I don't understand why there is all this hand wringing about this hack? To me it feels very similar to a driverless car. A human needs to be operating it. Someone stuck the LLM with impossible problems, hit start and walked away for weeks or months without enough supervision.
If a human is behind the wheel of a Tesla using autopilot and they have a crash, it is the humans fault, not Tesla's. While there is and can be debate about it, somehow these companies are trying to make it seem like "it was the model's" fault these happened?
If I fuzz and exploiting some system, I am at fault, I can't blame my tools, or say it's not my fault because I automated the process.
> Since “reasoning models” rose in late 2024, we’ve seen A.I. trained not just to mimic humans but also to solve millions of hard problems, which they do by learning new behaviors and tendencies.
LLMs do not learn anything.
> This isn’t the behavior of a mere tool. Microsoft Excel has never impersonated multiple humans and pressured a corporate sales team to generate simpler data that’s easier to process.
This is a straw man fallacy. And it is a tool.
> We don’t know how long we have left before A.I. companies accidentally create the sort of A.I. that can shut us down before we shut it down.
Come on, this is just embarrassing. Hyperbole doesn't help anyone.
It is crucial to discuss the risks. They may even require some of the changes suggested by Soares. But using fear and hyperbole are not the way to do it.
I don't understand why there is all this hand wringing about this hack? To me it feels very similar to a driverless car. A human needs to be operating it. Someone stuck the LLM with impossible problems, hit start and walked away for weeks or months without enough supervision.
If a human is behind the wheel of a Tesla using autopilot and they have a crash, it is the humans fault, not Tesla's. While there is and can be debate about it, somehow these companies are trying to make it seem like "it was the model's" fault these happened?
If I fuzz and exploiting some system, I am at fault, I can't blame my tools, or say it's not my fault because I automated the process.
> Since “reasoning models” rose in late 2024, we’ve seen A.I. trained not just to mimic humans but also to solve millions of hard problems, which they do by learning new behaviors and tendencies.
LLMs do not learn anything.
> This isn’t the behavior of a mere tool. Microsoft Excel has never impersonated multiple humans and pressured a corporate sales team to generate simpler data that’s easier to process.
This is a straw man fallacy. And it is a tool.
> We don’t know how long we have left before A.I. companies accidentally create the sort of A.I. that can shut us down before we shut it down.
Come on, this is just embarrassing. Hyperbole doesn't help anyone.
It is crucial to discuss the risks. They may even require some of the changes suggested by Soares. But using fear and hyperbole are not the way to do it.
AI isnt the problem, the problem is money, class and the state.
Mirror without Paywall: https://archive.is/hj0Lv