Meet the scientist behind Altman’s fall on ChatGPT – 11/19/2023 – Tech

All information indicates that the person responsible for the maneuver that ousted Sam Altman from his position as chief executive of OpenAI —startup developer of ChatGPT, OpenAI— was Ilya Sutskever, 37, the company’s chief scientist. Sutskever led the advances in artificial intelligence necessary to shape the most accessed chatbot on the internet.

Sutskever was the talent needed to get OpenAI off the ground, according to Elon Musk, who participated in the founding of the then-nonprofit entity, along with its chief scientist, Sam Altman and former board chairman Greg Brockman, who removed resigned from his position after the resignation of the former chief executive.

To do this, Musk had to convince Sutskever to leave Google’s artificial intelligence arm. At the time, Google brought together around two-thirds of the world’s artificial intelligence researchers and had almost unlimited money to develop the technology.

The owner of X, formerly of Twitter, said it was one of the most difficult recruitments of his entire career, in an interview with Fortune magazine. One of the arguments used by Musk was a supposed lack of security and responsibility perspective in the development of AI at Google at the time.

As Ronaldo Lemos, columnist for Sheetapparently, the rush to develop products based on artificial intelligence without the caution considered appropriate on the part of the team was the main reason for Altman’s dismissal.

There are rumors that the new products launched on the 6th had security flaws, including the possibility of leaking user data. This would have led Microsoft itself, OpenAI’s main investor, to suspend the internal use of the company’s products.

It was Sutskever, in person, who called Altman to notify him of the termination video conference. All the board members were present, except the then president of OpenAI, Greg Brockman. Investors, including the startup’s main backer, Microsoft, were not notified until the public announcement.

The then technology director Mira Murati took command of OpenAI on an interim basis.

The movement that ousted the then chief executive started a struggle. People close to Altman and Brockman say the two have plans to found a new artificial intelligence company and would be able to bring talent from OpenAI itself to the new business.

In addition, OpenAI investors mobilized to return Altman to command the company. According to the specialized website Verge, Altman calls for drastic changes in the company’s governance to avoid going through the same setback.

OpenAI is governed by a non-profit entity, which in turn controls a for-profit subsidiary. This is the one that has investors and generates money. If that changed, Sutskever and researchers who prioritize safety and responsibility would lose out in the split.

Why is Sutskever critical to OpenAI?

Canadian-Israeli, born in Russia, Sutskever began studying artificial intelligence and machine learning models at the age of 17 under the guidance of Geoffrey Hinton, a man known as the godfather of AI for conceptualizing the current form of artificial intelligence.

As a result, Sutskever has long been at the frontier of knowledge about so-called deep learning, a category of artificial intelligence in which data is processed with inspiration from the human brain. The structure of the algorithm relies on the functioning of a neuron, hence the name “neural network.”

Sutskever was the one who coined the hypothesis that a large-scale neural network with a large database could perform any task possible for a human being.

In partnership with Hinton and fellow artificial intelligence theorist Alex Krizhevsky, Sutskever developed AlexNet, an artificial intelligence capable of recognizing images and classifying tasks that surprised the world in 2012. The advance, at the time, was such that Google spent US$44 million (around R$200 million) to acquire a company founded by Hinton and his two students.

After leaving Google at Musk’s invitation, already at OpenAI, Sutskever realized that, if an artificial intelligence model learned to predict the next word in a given text, it could start training itself.

This was the essential advance in creating generative artificial intelligence, such as ChatGPT itself, which generates text.