Yes, how much damage has Skynet done. I don't think talking about a work that published its first episode almost 40 years ago (1984) is spoiling, but four decades ago there was no artificial intelligence beyond small programs that learned a little and let your imagination run wild. This is how the figure of Skynet was conceived, a program that learned by itself and improved and improved until it became aware of itself, "scared" to learn that it was going to be turned off and end humanity as it was known by launching missiles between the US and Russia. . The most advanced that exists now, at least that is known, is ChatGPT4, and its improvements have come to scare or, at least, annoy.
But let's get down to a more realistic point. The concern of the signatories of the open letter, including Elon Musk, have other concerns, or one without a full form. He their fear is of the unknown, one of the most common fears, is that OpenAI and other foundations, programs and projects are improving their artificial intelligences as if it were a race to see who gets further and when, and this can give disparate results.
ChatGPT 4 in the spotlight
With version < 4 of the OpenAI chatbot, things seemed more “funny” or “cute”, use whichever word you prefer. They ask him things and he answers; you are asked to correct a section of code and you do so; you ask him for the meaning of a letter from a song or a movie and he explains it to you; but ChatGPT 4 has gone further. Your answers have improved a lot, and you can even generate content that could be false.
Because this is one of the concerns that is mentioned: let's imagine that I write a very detailed article on why man has never reached the moon. This article has many visits and is shared everywhere, although in some cases it is to criticize it. Later, we ask ChatGPT what happened at that moment, to write us an article about that event, it writes us something based on what I have written and, since it does not have the ability to detect fake news, write us something false. This, or with a better example, could flood the networks.
6 month break to study control
What is required is thatAll AI Labs pause for at least 6 months the training of AI systems more powerful than ChatGPT-4. This pause should be public and verifiable, and include all key players. If the pause cannot be enacted quickly, governments should step up and institute a moratorium.«.
All laboratories and independent experts should use this time to implement a set of security protocols, which translated could be taken as another way of saying that you have to write laws similar to the famous laws of robotics.
Isn't it the same for governments?
What strikes me personally is a point where he says that, in parallel, AI developers need to work with policy makers to accelerate the development of artificial intelligence for government systems. That is to say, help governments with their own artificial intelligence to be able, in a few words, to control artificial intelligences. Do what I say, not what I do.
That the letter was signed by Elon Musk does not surprise me. It has been revealed that he co-founded OpenAI, and it seems that, after leaving the company, his success is hurting. For the rest, yes, it is true that artificial intelligence, however intelligent they may be, in decision-making is said to be no better than a 7-year-old child, and that it may be difficult for them to differentiate true from false information. , so the point of fake news is a possibility and action must be taken. For everything else, I think that as long as they are not given the missile launch codes, things don't look so bad. I hope I'm not wrong.