Muhammad Selim Korkutata | Anadolu | Getty Images
On Tuesday, Elon Musk’s XAI presented its latest artificial intelligence model, Grok 3, claiming that it could surpass the offer from Openi -a -a Chinese Deepseek based on early testing, which included standardized mathematics, science and encoding tests.
“We are extremely excited to present Grok 3, which, we think, is the order of size more capable of Groka 2 in a very short period of time,” Musk said on a demonstration of Groka 3, who streaming on his social media platform X.
The team also said that he was launching a new product called “Deep Search”, which will act as a “next -generation search engine”.
GOK 3 will be introduced to premium X subscribers later during the day, and will also be available through a separate subscription for the web -novels and versions of the model applications, said Xai Tim.
Speaking to the world government’s summit in Dubai last week Musk called the model “Scary Smart”, with powerful explanation capabilities, claiming that it outweighed all other existing models in XAI’s internal tests.
“This could be the last time Ai was better than a grave,” Musk said at the time, adding that he was dressed in “a lot of synthetic data,” and was capable of thinking about his mistakes to achieve logical consistency.
Xai team claimed that early iteria Grok 3 received better grades than existing competitors on Chatbot ArenaThe crowd website that throws different AI models against each other in blind tests.
Towards the end of the product demonstration, Musk said the company would continue to improve the model.
“We should emphasize that this is a kind of beta, which means you should expect some imperfections at first, but we will improve it quickly, almost every day,” he said, adding that the voice assistance for the model will be released on and later time.
Intense competition
Musk, who was quite loud On the potential dangers of artificial intelligence, he began XAI 2023, entering the generative market of AI, which includes Openai’s Chatgpt.
Last September Openi launched its most advanced model, O1, who came with Explanation abilities And they were able to solve relatively complex sciences, codes and mathematical tasks.
Musk, along with Altman himself, helped to create an Openi as a non -profit organization in 2015.
However, in recent years the leadership of Musk and Openii were arguing. Musk recently led a group of investors that filed a proposal To buy a non -profit parent AI startup for $ 97.4 billion – Openi’s offer rejected.
Last month, Chinese start-up Deepsek shocked the AI market when he published a technical work that claimed that one of his open code models was able to submit the performance of Openi O1 models, despite using a cheaper, less energy intensive process.
He made a feat of the US -Liming AI chips Nvidia From the sale of its top GPU-a-which they are used to train AI models-do China.
Xai has “Colossus supercompute,“For the training of AI, which he said was using a rally of 100,000 advanced NVIDIA GPUs for AI training last year. On Tuesday, the company revealed that it doubled the size of its GPU GP for Training Grok 3.
While many AI and technologies experts have said CNBC This deepseek has increased the competition, showing what can be done with less advanced technology, others are skeptical of their influence.
