• 5 Posts
  • 79 Comments
Joined 6 months ago
cake
Cake day: June 11th, 2024

help-circle
  • I think you (or I) misunderstand something. You have a test for a small well defined unit like a C function. und let the AI generate code until the test passes. The unit test is binary, either it passes or not. The unit test only looks at the result after running the unit with different inputs, it does not “go through millions of lines of code”.

    And you keep doing that for every unit.

    The writing of the code is a fairly mechanical thing at this point because the design has been done in detail before by the human.



  • Jeena@piefed.jeena.nettoTechnology@lemmy.worldThe GPT Era Is Already Ending
    link
    fedilink
    English
    arrow-up
    11
    arrow-down
    4
    ·
    4 days ago

    We already have to do that as humans in many industries like automobile, aviation, medicine, etc.

    We have several layers of tests:

    1. Unit test
    2. Component test
    3. Integration / API test
    4. Subsystem test
    5. System test

    On each level we test the code against the requirements and architecture documentation. It’s a huge amount of work.

    In automotive we have several standard processes which need to be followed during development like ASPICE and ISO26262:





  • I just bought a random UPS at MediaMarkt back in the day in Poland and calculated that it would be able to power all the CCTV cams, the CCTV recorder, a raspberry pi and the modem which is connected to a long range WiFi antenna for at least half an hour. This worked very well for a couple of years until the battery gave up. The one I had had a ethernet port but I never bothered to set it up to send the signal.

    Mine was running at my parents summer house in Poland while my parents live in Germany and I in Sweden and now in Korea, so if something breaks down it’s down for up to a year until someone goes there to fix it.

    Right now everything is down, my dad was there a couple of month ago and said that a marten chew up the Ethernet cables. Sadly my dad couldn’t fix it so now I hope I will be able to get there during Christmas.



  • I did it for a year or so, exactly the same thing, I have a bunch of websites and other services like matrix, mastodon, peertube running on that server.

    I found out the hard way that Lemmy is really not optimized for single user instances on a small VPS. In the beginning it was OK but the more they optimized for big instances like Lemmy.world and .ml the word it ran with few resources.

    In the end it was so bad that it would Hogg all the servers resources and bring the whole server down every couple of hour so that I had to hard reboot it.

    The worst part was that this impacted alł my other services and websites which went down with it every couple of hours. I tried to get a beefier server but that only helped until the next update.

    In the end I switched to Piefed and couldn’t be happier. It uses a lot less resources as a single user instance so I could downgrade the server again and it’s written in python so I can much easier help working on it, no need to compile anything, etc. And it has a cool template engine which let’s you easily modify how it looks and feels.




  • I’m running an instance and as a software it’s OK. But you never know if your subscription tab has nothing new because there is a bug in the ActivityPub code or just because nobody is posting on peertube.

    Another thing is that most of my video watching I do on the TV but there is no app for android on the TV which give you a UI like the many apps streaming providers offer. So at least for me it’s limited to very seldom watching something on the laptop when I remember.

    Discover ability is another problem, similar to Mastodon it’s difficult to find new people to watch. That is solved much better on Lemmy with the communities.


  • The difference is the intent and the background behind it.

    Sure for maximum mass adoption the computer can out-research any human and just find the blandest set of rules which cater to the highest percentage of the majority.

    What it still will have a hard time doing, and I predict it will be for quite some time - probably until we have quantum computers - is to come up with a new way of doing poetry which is not just copying what humans did but better.

    I think of AI like it’s China, they are super efficient in copeing things and gradually making them better and cheaper but the setup of their society makes it impossible to really innovate.

    And yeah I’m saying that it’s the setup, because in Taiwan they are able to innovate at a much higher rate.