shared brain?

[PapaSchlumf]

what if we can link the auto-gpt from one user to another via something like torrent , that multiple user can train or interact with auto-gpt in specific order to develop a big function in smaller parts.
as far as i understand, not many people have the hardware for bigger llm´s and the big bois like gpt4 needs hella lot and expensive hardware.
personaly i would let my pc run and help develop or test whatever i can if it would be possible.

i am just at the beginning of learning programming so excuse me if not possible and not much knowledge

[AdityaVishwakarmaIN]

The idea of linking auto-GPTs via a torrent-like network is interesting, but it raises serious security concerns. If multiple users are accessing and training the same model, it's a huge risk for malicious code or data that could compromise the model's integrity or the privacy of user data. Hackers could easily introduce poisoned data into the model, and that could mess up the output. Plus, if the model is trained on sensitive data, there's a risk of data leakage or theft if we don't put proper security measures in place.
And don't even get me started on software vulnerabilities! If we don't secure the software used to run and train the model, attackers could exploit vulnerabilities to get unauthorized access to our systems and data. We need to be careful and implement proper security measures like secure communication protocols, access controls, and authentication mechanisms. We also need to conduct regular security audits and vulnerability assessments to keep the system secure.

[PapaSchlumf]

oh boy, so a good or way better security system would be in need first, if at all.
same base problem as always, people with bad intentions

[Wladastic]

What you are talking about is a mix of local running llm's and a peer2peer functionality.
You may look into existing llm text-gen plugins if you want. But those don't work mostly with auto-gpt and require a different prompting strategy.