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

AI Police Reports: Year in Review

eff.org

Post reply on HN
Print view View original

141–150 of 224 posts

Re: AI Police Reports: Year in Review

#141

Post by sejje »

tialaramex wrote: post #68
lostnground wrote: post #27

Earlier quoted context omitted.

While I agree that officers should be accountable. More enforcement of them will not suddenly make them good officers. Other nations train their police for years prior to putting them into the thick of it. US police spend far less time studying, and it shows, in everything from de-escalation tactics to general legal understanding. If you create a pipeline to weed out bad officers, then there needs to be a pipeline pr…

AIUI US policing is descended from slave catching and strike breaking. Two activities which I think we'd say today are obviously bad. In many European states their policing starts as town guards tasked with ensuring order. Order is, at least, not obviously bad. So that's a philosophical difference in what these forces even think their purpose is.

You understand it wrong. That's not where police come from.

Re: AI Police Reports: Year in Review

#142

Post by impossiblefork »

roenxi wrote: post #128
impossiblefork wrote: post #121

Earlier quoted context omitted.

I haven't tried Dramatron, but my experience is that it isn't possible to do sensibly. With regard to the second part >AI transcription & summary seems to be a strong point of the models so I don't know what exactly you're trying to get to with this one. If you have evidence for that I'd actually be quite interested because humans are so bad at representing what other people said on the internet it seems like it shou…

I can see why they'd struggle, I'm not sure what you're trying to ask the model to do. What type of analysis are you expecting? If the model is supposed to represent the views expressed that would be a summary. If you aren't asking it for a summary what do you want it to do? Do you literally mean you want the model to perform conversational analysis (ie, https://en.wikipedia.org/wiki/Conversation_analysis#Method )?

Usually I use the format "Analyze the following ...".

For simple discussions this is fine. For complex discussions, especially when people get into conflict-- whether that conflict is really complex or not, problems usually result. The big problems are that the model will misquote or misrepresent views-- attempted paraphrases that actually change the meaning, the ordinary hallucinations etc.

For stories the confusion is much greater. Much of it is due to the basic way LLMs work: stories have dialogue, so if the premise contains people not being able to speak each other's language problems come very soon. I remember asking some recent Microsoft Copilot variant to write some portal scenario-- some guys on vacation to Teneriffe rent a catamaran and end up falling through a hole in the world of ASoIAF and into the seas off Essos, where they obviously have a terrible time, and it kept forgetting that they don't know English.

This is of course not obviously relevant for what Copilot is intended for, but I feel that if you actually try this you will understand how far we are from something like AGI, because if things like OpenAIs or whoever's systems were in fact close, this would be close too. If we were close we'd probably see silly errors too, but it'd be different kinds of errors, things like not telling you the story you want, not ignoring core instructions or failing to understand conversations.

Re: AI Police Reports: Year in Review

#143

Post by sejje »

bushbaba wrote: post #64

To me it’s a question of it they are on average better. It’s not like human based input is perfect either.

Do mistakes by the LLM carry the same weight as mistakes by the officer? Can the case be thrown out for an LLM mistake?

Does the officer face repercussions for introducing false information via this tool?

Re: AI Police Reports: Year in Review

#144

Post by a1j9o94 »

ptx wrote: post #108
hombre_fatal wrote: post #97

Earlier quoted context omitted.

Because it gives you an answer and all you have to do is check its source. Often you don’t have to do that since you have jogged your memory. Versus finding the answer by clicking into the first few search results links and scanning text that might not have the answer.

As I said, how are you going to check the source when LLMs can't provide sources? The models, as far as I know, don't store links to sources along with each piece of knowledge. At best they can plagiarize a list of references from the same sources as the rest of the text, which will by coincidence be somewhat accurate.

Pretty much every major LLM client has web search built in. They aren't just using what's in their weights to generate the answers.

When it gives you a link, it literally takes you to the part of the page that it got its answer from. That's how we can quickly validate.

Re: AI Police Reports: Year in Review

#145

Post by sejje »

thunderfork wrote: post #140
sejje wrote: post #138

Earlier quoted context omitted.

By the end of your paragraph, you decided that criminals are victims.

Everyone the policing system interacts with is not only a convinced criminal, but convicted justly, yeah? That's what you actually believe?

No, that's not what I believe, I didn't say any of those words.

Re: AI Police Reports: Year in Review

#146

Post by thunderfork »

sejje wrote: post #145
thunderfork wrote: post #140

Earlier quoted context omitted.

Everyone the policing system interacts with is not only a convinced criminal, but convicted justly, yeah? That's what you actually believe?

No, that's not what I believe, I didn't say any of those words.

The person you replied to didn't say "criminals are victims", either, so comprehending your post requires some inference.

Feel free to clarify what you did mean, it's a lot more helpful than insisting on what you didn't mean.

Re: AI Police Reports: Year in Review

#147

Post by Workaccount2 »

kylecazar wrote: post #124
futuraperdita wrote: post #16

What worries me is that _a lot of people seem to see LLMs as smarter than themselves_ and anthropmorphize them into a sort of human-exact intelligence. The worst-case scenario of Utah's law is that when the disclaimer is added that the report is generated by AI, enough jurists begin to associate that with "likely more correct than not".

Maybe it's just my circle, but anecdotally most of the non-CS folks I know have developed a strong anti-AI bias. In a very outspoken way. If anything, I think they'd consider AI's involvement as a strike against the prosecution if they were on a jury.

A core problem with humans, or perhaps it's not even a problem, just something that takes a long time to recognize, is that they complain and hate on something that they continue to spend money on.

Not like food or clothing, but stuff like DLC content, streaming services, and LLMs.

Re: AI Police Reports: Year in Review

#148

Post by closewith »

latexr wrote: post #105
hombre_fatal wrote: post #94

Earlier quoted context omitted.

Dan makes a case for being charitable to the commenter and how lame it is to neener-neener into the past, not that it has some opposite meaning everyone is missing out on.

Dan clearly references how people misunderstand not only the comment (“he didn't mean the software. He meant their YC application”) but also the whole interaction (“He wasn't being a petty nitpicker—he was earnestly trying to help, and you can see in how sweetly he replied to Drew there that he genuinely wanted them to succeed”). So yes, it is the opposite of why people link to it (which is a judgement I’m making, I’…

The comment isn't infamous because it was petty or nitpicking. It's because the comment was so poorly communicated and because the author was so profoundly out-of-touch with the average person that they had lost all perspective.

It's why it caught the zeitgeist at the time and why it's still apropos in this conversation now.

Re: AI Police Reports: Year in Review

#149

Post by kylecazar »

saghm wrote: post #134
theoreticalmal wrote: post #126

Earlier quoted context omitted.

Why do people in your circle not like AI? I have similar a experience about friends and family not liking AI, but usually it’s due to water and energy reasons, not because of an issue with the model reasoning

If your circle has any artists in it, chances are they'll also have a very negative perception, although influenced heavily by the proliferation of AI-generated art. At least personally, I've seen basically three buckets of opinions from non-technical people on AI. There's a decent-sized group of people who loathe anything to do with it due to issues you've mentioned, the art issue I mentioned, or other specific thin…

Pretty much hit the nail on the head -- while there are some artists, most are from traditional broadly "intellectual" fields. Examples: writers, journalists, academia (liberal arts), publishing industry...

Re: AI Police Reports: Year in Review

#150

Post by kylecazar »

Workaccount2 wrote: post #147
kylecazar wrote: post #124

Earlier quoted context omitted.

Maybe it's just my circle, but anecdotally most of the non-CS folks I know have developed a strong anti-AI bias. In a very outspoken way. If anything, I think they'd consider AI's involvement as a strike against the prosecution if they were on a jury.

A core problem with humans, or perhaps it's not even a problem, just something that takes a long time to recognize, is that they complain and hate on something that they continue to spend money on. Not like food or clothing, but stuff like DLC content, streaming services, and LLMs.

At least in my case, I suspect they also don't keep up with the progress. They did experiments in 2023/24, were thoroughly put off, have not fired it up since. So the impression they have is frozen in time, a time when it was indeed much less impressive.
Post reply on HN