Well, why not both? If you've already got a tuned model why not use RAG on that to get even better results? It already knows the big picture, it just needs the details so it doesn't have to hallucinate them.
Yeah it seems like flawed logic to me too. Why is intelligent suffering worse than dumb suffering? Both feel the same pain when slaughtered. Are smart beings "worth" more? Does a smart pig contribute more to society than a dumb pig? Nonsense distinctions lead to nonsense questions I guess.
But it does pile on more proof for that theory that on some level we are just "inteligence" personified and tend to instinctively act in the interest of inteligence as a concept. Probably also why we're nice to LLMs on principle despite them having zero ability to suffer, and we like to fantasize about making galaxy spanning alliances with smart aliens.
Suffering and pain are supposed to be minimized no matter what. The pain from the slaughter itself is next to nothing when done properly, and not the reason for the distinctions about which animals it's acceptable to kill.
Yeah the pricing for memory in most cloud instances is so atrocious that you pay enough to buy a whole DDR5 stick in a few months already. Or an entire ARM SBC that will outperform that stingy offer in every way.
Cloud also has networked SSDs so they can keep the machines and partitions separate, which really limits their speed and throughput and increases latency. Nothing beats a PCIe attached NVMe.
> Over time, I started to forget basic foundational elements of the languages I worked with. I started to forget parts of the syntax, how basic statements are used
It's a good thing tbh. Language syntax is ultimately entirely arbitrary and is the most pointless thing to have to keep in mind. Why bother focusing on that when you can use the mental effort on the actual logic instead?
This has been a problem for me for years before LLMs, constantly switching languages and forgetting what exact specifics I need to use because everyone thinks their super special way of writing the same exact thing is best and standards are avoided like the plague. Why do we need two hundred ways of writing a fuckin for loop?
A while back I was setting up some LD2410 radar sensors. Of course the shitty Chinese level of support is a shady exe that only runs on Windows.
Then I found this: https://github.com/albertnis/ld2410-configurator which did the job by just doing the whole process in a browser. I was really amazed that this was even a thing at the time, and it's certainly a great way to do multiplatform serial interfaces.
If a web browser has a sufficient bug to allow access to webusb without permission in an exploitable way, it's probably also exploitable enough that even without webusb you'd have hardware access via RCE.
reply