Demis Hassabis, co-founder and CEO of Google Deepmind, held a meeting at Mobile World Congress (MWC), the largest annual rally in the telecommunications industry, held on February 26, 2024 in Barcelona. Masu.
Pau Barrena | AFP | Getty Images
PARIS – Deepseek’s AI model is “probably the best job” from China, Google Deepmind CEO Demis Hassabis said on Sunday, but added that the company does not show new scientific advances .
Last month, China’s Deepseek released a research paper that rattles the global market after claiming that AI models were too highly trained as part of the cost of major AI players. nvidia Tip.
The Deepseek announcement has led to aggressive stock sales and considerable debate over whether large tech companies are too many in AI infrastructure.
Hassavis praised Deepshek’s model as “impressive work.”
“I think it’s probably the best job to come out of China,” Hassavis said at a Google-hosted event in Paris ahead of the city’s hosted AI Action Summit.
The CEO of Deepmind said the AI model shows DeepSeek can do “very good engineering” and it shows “changing things on a geopolitical scale.”
But from a technology standpoint, Hassavis said it wasn’t a major change.
“In spite of the hype, there are no real new scientific advances. It uses known techniques (in AI),” he said.
The CEO of Deepmind said the company’s Gemini 2.0 Flash model, released by Google this week, is more efficient than the DeepMind model.
Deepseek’s claims about the low cost and the chips it uses have been questioned by experts who believe that the development costs of Chinese companies are high.
Agi 5 years later
The world of AI has been debated for years as artificial general information or the arrival of AGIs. AGIs widely specify AIs that are smarter than humans.
Hassavis said the AI industry is “on the path to AGI,” and described it as “a system that demonstrates all the cognitive abilities that humans have.”
“I think we’re getting closer now. Perhaps we’re alone, maybe five years or rather off the extraordinary system,” Hassavis said.
“And I think society needs to prepare for it. And you make sure we are profiting from it, and that society as a whole will benefit from it, but , it also reduces some of the risks.”
Hassavis’ comments reflect comments from others in the industry that suggest that AGI could be closer to reality.
This year, Openai CEO Sam Altman said, “I am confident that I know how to build an AGI, as we traditionally understood.”
Still, many in the industry flag multiple risks associated with AGI. One of the biggest concerns is losing control over human-created systems. This is a view shared by renowned AI scientists Max Tegmark and Yoshua Bengio.
