/biz/ - Business & Finance

>>58007623
>With a decent translation you can run code for ChatGPT on a server stack from 1995.
no you cannot, it takes many gigabytes of RAM to have the model in memory all at once and do the operations quickly with the GPUs. You could run it on a very old computer if you swap to disk, sure, but that would mean it's (conservatively) 1000x slower, and even MORE expensive.

>when most of what it's being asked to do is write middle school papers?
you're lacking vision

Don't get me wrong NVDA is overbought. I sold my shit last week at 915 and I'm happy. But LLMs and generative AI are here to stay and will be a big deal (tm). MSFT is a good bet, picture fucking related.

>>58007649
>Your brain has a TDP of about 15 watts
so what you literal retard? do you not understand these models can do some human tasks and are MUCH less expensive than a engineer or graphic designer?

>>57967569
I know AI is the future but NVDA is overbought anyway.

>>57967557
>MSFT being 50% of your folio
I don't know sounds pretty based to me