recently had an injury, So.
Flip flow, who's gonna you're hem and Han who's gonna win?
Okay, okay on a good day.
I think Mark, if he's recovered from his injury, I think it suck because I think he's trained for this conditionally.
He posted on Instagram that he's doing rehab. He's going to be better than ever.
So sucked it.
Okay, good, something that'll keep in mind.
But we decided to just see just how brilliant Groc is. Right, we said, why don't we say why don't we hit Grock with the question. So, Scharen, you did you asked Groc who will win the fight? And it told you what.
It said. It also, like myself, gave a very kind of reasoned response, and it said, you know, listen out that Zuck is smaller. He gave his hide his weight. It said, as for their fighting styles, Zuckerberg's jiu jitsu training could give him an advantage in grappling and ground fighting, while Musk's height and reach advantage could be useful in striking.
Very clinical atwer distance.
Wow yeah, And it said ultimately, the outcome of the fight would depend on various factors such as training, preparation, and strategy. So let's wait and see if this hypothetical match ever becomes reality.
So a non answer, that's a lay of answer, and it says Musk is known for his his interest in physical fitness, eh and martial arts. What it doesn't say is that Musk is a life long brawler, which he's been fighting since he was you know, since he got out of the crib, and that's why he's going to win any of it. It's kind of indeed, Sharen, it's kind of a lame answer. So he might have done great on the news questions the cage match answer, I give it a zero one.
Yeah, here's the thing. They advertise it as a chat bought with a rebellious streak right and ready to answer spicy questions. And you know, you would think on something like this it would weigh in. By the way, I fact checked Elon Musk's height just now.
Correction.
According to the Internet, he is six ' two, not six ' one, so Grock maybe even underestimated his height.
Okay, enough with the cage match talk. Let's end it there. Thanks for listening. To Elon Ing and thanks to our panel Scharene Dana Mac.
Great to be here, always a pleasure, Thanks for having me.
This episode was produced by Stacy Wong. Naomi Shavin and Rayhan Harmansi are senior editors. The idea for this very show also came from Rayhan. Blake Maples handles engineering, and we get special editing assistants from Jeff Grocott. Our supervising producer is Magnus Henrickson, thanks a bunch of BusinessWeek editor Joel Weber. The Elon Inc. Theme is written and performed by Taka Yasuzawa and Alex Suira. Sage Bauman is the head of Bloomberg Podcast and our executive producer. I am David Papadopolis. If you have a minute, rate and review our show, it'll help other listeners find us. See you next week.