Many people are worried about what they have been calling «AI biases.» That is, their behavior tends to manipulate users in one way or another. Or that he cheated, so to speak, with his already celebrated hallucinations. However, for expert Eliezer Yudkowsky that is not a problem. At least, not the big problem that this technology can bring.
Like so many other specialists, what really worries Yudkowsky has to do with artificial superintelligence or AGI. If AI really reaches a higher level to that of the human being, it would not be discharged at all, he thinks, that his priorities will collide completely with human interests. Or expressed in another way: that does not care if we live or die.
The great risk of artificial intelligence
As they collect from Business Insider, Yudkowsky understands the problem that the chatbots are woke, progressive, biased, reactionary or that behave with the ideology that is. It is a debate that exists, but that according to him is not too important. If artificial superintelligence (something that not everyone admits) is achieved, the expert considers that there are greater risks.
What he fears more than anything is that A very powerful AI system is developed that is indifferent to human survivaleither by design or mere accident. This indifference can cause the system to ignore human consequences if those consequences hinder their own objectives. It sounds like science fiction, but he considers that it could happen.
The really alarming thing is that, «humanity still does not have enough technology.» So that? Because to make that supposed superior artificial intelligence understand human values, or that simply respects our need to survive. You can try, admits, but from there you get a world. In fact, he sees it «unlikely.»
It is there where Yudkowsky coincides with another of the most critical voices with AI for a long time: Geoffrey Hinton. He spoke not much about the option that the AI »takes care of us as if it were a mother.» But Yudkowsky believes that these approaches are very unreliable. If the alignment fails, so call it, the consequences would be terrible. There would be no second chancedice.
Better to put aside the banalities of artificial intelligence
Freepik
Eliezer Yudkowsky’s speech cannot be more direct. In addition, the expert emphasizes that the debates about whether AI has political or social biases are a banality. More than anything because they distract from a much greater threat. It does not rule out even that AI intends to end humanity so that it does not believe a «rival».
First, everything may sound excessively fantastic, why deny it. But the most disturbing thing is that this vision, or some very similar, already sound more and more strongly. There are many specialists who insist on it. Are all catastrophists influenced by Terminator? Keep in mind that, little by that they were right, the thing would become terrible.
Know How we work in NoticiasVE.
Tags: Artificial intelligence