Not surprising, i remember hearing of data from companies that increased efficiencies in processes expecting to give their employees more free time. But the end result was that all free time was used up by new tasks. I could easily imagine how if half of the BS tasks i had to do magically went away how i would actually probably end up busier because i would have more time to manage my projects in a more direct way. Which honestly would both be great and really annoying
Looks like the position AI can fill best is “idiot coworker whose best contribution to overall productivity is to call in sick”.
As much as I hate for Musk to have any kind of a win, I think this is probably still good news overall. It was ridiculous that OpenAI (or any company, really) would have ever been allowed to restructure from a non-profit to a for-profit entity. There’s a good reason that the California AG was investigating this proposal. But this is also good because I want OpenAI to crash and burn before they get any bigger and do more planet-burning shit than they’ve already done. SoftBank’s much-touted $40B investment (which OpenAI needs badly) was partially contingent on them converting to a for-profit, so interesting times are ahead!
Definitely looking forward to seeing Ed Zitron’s take because he understands this way better than I do.
Looks like there were a quite a lot of pictures of espresso machines in the training data.
So their in-house testing already identified the update as being a problem, but they went ahead and released the update anyway. Now they’re saying that in the future they’re “formally considering” having the potential to block updates if issues are found? Sounds like some top-notch, confidence-inspiring quality control.
As for the sycophantic behavior of ChatGPT, maybe that explains a lot about what all those AI executives are up to. You know that they’re constantly getting validation from their own product telling them how awesome they are at their jobs and what great businesses decisions they’re making.