Provide 1000+ AI efficiency tools丨 AIEVA.CN 丨Welcome to visit
AI奇点网-AI资讯特刊丨7月1日
On July 1, local time, "night owl" Musk revealed on Twitter (X) that the large language model Grok 2 version of his artificial intelligence startup xAI will be released in August!
"Blue Star Big Mouth" Musk announced the above news in a tweet in response to questions from netizens.
Lao Ma's revelations about the new model are not only that, he directly announced that the next generation of large models, codenamed "Grok 3", will be released at the end of this year. This is a large super-parameter model that will be trained on a cluster of 100,000 NVIDIA H100 graphics cards.
Musk stressed that Grok 3 will be very "special".
At present, there are only a handful of AI companies in the industry that can control 100,000 NVIDIA H100s, only Laoma and Microsoft. In the era of large language models where the "big flying bricks" rule still continues, the performance of Grok 3 is very exciting.
Musk also revealed more details, and the research team spent a lot of work cleaning the corpus needed for large language models from Internet data. Grok-2, which will be launched in August, will make great strides in semantic recognition.
It is understood that the source of the corpus data of the Grok series model, in addition to the public Internet text, Musk also has the unique Twitter (X) massive social platform data.
Lao Ma refers to Twitter (X) as the "Earth Group Chat".
Since last year, Musk has banned other artificial intelligence companies, including rival OpenAI, from collecting data from Twitter's (X) corpus.
However, social media platforms that are open and free to talk about have a similar disadvantage, which is that it takes a lot of time to purge useless "nonsense" information. This is exactly why Musk pointed out that Grok 2 consumes a lot of labor costs.
In March this year, xAI announced the open-source "Grok-1" large model, a large language model with 314 billion parameters, which is the first open-source large model with ultra-large-scale parameters in the open source industry. Since then, Musk has been nicknamed "Source God" by domestic netizens.
Soon after, xAI successively launched the Grok-1.5 large language model and its first multimodal model, Grok-1.5 Vision, which claimed to surpass GPT-3.5 in performance and comparable to GPT-4V in multimodal capabilities.
In May, Musk tweeted that xAI is a new company that has been around for less than a year, so the Grok AI system still has a lot of work to do to compete with industry leaders like Google and OpenAI, and perhaps catch up with the first group by the end of this year.
[This article is from丨AI Singularity Network丨The same name of the whole network account丨Welcome to search and follow]
AI Singularity丨Provide 1000+ AI efficiency tools丨 AIEVA.CN 丨Welcome to visit