AI chatbots are suddenly a dime a dozen. Beyond the massively successful and overhyped options (like OpenAI's ChatGPT and Google's Bard), now Elon Musk's xAI is getting on on the action. Musk's AI company has released the first version of Grok, an AI chatbot that the SpaceX leader heralds as a more humorous and free-thought version of ChatGPT. But what exactly is Grok, is it worth paying to use it?
According to xAI's website, Grok is an "AI modeled after the Hitchhikers Guide to the Galaxy." The Hitchhiker's Guide to the Galaxy is, of course, an old sci-fi novel (slash radio drama slash TV show slash movie), but the name "Grok" appears to be drawn from a different old sci-fi novel, Robert Heinlein's Stranger in a Strange Land. In that book, "grok" is a Martian word with a definition roughly along the lines of "to know something (or someone) on a truly intimate level."
The company says Grok is intended to be used answer almost anything, while also being able to suggest some questions users can ask it. Additionally, the company says that Grok is designed to answer with "a bit of wit" and that the AI has a "rebellious streak," and warns not to use it if you "hate humor." Sure. Elon Musk is, after all, famously hilarious.
The product is still in very early beta, and currently it only has around four to five months of training. As such, xAI says that Grok should improve exponentially over the coming months. But how is it right now?
Grok is...not the worst AI chatbot that has been released in recent months, and based on how hard Elon Musk likes to push his companies to iterate, it is likely we'll see some degree of "improvement" in the service in the months going forward. Whether or not that improvement means making the bot more like its premium competitors—ChatGPT and Claude—remains to be seen.
As it stands, Grok is probably on par with what we saw with ChatGPT's earliest iterations, though xAI continues to claim the service will improve rapidly as it is trained on the content and information freely available on X, formerly known as Twitter, which Elon Musk also owns. (X/Twitter is also, after all, famously filled with accurate information). So far, the company says the chatbot it has shown substantial improvements since Grok-0, the very first iteration of the large language model.
Because Grok is designed to respond with "wit" and to be "a bit rebellious," the chatbot is a good deal different than OpenAI's ChatGPT, or the even more ethically restricted Claude, from Anthropic. According to xAI, Grok was more than 60% accurate across three of the four tests that the company ran it through. However, it's unclear if that information has been tested and verified by independent parties.
Hallucinations—that is, instances of an AI making up facts or answers—are common with chatbots like Grok. And while Grok is still being trained, as noted, the model learns based on of all the information posted on X, which means it has license to leaf every dumb post sent out by millions of users, including all the ones full of misinformation and bad data. As such, I'd recommend double checking everything Grok tells you, because you never know what kind of "fact" it's going to pull from a random user of that service.
The Grok waitlist is currently only open to Premium+ subscribers on X. That means you will need to become a verified user on the service, which costs roughly $16 a month. Sure, it's cheaper than a subscription to ChatGPT or Claude—but consider that both those options have had more training, and also offer higher ethical restrictions. Any intensive use of AI chatbots is probably best saved for those two platforms, or others like them. And as I said, you can't even try it yet—you're only able to apply for the waitlist right now—so that monthly payment doesn't exactly guarantee you access.
Doubts aside, Grok isn't necessarily a bad thing, and more than any AI chatbot is. The more competition in the AI space, the more every company will have to strive to make their products better, which will hopefully be a win for consumers like you and me. If you're debating trying out an AI chatbot, though, I'd recommend giving Grok some more time to build up its features.
The fact that Grok can pull data from X is intriguing in theory, but there's so much dumb data, hateful content, and straight-up misinformation on there, the answers you get from the chatbot might not be be worth taking seriously.
