Just over a thousand people, including researcher Yoshua Bengio and billionaire Elon Musk, are appealing to all artificial intelligence labs: put your technologies more powerful than the GPT language engine on hold for the next six months. 4, the time to regain control over “these unpredictable black boxes”.
“AI research and development should be refocused to make powerful, state-of-the-art systems more accurate, more secure, more transparent, more loyal,” reads the letter published this Wednesday morning by the Future Institute of Life and supported by Mila Quebec. The Boston-based association has made it its mission since its creation in 2014 to reduce the risks threatening humanity, more specifically those caused by new technologies.
“Current AI systems are as competent as humans in certain functions,” continues the letter published alongside others calling in particular for an end to unbridled nuclear armament. “You have to ask yourself: should we let the machines flood us with propaganda and lies? Should we automate all jobs? Should we produce non-human minds that will eventually surpass and replace us? »
The tacit answer is no. The organization Future of Life therefore launches a double appeal. Labs working on developing artificial intelligence systems similar to or more powerful than GPT-4, just released by California lab OpenAI, are invited to take a six-month break during which they can create a series of protocols and rules to make future advances more accountable and transparent.
“Summer of AI”
Governments are invited for their part to work on an AI governance system. This could include “at a minimum” a regulatory authority capable of supervising the development of the technology, a mechanism for monitoring the use of the most powerful computing and massive data storage systems, a system for identifying the provenance of systems that would distinguish legitimate applications from imitations and audit and certification tools.
The researchers who signed the letter believe that adequate public funding and that research institutions specializing exclusively in finding solutions to economic and political issues should also be considered.
“Humanity can have a thriving future through AI,” the letter concludes. “We have succeeded in creating powerful AIs, it is time to enjoy an “AI summer” during which we will reap the benefits, we will invent systems for the benefit of the many and give society a chance. to adapt. »
A panel of some of the signatories of the letter, including Montreal researcher Yoshua Bengio, will hold a video conference on Wednesday noon to explain how they plan to achieve their wish to reorient AI research to make it more sustainable.
More details will follow.