OpenAI worked on ‘super powerful’ AI before firing CEO

Before Sam Altman’s dismissal, OpenAI was working on artificial intelligence (AI) powerful enough to worry the company. Called Q* (pronounced “Q-Star”), the model will be capable of solving never-before-seen mathematical problems, something that would mark a breakthrough in technology. Information taken from Reuters.

The concern was expressed in a letter sent to OpenAI’s Board of Directors shortly before the dismissal of Sam Altman, now the company’s CEO. This will be one of the reasons why the executive is removed from office.

How dangerous is Q*?

Currently, language models (LLMs) are capable of formulating phrases and sentences based on probabilistic models. It’s as if AI, with its incredible processing power, can only deal with questions that have been presented to it before.

According to Reuters sources, Q* has overcome this limitation. The platform has been proven to be able to solve elementary math problems, unlike a calculator that can perform a limited number of operations.

Experts point out that the Q* project will be an important step towards artificial general intelligence (AGI), an autonomous system smarter than humans. Don’t talk at length, artificial intelligence will have cognitive capacity similar to that of humans.

Thus, the model will not only have access to previously provided knowledge, but will also be able to generalize, learn and understand an infinite number of mathematical solutions.

Sam Altman used to write about himself The evolution of Q* at a conference. “Oh [descoberta] The last one was just in the last few weeks, I was able to be in the room as we pushed back the curtain of ignorance and the frontier of discovery forward, and to be able to do that is the professional honor of a lifetime. “He mentioned it during the incident. He was fired a day later.

What are the dangers of Q*?

In the letter, inspectors would warn the council about the dangers of this new project, but would not specify what they were. Reuters did not have access to the aforementioned letter and was unable to obtain a response from OpenAI. It is therefore not possible to guarantee the content of the message or the true potential of the tool.