The list of iconic AI models that have missed their promised release windows continues to grow.
Last summer, billionaire Elon Musk, founder and CEO of AI company xAI, said that xAI's next major AI model, the Grok 3, would arrive by the end of 2024. GrokxAI's answer to models like OpenAI GPT-4o and Google's MethonIt can analyze images, respond to questions, and power several features on X, Musk's social network.
“After training 100k H100s Grok 3 should be something really special by the end of the year” Musk wrote. A July post on XAI refers to xAI. Memphis-based ensemble GPUs. “The Grok 3 will be a huge leap forward,” he said. He said. A follow-up post in mid-December.
But it's January 2nd and the Grok 3 hasn't arrived yet – and there's no sign of its imminent release.
in fact, AI tipster Tibor Blaho spotted some code on xAI's website. recommended. The mid-range model “Grok 2.5” will land first.
Grok..com has launched the Grok 2.5 model (grok-2-latest – “Our most intelligent model”) – thanks for the hint. pic.twitter.com/emsvmZyaf7
— Tibor Blaho (@btibor91) December 20 2024
It's not the first time Musk has set a lofty goal and missed it. That's it It is well established. Musk's announcements about the timing of product launches are often unrealistic.
To be fair Musk in an interview with podcaster Lex Fridman in August He said. The Grok 3 is expected to be available in 2024 “if we're lucky”.
But the Grok 3's MIA status is interesting because it's part of a growing trend.
Last year, AI startup Anthropic has failed to appoint a successor to its top ranks. 3 Close the job. Model. Months later Announcing The new generation model Claude 3.5 Opus will be released by the end of 2024 Anthropic was cancelled. All descriptions of the model from its developer documentation. (according to (According to a report, Anthropic completed training on the Claude 3.5 Opus last year, but decided it didn't make economic sense.)
Google and OpenAI are also reported to be. felt. discomfort, With their flagship models in recent months
Perhaps this is proof of limitations. Current AI scaling laws – Companies use methods to improve the performance of their models. In the not too distant past, Large performance improvements can be achieved from training models using larger and larger datasets. But profits shrink with each model generation, leading companies. to follow Alternative technologies.
Grok 3 is training at 10X; Soon Grok 2's calculation is 20X.
— Elon Musk (@elonmusk) September 21 2024
Musk himself said this in an interview with Fridman.
“Are you expecting to modernize (Grok 3)?” Fridman asked.
“I hope so,” Musk replied. “I mean, that's the goal. It may fail in this goal. That's a wish.”
There could be other reasons for Grok 3's delay. xAI has a much smaller team than many competitors. Nevertheless, The slippage time frame adds to the evidence that traditional AI training approaches are running into a wall.