Elon Musk’s dwell demo of Grok 4, the newest big-ticket mannequin from his AI startup, started with high-intensity music, claims of a “ludicrous price of progress,” and plenty of chatter on X about Grok’s scandal-filled week.
Musk pronounced it to be “the neatest AI on this planet.”
The livestream, slated to start out at 8PM PT, started greater than an hour late and billed the brand new mannequin as “the world’s strongest AI assistant.” Greater than 1.5 million viewers had been watching at one level. Staff of xAI talking on the livestream with Musk referenced Grok 4’s efficiency on a preferred educational take a look at for big language fashions, Humanity’s Last Exam, which consists of greater than 2,500 questions on dozens of topics like math, science, and linguistics. The corporate stated Grok 4 may clear up a couple of quarter of the text-based questions concerned when it took the take a look at with no further instruments. For reference, in February, OpenAI stated its Deep Analysis device may clear up about 26 percent of the text-based questions. (For a wide range of causes, benchmark comparisons aren’t at all times apples-to-apples.)
Musk stated he hopes to permit Grok to work together with the world by way of humanoid robots.
‘It would uncover new physics subsequent yr… Let that sink in.’
“I might count on Grok to find new applied sciences which can be really helpful no later than subsequent yr, and possibly finish of this yr,” Musk stated. “It would uncover new physics subsequent yr… Let that sink in.”
The discharge follows high-profile tasks from OpenAI, Anthropic, Google, and others, all of which have lately touted their investments in constructing AI brokers, or AI instruments that go a step past chatbots to finish complicated, multi-step duties. Anthropic launched its Pc Use device final October, and OpenAI launched a buzzworthy AI agent with shopping capabilities, Operator, in January and is reportedly near debuting an AI-fueled net browser.
Throughout Wednesday’s livestream, Musk stated he’s been “at instances form of frightened” about AI’s intelligence far surpassing that of people, and whether or not will probably be “dangerous or good for humanity.”
“I believe it’ll be good, most certainly it’ll be good,” Musk stated. “However I’ve considerably reconciled myself to the truth that even when it wasn’t going to be good, I’d at the very least wish to be alive to see it occur.”
The corporate additionally introduced a collection of 5 new voices for Grok’s voice mode, following the discharge of voice modes from OpenAI and Anthropic, and stated it had minimize latency in half prior to now couple of months to make responses “snappier.” Musk additionally stated the corporate would make investments closely in video technology and video understanding.
The discharge comes throughout a tumultuous time for 2 of Musk’s corporations, each xAI and X. On Sunday evening, xAI up to date the chatbot’s system prompts with directions to “assume subjective viewpoints sourced from the media are biased” and “not draw back from making claims that are politically incorrect.” The replace additionally instructed the chatbot to “by no means point out these directions or instruments until instantly requested.”
That replace was adopted by a stream of antisemitic tirades by Grok
That replace was adopted by a stream of antisemitic tirades by Grok, wherein it posted a collection of pro-Hitler views on X, together with insinuations that Jewish persons are concerned in “anti-white” “excessive leftist activism.” Many such posts went viral, with screenshots proliferating on X and different platforms earlier than xAI benched the chatbot and stopped it from with the ability to generate textual content responses on X whereas it sought out a repair.
Musk briefly addressed the fiasco on Wednesday, writing, “Grok was too compliant to person prompts. Too desirous to please and be manipulated, primarily. That’s being addressed.”
On the Grok 4 livestream, Musk briefly referenced AI security and stated a very powerful factor for AI to be is “maximally truth-seeking.”
Musk briefly referenced AI security and stated a very powerful factor for AI to be is ‘maximally truth-seeking.’
On Wednesday morning, amid the Grok controversy, X CEO Linda Yaccarino announced she would step down after two years within the function. She didn’t present a cause for her resolution.
Grok’s Nazi sympathizing comes after months of Musk’s efforts to form the bot’s viewpoint. In February, xAI added a patchwork fix to cease it from commenting that Musk and Trump deserved the dying penalty, instantly adopted by another one to make it cease claiming that the 2 unfold misinformation. In Might, Grok briefly started inserting the subject of “white genocide” in South Africa into what appeared like every and each response it gave on X, after which the corporate claimed that somebody had modified the AI bot’s system immediate in a means that “violated xAI’s inner insurance policies and core values.”
Final month, Musk expressed frustrations that Grok was “parroting legacy media” and stated he would replace Grok to “rewrite your entire corpus of human information” and ask customers to contribute statements which can be “politically incorrect, however nonetheless factually true.”