The record of flagship AI fashions that missed their promised launch home windows continues to develop.
Final summer time, billionaire Elon Musk, the founder and CEO of AI firm xAI, mentioned that Grok 3, xAI’s subsequent main AI mannequin, would arrive by “finish of 12 months” 2024. Grok, xAI’s reply to fashions like OpenAI’s GPT-4o and Google’s Gemini, can analyze pictures and reply to questions, and powers quite a lot of options on X, Musk’s social community.
“Grok 3 finish of 12 months after coaching on 100k H100s must be actually one thing particular,” Musk wrote in a July submit on X, referring to xAI’s huge Memphis-based cluster of GPUs. “Grok 3 shall be a significant leap ahead,” he mentioned in a mid-December follow-up submit.
But it’s January 2, and Grok 3 hasn’t arrived — nor are there indicators its rollout is imminent.
In reality, some code on xAI’s web site noticed by AI tipster Tibor Blaho suggests that an intermediate mannequin, “Grok 2.5,” could land first.
Grok[.]com is probably coming quickly with Grok 2.5 mannequin (grok-2-latest – “Our most clever mannequin”) – thanks for the trace, anon! pic.twitter.com/emsvmZyaf7
— Tibor Blaho (@btibor91) December 20, 2024
Granted, this isn’t the primary time Musk has set a lofty purpose and missed it. It’s well-established that Musk’s pronouncements concerning the timing of product launches are sometimes unrealistic at finest.
And to be truthful, in an interview with podcaster Lex Fridman in August, Musk mentioned that Grok 3 would “hopefully” be out there in 2024 “if we’re fortunate.”
However Grok 3’s MIA standing is attention-grabbing as a result of it’s a part of a rising development.
Final 12 months, AI startup Anthropic did not ship a successor to its top-of-the-line Claude 3 Opus mannequin. Months after asserting {that a} next-gen mannequin, Claude 3.5 Opus, would by launched by the tip of 2024, Anthropic scrapped all point out of the mannequin from its developer documentation. (In accordance to at least one report, Anthropic did end coaching Claude 3.5 Opus someday final 12 months, however determined that releasing it didn’t make financial sense.)
Reportedly, Google and OpenAI have additionally suffered setbacks with their flagship fashions in current months.
This may very well be proof of the constraints of present AI scaling legal guidelines — the strategies corporations are utilizing to extend the capabilities of their fashions. Within the not-too-distant previous, it was potential to attain substantial efficiency boosts by coaching fashions utilizing huge quantities of computing energy and bigger and bigger knowledge units. However the positive aspects with every technology of mannequin have begun to shrink, main corporations to pursue various strategies.
Grok 3 is coaching with 10X, quickly 20X the compute of Grok 2
— Elon Musk (@elonmusk) September 21, 2024
Musk himself alluded to this within the Fridman interview.
“You’re hoping for [Grok 3] to be state-of-the-art?” Fridman requested.
“Hopefully,” Musk replied. “I imply, that is the purpose. We could fail at this purpose. That’s the aspiration.”
There may very well be different causes for Grok 3’s delay. xAI has a a lot smaller staff than lots of its rivals, for one. Nonetheless, the slipped launch time-frame provides to the physique of proof that typical AI coaching approaches are operating up towards a wall.
