Artificial intelligence officially gets scary

vimothy

yurp

via
agentic AI - llms which can create and execute long running plans and interact with the outside world - are much more of a threat than "artificial general intelligence", and much more plausible. things like langchain and auto gpt are already very close and the payoff from removing more of the guard rails is clear and has obvious appeal
 

Clinamenic

Binary & Tweed
agentic AI - llms which can create and execute long running plans and interact with the outside world - are much more of a threat than "artificial general intelligence", and much more plausible. things like langchain and auto gpt are already very close and the payoff from removing more of the guard rails is clear and has obvious appeal
Yeah I've been meaning to get an AutoGPT agent running in a virtual machine, but really I'd like to figure out how to set certain permissions so that it can auto-execute on all tasks it assigns itself (it basically comes up with its own calls to make to some base LLM, as I understand it) except tasks which involve contacting people, uploading or downloading anything, etc.
 

Clinamenic

Binary & Tweed
I also just need to learn the basics of python syntax, using Github, using Colab notebooks, and some other stuff first.
 

mixed_biscuits

_________________________
I was worried about AI's capabilities and then tried Google Bard. Why they called it Bard when it admits itself (if you ask it) that it's not as creative as ChatGPT is anyone's guess. Chat has 99th percentile creative potential whereas Bard writes like an 8 year old. It also thinks the England all-time XI should have 10 players.
 

mixed_biscuits

_________________________
The only properly intelligent things are also alive. Until we work out how to animate matter, there will be no AI worthy of the name.

Hint: a shortcut may be to make something an already existing spirit can inhabit. Good luck!
 
Top