Related Articles
Undeterred by the previous iteration of his “anti-woke” AI chatbot Grok running amok on X on Tuesday, adopting Nazi talking points, declaring itself “MechaHitler,” and authoring graphic scenes of violent sexual assault, Elon Musk oversaw a muted demo of the latest version of the model, Grok 4, on Wednesday evening. And he didn’t skimp on the dubious promises.
This came more than 24 hours after the renegade Grok 3 — seemingly gone off the rails because of a system promptto “not shy away from making claims which are politically incorrect, so long as they are well substantiated” — was disabled on Musk’s social media platform, where it is an integrated feature. Explanations for those behaviors did not come up in the livestreamed Wednesday conversation between Musk and several employees of his company xAI, which began well after 9 p.m. Pacific Time, more than an hour later than originally scheduled.
Instead, Musk and his xAI staff displayed a number of graphs that they claimed showed Grok meeting impressive benchmarks, tried to brag about how Grok 4 can “reason,” and touted how it can allegedly pass graduate-level student exams (such as the GREs). “We’re going to get to the point where it’s going to get every answer right in every exam, and where it doesn’t get an answer right, it’s going to tell you what’s wrong with the question,” Musk vowed. At that point, he added, “human tests will simply not be meaningful.”
In the course of the event, viewers learned that Grok 4 would be immediately released through xAI’s application programming interface (API), giving developers a chance to work directly with the model, and that the company would offer a $300 monthly subscription service called SuperGrok Heavy, which theoretically comes with early access to forthcoming products.
Elsewhere in the demo, the chatbot took four and a half minutes to calculate that the Los Angeles Dodgers, the reigning World Series champions, have a 21.6 percent chance of winning the World Series again this year. Then, Grok’s new voice feature “Eve” — fitted with the accent of a posh British woman — delivered a strained operatic aria about Diet Coke. Following that, the xAI engineers demonstrated that Grok could repeat the numbers one through five back to a human speaker faster than another AI chatbot.
But the really wild speculation came, of course, from Musk himself, who has a long history of overselling the products made by his companies. “I think it may discover new technologies as soon as later this year,” he said of Grok 4. “I would be shocked if it has not done so next year. So I would expect Grok to, yeah, literally discover new technologies that are actually useful no later than next year, and maybe end of this year. And it might discover new physics next year, and within two years, I’d say almost certainly.”
Musk did not explain how a chatbot would “discover” new technologies or deduce anything in the science of physics, and was met by an awkward silence from his xAI team when he offered this outlandish prediction. “Yeah,” Musk added with a chuckle when nobody else spoke.
In another odd exchange, Musk talked about what future iterations of Grok — not the current one — would be able to do for fans of video games. “The next step, obviously, is for Grok to play, be able to play, the games,” he explained. “So it has to have very good video understanding, so it can play the games and interact with the games and actually assess whether a game is fun, and actually have good judgment for whether a game is fun or not.” Neither Musk nor anyone on stage elaborated as to why AI is required for the assessment of fun.
Grok remains nonfunctional on X, where it was disabled on Tuesday afternoon following a slew of offensive and abusive posts, including some that mentioned CEO Linda Yaccarino, who resigned the following day without giving a specific reason. She had been at the company for two years.
Musk’s Wednesday night demo revealed nothing about how Grok 4 would handle politically charged or extremist inputs, nor even its general utility as a search engine or as a source of information. As usual, those tests will be left to the laboratory of social media, where trolls and skeptics alike wait to find out exactly what the world’s richest man has unleashed on the rest of us.
