Meta Open Source AI Warning

Meta wants to open-source a GPT-5-level model and seems dead-set on open-sourcing right up until AGI. I want to be clear about what this means: There is no kill-switch. If something goes wrong—an agent gets out of control or a bad actor weaponizes it—there's no easy way to turn it off. It could be running on any small cluster. There will be no security. Safety research becomes meaningless. All the work people have done into making AI systems honest, aligned, ethical, etc becomes (mostly) moot. The population of AI's out in the world will evolve towards whichever systems produce the most economic output, irrespective of what values or motives they have. There will be no guardrails. Anyone can change their AI's values or capabilities as they want, for good or bad. If Meta continues to open-source as we get closer to AGI, it's pretty clear to me that things will become a shitshow. The arrival of these alien intelligences in the world is already going to be chaotic, but much much moreso if we just fling off what little levers of human control we have. As far as I can tell, Meta's wish to open-source stems mostly from some software industry dogma that "open-source good". And as far as I can tell, they weren't so in favor of open-source until their first model, Llama, was accidentally leaked, after which they pretended they were in favor of it the whole time. Overheard at a Meta GenAI social: "We have compute to train Llama 3 and 4. The plan is for Llama-3 to be as good as GPT-4." ...

Скриншот поста в социальной сети, где автор саркастически предупреждает об опасности открытого исходного кода для сверхмощных ИИ-моделей Meta. Шутка высмеивает драматичную реакцию на открытые модели и пародирует аргумент, будто только одна компания должна контролировать AGI; мем используют как реакцию на споры об open source AI и безопасности ИИ.

• @gentoomemes / 5980 2023-08-26