Something has shifted at OpenAI.

If this is the way to superintelligence, it remains a bizarre one. “This is back to a million monkeys typing for a million years generating the works of Shakespeare,” Emily Bender told me. But OpenAI’s technology effectively crunches those years down to seconds. A company blog boasts that an o1 model scored better than most humans on a recent coding test that allowed participants to submit 50 possible solutions to each problem—but only when o1 was allowed 10,000 submissions instead. No human could come up with that many possibilities in a reasonable length of time, which is exactly the point. To OpenAI, unlimited time and resources are an advantage that its hardware-grounded models have over biology. Not even two weeks after the launch of the o1 preview, the start-up presented plans to build data centers that would each require the power generated by approximately five large nuclear reactors, enough for almost 3 million homes.

https://archive.is/xUJMG

    • 14th_cylon@lemm.eeEnglish
      31·
      3 days ago

      sure. 100% means something will happen every single time in the observed set. if something does not happen every single time, then it is not 100%.

      this will not happen every single time. among all possible results, there will be results where none of the monkeys start any kind of shakespeare. there will be instances where every single work they start will be just the paper full of letter “a”. or something else than shakespeare. as you add monkeys (approach the infinity) the smaller such chancegets, until it gets extremely unlikely, but it is not going to be zero.

      imagine you are throwing a 6 sided dice hundred times and i ask you - is it possible there will be no 6 among those one hundred throws?

      anyone who passed some basic math understand it is indeed unlikely, but it is not impossible. if you keep throwing long enough, there will be cases with zero 6s in it.

      probability of that happening is (5/6)^100, which is 1,2 x 10^-8, eg it will happen roughly 1,2 times in ten million cases. not likely, but not impossible.

      in 1000 dice throws, the chance drops to (5/6)^1000, roughly 6,6 x 10^(-80), or 6,6 in 100 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 cases.

      fun fact: the above number (1 with 80 zeros) is called One Hundred Quinvigintillion (had to google that indeed).

      if you further increase the number of throws in the series, the chance of not having single 6 will be getting even smaller, but never zero.

      or in other word, if you raise 5/6 to any positive number, the resulting number is always positive number.

      https://www.wolframalpha.com/input?i=%285%2F6%29%5En

      for further study, the relevant concept here is limit of the function

      we say that limit of the above function is zero, which means it will approach the zero really close (infinitely close), but will never reach it.

      • deranger@sh.itjust.worksEnglish
        2·
        3 days ago

        Thanks for elaborating. I knew limits were going to show up.

        You make a good point, although I would like to point out that one hundred quinvigintillion is basically right next to the number 1 on the number line that goes to infinity. The chance of the monkeys not writing Shakespeare is infinitesimally small. You winning every possible lottery every day for the rest of your life is infinitely more probable than the monkeys not writing Shakespeare.