The «ultimate risk.» This is how Jared Kaplan, chief scientist at Anthropic, refers to the moment when artificial intelligence begins to train itself. A point of no return that, at least in his opinion, will be inevitable. What could happen then? The AI expert does not know for sure, but his predictions are disturbing to say the least.
He himself recognizes it: «No one knows how it will end,» he declared, referring to the evolution of this technology. to which more and more specialists refer. Only one thing is clear: «It will be terrifying.» Mainly, he assures, because it will not happen in the distant future, but rather, according to his predictions, it is just around the corner.
The disturbing evolution of AI, according to the head of Anthropic
To some extent, it is very understandable that there are people afraid of AI. Not only because of what has traditionally been seen in fiction, with machines that tried to wipe out humanity and those types of scenarios typical of novels and movies. Today, there are already many companies that are laying off workers en masse. But Anthropic’s Jared Kaplan goes further.
The first thing he has done, in recent statements given to The Guardian, is to set a date for when the AI begins to train itself. He thinks it will happen sometime between 2027 and 2030. There are many experts who agree in this last year. That would be when artificial intelligence created another, better artificial intelligence.
The big question is, what will happen then? For Kaplan, there are two possibilities. The first, that the result is extremely beneficial, as people like Bill Gates or Sam Altman predict. But the opposite can also happen: that human beings lose control over this technology. If this were to happen, he insists, it is impossible to know how everything will end.
Another of the biggest concerns of the head of Anthropic is that this artificial superintelligence would fall into the wrong hands. If today the number of scams, attacks on the security of numerous entities, etc., have already multiplied, in the future these risks could be incalculable. «I think preventing the misuse of technology is very important,» he warns.
The best intelligence in history
Throughout his review of the current state and future of artificial intelligence, Jared Kaplan raises questions that are as interesting as they are uncertain. Are AI good for humanity? Are they really useful? And above all: Are you going to allow people to continue having control over their lives and the world?
Finally, the Anthropic scientist makes another reflection to take into account. Everyone thinks that AI will continue to improve, in some way, but what if it doesn’t? Kaplan does not rule out that we are wrong and that the best possible AI is simply the one we know now.