Hacker News new | ask | show | jobs
by anthonyskipper 590 days ago
Some of us look forward to that future where you mostly just interact with AI. The one depressing us is not turning running over our goverment to AI. The sooner we can do that the better, you can't trust humans.
7 comments

That's still trusting humans, either the ones who created the AI and gave it it's goals/parameters or the humans that actually implement it's edicts. Can't get away from people, it's a lesson all the DAO hype squad learned quickly, fundamentally you still need people to implement the decisions.
If you don't trut humans you shouldn't trust AI-.

AI is based on human input and has the same biases.

Minus the accountability.
> Some of us look forward to that future where you mostly just interact with AI.

What is it about that that appeals to you? I'm genuinely curious.

A world without human interaction feels like a world I don't want to exist in.

If you are autistic (for example) I'm guessing human interaction can be extremely difficult and very stressful and triggering. Machines are much more amenable and don't have loads of arbitrary unwritten rules the way humans do. Maybe the idea of being entrapped by bureaucracy introduced by machines will be better than the the bureaucracy introduced by humans?
> Machines are much more amenable and don't have loads of arbitrary unwritten rules

I'm sure system prompts of the most famous LLM are just that

They are not as arbitrary as body language for example.
The difference between a standard human-written algorithm and machine learning is exactly that inability to find the rules transparent, predictable, and not arbitrary.
I can see this but I think humans are much more random than most LLMs - they lie, they have egos, they randomly dislike other humans and make things difficult for them. Never mind body language, networks of influence, reputation destruction and all the other things that people do to obtain power.

I think LLMs are much more predictable and they will get better.

They expect AI bureaucracy to be more effective than human bureaucracy.

I expect this to be entirely true in some cases.

"You can't trust humans" but you can trust a non deterministic black box to take their place?
Humans already are non-deterministic black boxes, so I'm not sure I would use that comparison.
What's better, a non-deterministic black box or a non-deterministic black box designed by non-deterministic black boxes?
How are humans designed, for comparison?
Do you believe humans are designed by someone?
Nope. Precisely my point.
For me they are more a gray box. That is why publicity and propaganda work.
Humans are accountable. You can sue a human.
And you can't sue the corporation that made an AI?
In a world where all lawyers and prosecutors use its product for professional and private purposes?
In theory yes, but good luck with that.
Are you suggesting humans are deterministic?
A little bit, we are. With some degree of confidence, given the incentives you can predict the output.
You won't receive better outcomes in this world. The people in charge will simply change what they're measuring until the outcomes look better.
What looks like turning things over to AI is really turning things over to the people who own the AI, which is another thing entirely.
Can we trust AI?
Define “trust”, because that singular word carries immeasurable weight.

Can we trust AI to make consistent predictions from its training data? Yeah, fairly reliably. Can we trust that data to be impartial? What about the people training the model, can we trust their impartiality? What about the investors bankrolling it, can we trust them?

The more you examine the picture in detail, the less I think we’re able to state it’s trustworthy.

>What about the investors bankrolling it, can we trust them?

This is the crucial question. We live in capitalism and maximising profits is the most dominant axiom.

Without any legislation in place anywhere in the world and AI-supported diagnosis and therapy very much already in place what would prevent the companies bankrolling the software/hardware from bricking the tool on various grounds, like payments not made, political reasons: Cory Doctorow wrote a blog-post in 2022 where John Deere bricked Ukrainian tractors [0]; or a manufacturer of respirators refused to repair the devices at the height of COVID, so that only Hackers enabled the devices to run and save human lifes? [1]

[0] https://doctorow.medium.com/about-those-kill-switched-ukrain...

[1] https://www.hackster.io/news/polish-hacker-shares-software-s...

Yes, we can. But should we?