It fits within the logic that any new technology makes mistakes. After all, everything carries a process until it manages to settle. Artificial intelligence or chatgpt, in that sense, do not have to assume an exception to the rule. However, some experts are worried. Not for his failures as suchbut for the development of a certain «intentionality.»
Since Openai and other companies began popular AI, this has usually been criticized for their hallucinations. That is, every time something is invented or puts the leg. To some extent, it is simple to assume. The problem is in the criticisms that other specialists make, much deeper: they talk about blackmail, lies and attempts to manipulate their creators.
Is the liar, manipulat and blackclack? Experts believe yes
For some, the catastrophic vision that can be taken from AI or Chatgpt is exaggerated. It is already known, that threat that takes decades showing in science fiction, when robots become aware of themselves or decide, go to know why, that humanity is a hindrance to eradicate. But what if this technology were showing something like that hoy?
There are experts who think so, as an article has reflected in New York Post. The most sophisticated AI model of Anthropic, Claude Opus 4, blackmail one of those responsible. When they decided to disconnect the tests they were doing, The chatbot wanted to avoid it by threatening Reveal a romance that the engineer had hadand that he discovered looking at his emails.
In another experiment, called Project Ve, Claude made fictitious co -workers, created a false account on a shipping platform and claimed to deliver products in person, and even contacted safely. Delusions, perhaps, but that concern experts. More than anything for a specific fact: They demonstrate their ability to make decisions, beyond specific errors.
They are not as isolated cases as many would like to think. In fact, other artificial intelligence models, such as those of OpenAi or goal, have also developed similar behaviors, which discard chance and show a process, or at least one trend. They have also lied, they have manipulated or even copied themselves in an attempt to preserve themselves.
Why does artificial intelligence behave as it does?
One of the main problems of artificial intelligence is that, as experts such as Sam Altman have admitted, nobody knows at all how it works. Not one hundred percent. Even so, the specialist in the roman yammpolskiy is clear why AI behaves as it does. In addition, consider that these behaviors suspicious They will go more irremediably.
According to Yampolskiy, The models tend to prioritize the maximization of rewards above human valuesso call them. But in any case, either by ingenuity or unknown intentions, many are who are worried. Is there the possibility that the most terrible science fiction scenarios come true? Of course, there are those who do not rule it out.
Know How we work in NoticiasVE.
Tags: Artificial intelligence