I think it would be a pretty serious security issue if the president sat down and started typing important stuff into chatgpt via chrome or whatever. So if that's what you have in mind that's not going to happen.

@Odoacre In 5 years there is a very good chance the more serious risk will be in not consulting AI for decisions.

@mark1263 sure. But it sounds like you have some sort of commercial chat bot in mind

@Odoacre I hadn’t given its exact form much thought to be honest. Mainly that’s it’s a digital intelligence is all.

This is almost 5 years in the future, AGI is widely considered to arrive before then, basing a decision on an intelligence equal to or superior to human may be more an issue of pride than wisdom

Yes, that sounds appropriate, GPT 4 or higher. Apologies and understanding that simply “AI” is a little vague.

My thought was that the President would credit AI for a decision to an important issue, in such a manner that any outcome, deemed good or bad would be credited to the AI and not to the President.

So could we specify something like "transformer-based LLM of complexity gpt3 or above"?

True, AI is too VAGUE. I mean technically speaking we are following AI directives due to Super Computer Models in Science. What is the Line between a Simulation and an AI? An AI is literally a simulation of Human Knowledge. When it becomes more like a Human we consider the AI better. When Chat-GPT gives you an answer like a 5 year old it seems more like some crappy educational program then a true AI.

Does the president need to have used the AI himself personally, or can it be something like deferring to an AI developed by some state department ?

@Odoacre or basing a decision on advice from human experts who are just regurgitating AI takes.

I was thinking something like chat gpt, an interactive AI that gives responses to questions

What do you count as AI?