Lets say you give access to the internet to it and say "hack into a pentagon". Well, surely, if it succeeded it means that Chinese and Russian hackers did this a long time ago. If you told it "destroy humanity", well, again, surely, it would not be able to do that even without imposed restrictions by OpenAI.
So, how exactly AI would lead to something horrible? That is the first question.
And the second one is: why is OpenAI pushing for regulation so much? Do you think they are honest? What could be alternative motives for pushing this agenda?
Could it be that Microsoft and Google are doing something that top corporations always have tried to do? Make deals with governments and impose restrictions so competition is unlikely to emerge?
OpenAI pushes for regulation because
- It drives excitement, which helps their business
- It will make it harder for competitors to succeed
- It aims the regulation in the wrong direction
- It lets them dismiss criticism from other sources
None of the robot god stuff is real; it's a honey trap for the credulous, laid by profiteers. LLMs are really exciting, but they aren't actually an existential threat, and the people who trumpet that are doing their own (otherwise interesting and valuable) work a massive disservice.
Another could be that they see that it will be regulated anyway, so they are getting ahead of the narrative and getting some good will with the regulators at the same time.
1. Open source AI research is a threat to O̶p̶e̶n̶AI.com's business [0] and they cannot win the race to zero if the participants (Stability, Apple, etc) are already at the finish line.
2. Scapegoating 'AI safety' and spreading doomsday tales to governments so that they can wipe out open source AI models and setup 'licensed' and 'compliant' AI models (whilst O̶p̶e̶n̶AI.com open sources their own compliant LLM) [1]. ie regulatory capture.
3. Promoting Worldcoin crypto snake-oil as an 'antidote' against everything getting faked by AI to 'verify' human eyeballs once their doomsday tales becomes true. [2]
Sources:
[0] https://www.semianalysis.com/p/google-we-have-no-moat-and-ne...
[1] https://www.reuters.com/technology/openai-readies-new-open-s...
[2] https://worldcoin.org/blog/engineering/humanness-in-the-age-...
To raise legal and regulatory barriers to entry for competitors. They regulations they are pushing for are the sort that will enshrine their business model into law, locking out newcomers to the technology.
That is a cynical take, which I cannot prove true, but I think it is part of the answer.