4 ms·

Has an AI already been used to improve itself, or create another AI? I think that's going to be a very near inflection point where we'll see exponential improv

• ilaksh 3y ago
  It is something that we need to outlaw. All it takes for that to be dangerous is for someone or some groups in competition to give it an overly broad goal like "take control". But also people are stupidly trying to find ways to emulate life in these systems (such as reproductive and self-preservation instincts). Combine that with hyperspeed or hyperIQ AI and you just have to hope to stay on its good side because you cannot possibly keep up with the speed and effectiveness of those systems.
  • fragmede 3y ago
    I think if we have to rely on humans to not glue an AI to itself because it's merely against the law, that we will find ourselves on the losing side real quickly. if you told me there is this piece of code that will make me money on the stock market but it is illegal to run, I can't say that I wouldn't run it.
    • ilaksh 3y ago
      True but we should at least try to actively discourage it and educate people about existential risk. These will be so far beyond the scope of stock market trading. My main thought on this stuff that I have been repeating is that the only real mitigation is to prohibit the manufacture of certain levels of AI hardware. Again, hard to enforce, but you have to try. And not having the hardware exist will be more effective than laws. If we make an effort then we might be able to hold off the post-human era for a few generations. By post-human I don't mean that (necessarily) normal humans have been killed off, just that they don't control the planet anymore.
• TeMPOraL 3y ago
  > Has an AI already been used to improve itself, or create another AI? To the extent OSS community uses GPT-3 and GPT-4 to generate training data for training open models, AI is indeed being used to improve itself. Runaway self-improving AI doesn't need to be - and is, initially, unlikely to be - a single instance of a model improving itself at runtime. AI models being used to directly or indirectly increase capabilities of other AI models fits the bill too.
  • bsder 3y ago
    Ever see what happens when you compress an image over and over?
    • TeMPOraL 3y ago
      Yes. But I'm not talking about looping on that in particular. GPT-sourced data is just fine if you're trying to cheaply bootstrap LLM of a different family/architecture that is meant to compete with GPT, and need something to cover for training data you can't otherwise afford to get. It's a one-time deal, but ten GP was askind: > Has an AI already been used to improve itself, or create another AI? And IMO this fits the bill.
• IshKebab 3y ago
  No. The GPTs are very impressive but they're quite far from being clever enough for that, and they're not great at coming up with new ideas either. They're much better at remixing existing ideas.
  • TeMPOraL 3y ago
    > coming up with new ideas > remixing existing ideas Those two are the same thing. That's how we do it too - we come up with new ideas through remixing ideas we've learned.
    • IshKebab 3y ago
      A lot of new ideas are remixes but plenty aren't. They involve a novel creation. Think about something as simple as the wheel. Stone age ChatGPT probably couldn't come up with that because nobody had talked about it before and it doesn't do well at thinking of totally new things.
      • TeMPOraL 3y ago
        Wheel was incremental too. Plenty of fruits and stones have spherical shapes. If you cut a fruit, or a tree, or a trunk, into slices, you get wheels. Early humans had plenty of opportunities to observe and play with wheel-shaped things.