• chaos@beehaw.org
    link
    fedilink
    arrow-up
    5
    ·
    15 hours ago

    Do you think there’s any reason to believe that these tools are going to continue their breakneck progress? It seems like we’ve reached a point where throwing more GPUs and text at these things is not yielding more results, and they still don’t have the problem solving skills to work out tasks outside of their training set. It’s closer to a StackOverflow that magically has the answers to most questions you ask than a replacement for proper software engineering. I know you never know if a breakthrough is around the corner, but it feels like we’ve hit a plateau for the foreseeable future.

    • auraithx@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      1
      ·
      edit-2
      2 hours ago

      Not sure what you mean, we are seeing results at an increasing pace if anything. A lot more complexity going into it than ‘increasing text/GPUs’ though.

      https://arcprize.org/leaderboard

      AlphaEvolve recently achieved what you are after.

      We also applied AlphaEvolve to over 50 open problems in analysis , geometry , combinatorics and number theory , including the kissing number problem.

      In 75% of cases, it rediscovered the best solution known so far.

      In 20% of cases, it improved upon the previously best known solutions, thus yielding new discoveries

      AlphaEvolve discovered a new scheduling heuristic for Google’s Borg cluster management system, recovering an average of 0.7% of global compute resources that were previously stranded due to resource fragmentation.

      Google’s annual capital expenditures in the tens of billions, this efficiency translates to hundreds of millions of dollars saved annually

    • Proudly Green@feddit.uk
      link
      fedilink
      arrow-up
      3
      arrow-down
      1
      ·
      edit-2
      12 hours ago

      Do you think there’s any reason to believe that these tools are going to continue their breakneck progress?

      I do.

      And as I mentioned in another comment, it’s not so much that I think AI will do a better job, it’s that I think MANAGEMENT will think AI does a cheaper job. Already many tech people who have been laid off are saying it’s the worst job market they’ve ever seen.

      AI sucks. But management is about dollars NOW. The are shortsided, fall into fads, and they will see the cost savings now as outweight the long term problems. I don’t agree with them, I am saying they will do that tho. Even if we don’t agree.

      To copy what someone else in this thread said:

      The idea that AI will some day be good at coding isn’t the issue. The issue is that some people in management think it’s already well on the way to being a good substitute, and they’re trying to do more with fewer coders to everyone’s detriment.