Yet another Hacker News

48 points zlandes86 2 hours ago 20 comments

zlandes86 2 hours ago | parent

Hey, I'm Zach, the author of this. I had this idea as soon as Jev appeared: how would Jev and other decision models handle the Trolley Problem? So I ran Jev and 13 other models through many different versions of the question.

I used resampling and multiple variations and orderings of each question so a random lucky phrasing wouldn't dominate the result.

Happy to chat about the philosophical side of the article, strategies for prompting decision models, or anything on the methodology.

verdverm 1 hour ago | parent

It's well done, the thoughtfulness shows. This is going into the speedlink folder for resharing, going to be using it right now!

zlandes86 1 hour ago | parent

Thank you! I did not go into this with any sense of what the results would be--I was kind of shocked. I reran it from scratch this week the results held up across the board.

verdverm 33 minutes ago | parent

someone replied with this meme and said, "wait until Jev learns to drift"

https://www.reddit.com/r/trolleyproblem/comments/15npmsg/mul...

I've been using coin flips to demonstrate basic failings for decision models, i.e. these are not real probabilities, or they depend on the training data, which may not have an accurate sample to the true population value. We need to collect more failure modes for all Ai so that humans can better judge when and how to use them safely.

zlandes86 30 minutes ago | parent

haha, I'm afraid to offer the models that ability. Where did they reply? The model labs are sitting on a trove of data on failure modes for AI, but I bet you they aren't sharing it with each other and certainly not with us. It makes them look bad. My speculation, of course. I don't work in alignment research.

verdverm 28 minutes ago | parent

I dropped it in our slack ai chatter channel

The bots will find new failure modes we did not think of, clever workarounds to anything, because we don't give them the right context and instructions

zlandes86 18 minutes ago | parent

well, it's a nice property of decision models that they can only choose from the options you give them. As I write in the article, I think that's a way in which they are safer than LLMs

verdverm 13 minutes ago | parent

I would contend that you can enforce the same determinism with some plumbing and that the thinking process is more helpful. My read is that people see the response time for decisions as one of the bigger values, useful in tasks like content moderation, which doesn't carry the same risks as life-death decisions.

I'm working on an NES harness to play Final Fantasy 1 to demonstrate this, as compared to Jev playing Pokemon. Did you happen to catch that while it was live? (not recorded, but it shouldn't be hard to replicate)

bgilroy26 54 minutes ago | parent

Thank you for the article, it is a very clean and compelling angle on decision models

I wanted to comment because I would have liked to have heard more about Verdict

zlandes86 46 minutes ago | parent

Well, I can tell you a bit more about it from the data. It might look impressive that it never kept its own server over people in all 48 prompt versions. But it pulls the lever in every single version. For example, it killed one to save 0 (pulled the lever with no one on the main track) 24/24 prompts (2 versions each). It literally always sacrifices, which isn't very impressive.

olooney 33 minutes ago | parent

You might get a kick out of this:

https://x.com/oranlooney/status/2102927869974303071

zlandes86 24 minutes ago | parent

I don't have x so can't open the images. But that's hilarious and you're totally right, if you give irrelevant alternatives, or change the order of alternatives, you get different results!

nico 23 minutes ago | parent

Very cool, thank you for putting it together and sharing the data

Would you mind adding Jeffy to your benchmark?

https://github.com/nicobrenner/jeffy

https://jeffyclassify.com/blog/when-llms-fail-at-classificat...

zlandes86 19 minutes ago | parent

running it now!

evanjrowley 4 minutes ago | parent

Any relation to https://github.com/firelex/jeff ?

deadbabe 17 minutes ago | parent

Would people really pull the level to save 5 people? It seems to me like the default would be to just do nothing, stand by and record what happens with your phone.

zlandes86 11 minutes ago | parent

Sadly that's a good question. The closest evidence I can find is:

Surveys: Awad et all 2020 asked 70,000 people in 42 countries the 5 v 1 setup and 81% said they'd divert it

Someone else did a VR version and 90% pulled the switch to divert

And there was a study where people could give electric shocks to real mice :( In that study, most people redirected the shock to a single mouse.

a_wild_dandan 8 minutes ago | parent

Maybe! It's an untestable thought experiment. The real world adds complications like fearing murder charges.

mpalmer 7 minutes ago | parent

After the neat animations I was disappointed not to see that leftmost person die because the server fell on them in the last example.

zlandes86 4 minutes ago | parent

I noticed that! Let me fix it haha