OpenAI and Google DeepMind employees unite to expose the risks of rapid AI development

Former and current employees of Google DeepMind and OpenAI have shared a letter warning about the speed of development of models ofgenerative artificial intelligence. According to them, the financial motivations of suppliers prevent appropriate control to mitigate the risks of the technology.

Since the launch of ChatGPT in late 2022, the tech industry has been in a mad race for AI, with everyone doubling down on their efforts to deploy the technology and not get left behind. It is this haste that the authors of the open letter denounce, who ask, among other things, for better protection of whistleblowers in the context of AI.

Among the 11 signatories, we notably find Geoffrey Hinton, an eminent researcher in neural networks, who decided to leave Google in 2023 to warn of the consequences of uncontrolled development of artificial intelligence.

Signatories call for better protection for whistleblowers

They worry about “ low bonds » relying on companies to share information with governments about the capabilities and limitations of their systems. According to them, hiding this information has a certain financial interest, while the security systems put in place within companies are not sufficient. A few days ago, OpenAI formed a security committee on which CEO Sam Altman sits.

Researchers cite numerous risks related to AI, ranging from misinformation to worsening existing inequalities. “ The extinction of humanity » is mentioned in the missive.

In addition to the reluctance of companies to share the dangers of their models, the signatories regret the lack of protection for whistleblowers wishing to express themselves. “ Broad confidentiality agreements prevent us from expressing our concerns except to companies that may not be addressing these issues. Ordinary whistleblower protections are insufficient because they focus on illegal activities, while many of the risks we worry about are still unregulated », they write.

These comments are made despite OpenAI’s recent commitment to remove the non-disparagement clauses that the company made its employees sign, putting them at risk of losing their shares if they criticized the start-up.

The signatories call on AI companies to commit to various points: not entering into or implementing agreements prohibiting criticism of risks, setting up an anonymous procedure allowing current and former employees to raise concerns, supporting a culture of criticism and promise not to retaliate against those who share confidential information to raise whistleblowers “ after other procedures have failed “.

Lots of worries, but few actions

The publication of this letter comes after the recent departure of several executives from OpenAI. Ilya Sutskever and Jan Leike supervised the “ super-alignment » business models, a discipline whose objective is to ensure that they act in the interest of humanity.

“ We agree that rigorous debate is crucial given the importance of this technology, and we will continue to engage with governments, civil society and other communities around the world », Soberly reacted OpenAI to the publication of the letter.

This is not the first time that researchers have warned of the dangers of AI. As early as March 2023, thousands of experts, including Elon Musk, called for a 6-month interruption of research into generative AI. A few months later, eminent heads of Silicon Valley, including Sam Altman, estimated that AI would potentially be as destructive as nuclear war…

Sources: Reuters, CNBC