/vr/ - Retro Games

>>10124218
> you've been sold the idea that ... this is an inevitability
well, the current data shows that GPT / LLMs with more horsepower and data tend to develop new and emergent abilities, and as more horsepower and data thrown at this thing, the current data points to this trend continuing. Ilya Sutskever believes this, the man pictured in my last post and the lead architect behind GPT 4. but you didn't know that, did you?
the main problem right now is horsepower, and as the last 40 years of technological progress has shown, that will not be a problem for long...

>>9891970

>>9883114
Those are some good steps, i'm proud of you.
> I welcome any tips
Mesen supports LuaSocket out of the box, if BizHawk does, you can get data in and out of the emulator. If you can get it shooting data to localhost, this should give you flexibility later with what models you choose, and also the possibility of using some out of the box ML models locally. If you can get this running now, you will likely save a metric tons of work in the future.
I don't know much about ML, but that would be my advice. ML HF

>>9680516
yes, its good for basic concepts, but as you get more specific into advanced topics it will start spitting out correct sounding bullshit so make sure you verify what it tells you