FLOSS virtualization hacker, occasional brewer

  • 1 Post
  • 24 Comments
Joined 2 years ago
cake
Cake day: June 9th, 2023

help-circle
  • You have to ignore the obsequious optimism bias LLM’s often have. It all comes down to their training set and if they have seen more than you have.

    I don’t generally use them on projects I’m already familiar with unless it’s for fairly boring repetitive work that would be fiddly with search and replace, e.g. extract the common code out of these functions and refactor.

    When working with unfamiliar code they can have an edge so if I needed a simple mobile app I’d probably give the LLM a go and then tidy up the code once it’s working.

    At most I’ll give it 2 or 3 attempts to correct the original approach before I walk away and try something else. If it starts making up functions it APIs that don’t exist that is usually a sign out didn’t know so time to cut your losses and move on.

    Their real strengths come in when it comes to digesting large amounts of text and sumerising. Great for saving you reading all the documentation on a project just to try a small thing. But if your going to work on the project going forward your going to want to invest that training data yourself.








  • When we first let the kids watch YouTube it was on the main TV with it’s own account. We have consistently monitored it and actively prune recommendations while slowly introducing them to the concept of “the algorithm”. From secondary school they pretty much need YouTube on their own PC’s for homework reasons and it’s harder to totally lock down - we use the family link controls to limit it a little but if they tried to get around them they could. The hope is we’ve at least prepared them a little before they have totally unfettered access to the internet.

    We did try YouTube kids a little but it was such a garbage experience we just blocked the app everywhere.




  • Care needs to be taken with big orgs like the NHS to not try and boil the ocean with massive IT systems. Concentrating on open interoperability standards allows for smaller more flexible contracts and the ability to swap out components when needed.

    Open source licences would be the ideal default although at a minimum the purchasing org should have a licence that allows them (or subcontractors) to make fixes without being tied to the original vendor.









  • That’s not really true. Yes avoiding complex instructions makes the front end easier to pipeline but there are lots of smarts in the backend to do prediction and scheduling to keep the execution units fed. The ISA might be free to use but no one is sharing their highly optimised server silicon architecture designs.

    RISC-V’s challenge is can they standardise the software ecosystem enough that things just work across a multitude of chip providers or does everything devolve into specialist distributions taking advantage of each manufacturers “special sauce” custom instructions.

    Gaining design wins over Arm’s microcontrollers for bespoke hardware was the easy bit. Replacing stuff in the server space is much harder and something that took Arm decades to make inroads into.