Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

cbc.ca

Post reply on HN
Print view View original

21–30 of 38 posts

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#21

Post by justinclift »

DoreenMichele wrote: post #4

"I did not intend to generate or refer to fictitious cases in this matter. That is clearly wrong and not something I would knowingly do," Ke wrote in her deposition. "I never had any intention to rely upon any fictitious authorities or to mislead the court." So beware what ChatGPT tells you. It might not be real and this can be a big problem in some situations.

Directly related to this, there is the recently submitted project named like a famous award: https://news.ycombinator.com/item?id=39514192

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#22

Post by graemep »

mdp2021 wrote: post #19
ifdefdebug wrote: post #17

They have to realize that the language model did it's job just fine. It produced a convincing piece of well-formed language that fits the context window. If they wanted knowledge instead, they should have used a knowledge model.

> convincing piece of well-formed language that fits the context window I hope you do realize that "artificial conman" is hardly a goal.

If people are willing to pay for it, then it is.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#23

Post by mdp2021 »

DoreenMichele wrote: post #7
255kb wrote: post #6

Earlier quoted context omitted.

Every time I see a company/startup claiming to do an AI about X or Y, I wonder how they can be trusted at all. What about all the hallucinations, false positives, etc.? Aside from some sort of tool that classify cases, or extract key points from an agreement, and talk in probabilities rather than asserting things I don't see a use for these. And I don't know if we will be able to 100% trust these tools one day.

We haven't figured out how to deal with human intelligence effectively. It's jumping the gun to imagine we can do AI well. /2 cents -- as if I need a "disclaimer" but the internet being what it is....

And yet Columbus made the egg sit firmly on the table.

The "«we»" that includes the lawyer in question has so little to do with the "«we»" that may include Richard Feynman or Marvin Minsky (and further in the right directions), that you may want to refine your model past the problem of Porphyry and hyper-realism in the universals: that "«we»" is insubstantial. It is not some collective humanity that arrived to non-Euclidean geometry, but a small number of people that had the skills and dealt with the problem intensively.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#24

Post by xxs »

rcbdev wrote: post #9

Despite what some in the tech space might think, I'm not sure I can lay the blame 100 % on the lawyer. It's true that this is definitely malpractice and they should have known better, it's also true that big tech for years overplayed their hand with the marketing of "AI" to laypeople and management, which has created completely unreasonable expectations. As long as OpenAI has a profit motive to make people believe Ch…

so you are saying the lawyer has been conned? I'd not think twice about her not cross checking her sources.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#25

Post by mdp2021 »

miningape wrote: post #20
mdp2021 wrote: post #19

Earlier quoted context omitted.

> convincing piece of well-formed language that fits the context window I hope you do realize that "artificial conman" is hardly a goal.

I hope you do realise this is what it was trained to do, not be correct, not be inventive/creative, but to make text that _sounds_ good.

I very well realize it, I and I hope that re-reading again my original comment you will see that I am condemning that specifically.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#26

Post by mdp2021 »

graemep wrote: post #22
mdp2021 wrote: post #19

Earlier quoted context omitted.

> convincing piece of well-formed language that fits the context window I hope you do realize that "artificial conman" is hardly a goal.

If people are willing to pay for it, then it is.

Clearly that is what some seem to think, and it should be also clear that it has sinister sides.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#27

Post by pjc50 »

ifdefdebug wrote: post #17

They have to realize that the language model did it's job just fine. It produced a convincing piece of well-formed language that fits the context window. If they wanted knowledge instead, they should have used a knowledge model.

Man, this entire conversation is full of is/ought conflation. And deceptive marketing.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#28

Post by pjc50 »

255kb wrote: post #6
DoreenMichele wrote: post #4

"I did not intend to generate or refer to fictitious cases in this matter. That is clearly wrong and not something I would knowingly do," Ke wrote in her deposition. "I never had any intention to rely upon any fictitious authorities or to mislead the court." So beware what ChatGPT tells you. It might not be real and this can be a big problem in some situations.

Every time I see a company/startup claiming to do an AI about X or Y, I wonder how they can be trusted at all. What about all the hallucinations, false positives, etc.? Aside from some sort of tool that classify cases, or extract key points from an agreement, and talk in probabilities rather than asserting things I don't see a use for these. And I don't know if we will be able to 100% trust these tools one day.

> What about all the hallucinations, false positives, etc.?

There's a very small disclaimer at the bottom that says all those things are your problem.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#29

Post by DoreenMichele »

mdp2021 wrote: post #23
DoreenMichele wrote: post #7

Earlier quoted context omitted.

We haven't figured out how to deal with human intelligence effectively. It's jumping the gun to imagine we can do AI well. /2 cents -- as if I need a "disclaimer" but the internet being what it is....

And yet Columbus made the egg sit firmly on the table. The "«we»" that includes the lawyer in question has so little to do with the "«we»" that may include Richard Feynman or Marvin Minsky (and further in the right directions), that you may want to refine your model past the problem of Porphyry and hyper-realism in the universals: that "«we»" is insubstantial. It is not some collective humanity that arrived to non-Eu…

I have absolutely no idea at all what you are saying to me.

For clarification of some kind: I homeschooled my twice exceptional --"gifted and (learning) disabled" -- sons and for a time did pro bono professional work for a gifted organization in support of me being the teacher in a private school with two students for some years under California law and designing the custom curriculum for my special-needs sons so as to make it digestible without talking down to them and accommodating their big strengths and big weaknesses at the same time.

My emails were seemingly "popular" for a time on a gifted homeschoolers list and you can blame their enthusiasm for my writing for my habit of blogging to this day, though I have no traction and it makes damn little money and is no doubt basically me wasting my life.

Fortunately, we are all dust in the wind anyway, so I'm sure none of it really matters, but I have Opinions about the poor job we generally do with dealing with other humans, especially short ones under the legal age of majority.

So maybe that gives you some context for my stupid comment. No, I don't care to talk further about any of this really, tbh.

Re: B.C. lawyer reprimanded for citing fake cases invented by ChatGPT

#30

Post by mdp2021 »

DoreenMichele wrote: post #29
mdp2021 wrote: post #23

Earlier quoted context omitted.

And yet Columbus made the egg sit firmly on the table. The "«we»" that includes the lawyer in question has so little to do with the "«we»" that may include Richard Feynman or Marvin Minsky (and further in the right directions), that you may want to refine your model past the problem of Porphyry and hyper-realism in the universals: that "«we»" is insubstantial. It is not some collective humanity that arrived to non-Eu…

I have absolutely no idea at all what you are saying to me. For clarification of some kind: I homeschooled my twice exceptional --"gifted and (learning) disabled" -- sons and for a time did pro bono professional work for a gifted organization in support of me being the teacher in a private school with two students for some years under California law and designing the custom curriculum for my special-needs sons so as…

So I understand you experienced difficulties in dealing with developing intelligences, and you entailed that it should be difficult to create artificial ones.

But nobody said it was easy (also because ideas like "to invent the lightbulb is not easy" in that form make little sense), and yet the "right" people - not "«we»" - will probably reach the goal, partially owing to a piece of the "«we»" and largely in spite of a large piece of the "«we»".

Just like some achieved the lightbulb, and some (Columbus) achieved making an egg stand straight on the table while those nearby claimed it was impossible.

Now, the problem in the original context was not about those people able to create a hammer, but about the people unable to use it. The problem here is not about creating intelligence but using it.

Post reply on HN