“I asked how I would open this audition. And you just heard the result.” The voice heard was the same as that of Senator Richard Blumenthal, who chairs the US Senate’s privacy subcommittee, but it was not actually his. Ideas and words, although recognizable and respected, are not really so either.
The voice? Synthesized by Artificial Intelligence (AI) from the voice and speeches of the senator.
The text? Written by ChatGPT from the question: “How would I open the proceedings for a discussion about the future of AI regulation before a judicial subcommittee based on my previous speeches in Congress?”
In less than five minutes he had transmitted a message: at the age of 77, he made his voice heard in a room -without opening his mouth- while he transmitted his ideas -without having them written down- about a world for which they were trying to find rules, but of which who have already benefited.
The Senate is concerned about the AI revolution and sat down this Tuesday with three specialists to try to understand what the dangers of this new world are and how far they go, but also how the new powers that are spreading can be controlled.
The initial theatrical blow carried out by Blumenthal – or rather, by what the AI built from his voice and his ideas – had already left the room restless. In fact, he was left uneasy to the point of commenting: “This may have seemed funny to you, but what if [a IA] supported the surrender of Ukraine?
The room full of senators, mostly over the age of 60, quickly assumed that they knew little to nothing about AI. So they called a senior IBM official, Christina Montgomery, a university professor, Gary Marcus, and the CEO of OpenAI, ChatGPT’s “parent company,” Sam Waltman.
It is on the latter that attention falls, especially since it promised a technology capable of doing what humanity has not yet achieved on its own.
“We are working on tools that one day can respond to humanity’s biggest challenges,” he says, “like climate change and a cure for cancer.”
It is true, he acknowledges, that the current systems “still do not work, but it has been extraordinary to see so many people around the world, benefiting from what can already be done.”
The risks, these, are as great as the feats the future may bring, and AI can be used, he acknowledges, to rig elections, to revolutionize the world of work, and can even be used to start a war.
Waltman responded without much thought to questions from another senator, Kennedy. He wants to know if, by giving coordinates to a drone flying over, for example, a target or a car, “artificial intelligence can create a situation where a drone can select a target. Can this be done?”
Confirmation came shortly: “Of course,” Waltman said. But not without remembering that AI is a tool and not a creature. And as the creator of ChatGPT, he has no problem recognizing that you have to create a set of rules. He has three suggestions.
The first is to create an “agency to license some areas of activity and which can also withdraw this license if the rules are not followed.”
The second is to “define a list of security procedures to assess the degree of risk and define the steps that must be taken before launching any model” of AI.
And the third is the performance of “independent audits, by experts, who can say whether or not this model conforms to the standards,” he defined.
At the suggestions, the space for the invitation arose: can the man from ChatGPT accept the task of designing the AI regulator?
The answer is no: “I love my current job.”
Source: TSF
