Hacker News Reader

Comment by elicksaur | Hacker News Reader

Good. Totally fine with trying to use AI to give legal advice, but it should be done with a lawyer’s license on the line. A company that explicitly disclaims being a law firm and states is not giving legal advice should also not get to tweet that they are “representing” someone in court.

A good bar (pun intended) for the quality of the tech is if it is good enough that a licensed attorney trusts it to give legal advice with their livelihood at stake. If this product doesn’t work for DoNotPay, they can just walk away and do something else, as they are doing anyways here. If it doesn’t work for a lawyer, they’d get sued for malpractice and possibly disbarred, ruining their career. When someone trusts it to that level, have at it.

View on HN

No, bad. I may also disagree with some of the tactics DoNotPay used to represent themselves. But in a larger sense Lawyers cost money a lot of money. it's wonderful living in a time where the cost of filing suits is so low compared to defending against said suits.

AI lawyers can help lower those barriers to entry. the court in question is fucking traffic court. please lower the barrier to entry for defense and allow normal not rich folks to get on with their lives.

the costs involved in legal are so ridiculous, our society basically encourages blind folding ourselves to most business ethical standards in hopes that our product demands a high enough margin to then pay for what ever legal fuck ups that cost too much to figure out on the front end.

Do you imagine that a pro se litigant is going to have a hard time or will be required to spend small fortunes to defend themselves in "fucking traffic court?"

You're basically trying to cut the argument both ways. Administrative courts are not criminal courts, and an AI would never be allowed near a criminal defense trial for obvious reasons.

Yes because a lawyer costs their hourly rate whether it’s traffic or criminal court. Or more generally as an economy wide trend — paying a human to do something is often the most expensive route you can take.

We’re talking about traffic tickets that are usually in the hundred dollar range. If anyone is going to court rather than just paying it, it’s because $100 is a non-trivial amount of money to them.

It doesn’t have to be AI lawyers but any change to the system that reduces the total amount of work needed to be done by humans is a win.

>Lawyers cost money a lot of money. AI lawyers can help lower those barriers to entry.

For traffic court, perhaps, but these AI tools don’t seem guaranteed to not make this problem worse at broader scales.

These AIs will also be available to large firms, who are equally incentivized to use AI for augmenting argumentation through existing lawyers, but will also be incentivized to train their own walled garden powerful models in a way that poorer clients still would likely not have access to, and which individuals and smaller firms will not have the resources to train themselves.

These kinds of AI models could very easily serve to entrench and raise the cost of a defense by making it so you not only need a lawyer, you need a lawyer-backed-by-a-firm-with-a-LLM to be competitive at trial - making all existing problems even worse.

>walled garden powerful models in a way that poorer clients still would likely not have access to

Even if that becomes the case, this is some how worse than what we have now?

>raise the cost of a defense

Highly doubtful

> you need a lawyer-backed-by-a-firm-with-a-LLM to be competitive at trial - making all existing problems even worse.

First, most things don't go to trial so you're completely missing the cost savings associated with avoiding trial because of these types of AI assisted scenarios.

Second, the cost associated with training models will go down. The cost of a Harvard law degree has ... never gone down.

None of your objections are contradictory to requiring a real lawyer put their license on the line to vouch for the AI. Structural engineering jobs require a licensed PE to sign off on all designs. That doesn't mean they have to do all the work themselves, just supervise enough and trust their delegates enough to be confident the job is being done correctly. Licensed Electricians are responsible for all the work under their permits. Registered Nurses are largely responsible for all the LPNs working under them. All of these arrangements save money, but still require someone licensed to take responsibility. There is no reason AI couldn't be used in the same manner and save even more money, if done properly.

Yeah, I agree with basically everything you’ve said, but the standard of AI legal advice should be at least the standard of services required by current attorney regulations. They should probably be higher even. It is certainly likely that in the future, technology can provide that level of quality at a massive, extremely cost effective scale, but ChatGPT and DoNotPay is not that.

No direct complaint here, other than some president has to be set at some point.

Someone will have to take the initial risk. ChatGPT like AI may not be "the thing" but for some in this forum to be afraid of AI defense attorney is completely missing the forest through the trees.

> but it should be done with a lawyer’s license on the line.

Either that, or the client is fully aware they are defending themselves with the help of an AI that's not, and cannot at the moment be, a lawyer. As much as I want to believe AGI is just around the corner, LLMs are not individuals with human-level intelligence.

It's too bad programmers don't also have some sort of licensure as well. It would be helpful in keeping humans employed in creating and maintaining code, instead of letting AI run off with all our jobs.

What do you believe licensure would solve?

It would help in having some sort of body that says, we want to have humans involved in the chain of responsibility when creating code and not willfully hand over the control to AI.

Would that not be the person running the AI? The one giving prompts and verifying that prompts are fulfilled adequately?

If I've learned anything, its people often never verify the correctness of anything automated.

Reply on news.ycombinator.com