untitled design

Sam Altman plans to create AGI

OpenAI CEO Sam Altman in an interview Financial Times announced its intention to attract additional funding from Microsoft to create artificial general intelligence (AGI).

According to him, the partnership with the main investor, in particular with the corporation’s CEO Satya Nadella, “is working very well.”

Earlier this year, Microsoft invested $10 billion in OpenAI as part of a “multi-year” agreement. The San Francisco company was valued at $29 billion.

Altman noted the project’s “revenue growth” without disclosing financial details. However, he explained that the company still remains unprofitable due to the high costs of training language models.

AGI is AI-based software that has human-level intelligence.

“People say, ‘You’ve got this research lab, you’ve got this API, you’ve got this partnership with Microsoft, you’ve got this ChatGPT thing, and now the GPT store.’ But in reality these are not our products. These are channels to our only product – artificial intelligence,” Altman added.

Now OpenAI faces two tasks: exploring options for creating “superintelligence” and increasing the computing power for this process. According to Altman, ensuring the safety of AGI is an important issue.

Developers are also working on creating “autonomous agents” that can perform a number of tasks or actions, such as making payments or sending emails.

In addition, Altman spoke about the next generation of artificial intelligence model – GPT-5. He did not give a specific launch date for the product, but said that training AI will require a much larger amount of data.

Although the neural network will be more technically complex than its predecessors, it is difficult to say exactly what new capabilities AI will gain, the head of OpenAI added.

“Until we start training this model, it will be like a funny guessing game for us. We’re trying to get better because from a security perspective, it’s important to anticipate opportunities. But I can’t say what exactly GPT-5 will do that GPT-4 didn’t do,” Altman shared.

Like most AI companies, OpenAI uses Nvidia H100 chips to process training data. At the same time, the head of the company emphasized that there has been a “severe” shortage of them all year.

In his opinion, the dependence on Nvidia is unlikely to last long, as other players – Google, Microsoft, AMD and Intel – are preparing to release competing AI chips.

Altman believes that language is “a great way to compress information” and therefore develop intelligence. However, this factor is missed by many big players like Google Deepmind.

“There was a long period when [Исаака] Newton’s correct approach was to read more mathematics textbooks, talk to professors, and solve problems. This is what our current models do,” the company’s CEO gave an analogy.

In early November, OpenAI held a presentation at which it showed turbo versions of ChatGPT and introduced an assistant for developers.

Source: Cryptocurrency

You may also like

Get the latest

Stay Informed: Get the Latest Updates and Insights

 

Most popular