Ok, we can stop now. I (and plausibly many others before me, but I haven't seen) have definitively proven that #ChatGPT is utterly useless as a chat bot... #IPutOnMyRobeAndWizardHat
| blog (in bulgarian) | https://undef.me |
| ✔ | https://kunev.me/ |
Ку 🐻🇧🇬🇪🇺
- 10 Followers
- 320 Following
- 523 Posts
I break things to see how they (don't?) work | I pass the butter | toots may contain Bulgarian, да го еба | !! и ≠ й ≠ ѝ !!
I am actually excited! I think the engine itself hasn't been touched for this, it's "just" the UI/UX stuff, so it will probably still take a while to get to a 250 year world. But it feels like it's been tamed a bit. The free version used to go fully unresponsive around 100 years or so, often dragging the machine in general to barely respond to anything. Now it might be slower, but it seems it's capped at around 2.5 cores so you can use your computer while you wait for the world to generate.
Не съм изненадан, даже леко се чудя що изобщо пробвах. И на български е зле в тая работа.
Can't wait for someone to train #GPT on a better corpus of Eastern European languages and start generating shit like that at a rate in the high 3 digit numbers per day.
The ability to fully automate journalistic BS will truly do wonders for fucking up whatever remnants of a remotely adequate society we still have.
The ability to fully automate journalistic BS will truly do wonders for fucking up whatever remnants of a remotely adequate society we still have.
Не разбирам от къде идва нееднозачността на „по“ в случая, но поне явно може да заподозре, че не разбира въпроса и да даде повече отговори, от колкото е необходимо.
So I saw some posts by people telling ChatGPT to just prove some obviously false things for which it happily compiles nicely sounding pseudo-proofs. Obviously I had to try with something outrageously stupid. The result was quite lovely, actually way better than what I could have ever hoped for. 😄