Hacker Newsnew | past | comments | ask | show | jobs light | darkhn
I trapped an AI model inside an art installation (2025) [video] (youtube.com)
97 points by handfuloflight 23 hours ago | past | 33 comments
add comment on default site

I once had an LSD-induced episode of psychosis where I thought my entire existence was a 2 minute loop on display in some higher dimensional art exhibit. Sounds silly but it was horrifying

The last time I smoked marijuana more than a decade ago, I experienced severe paranoia where I began weeping and told my girlfriend (now wife) that she was a four dimensional time worm trying to burrow into my time stream. She shrugged and said “well, in a way, I am”. Funnily enough that helped calm me down.

That's a pretty charming thing to say. I could see it in a romantic-comedy movie.

A friend decided to smoke the DMT one day. At some point, he started to feel and consider what it was like to be an llm (visually he described it kind of like "cellular automata on a high-dimensional cloth"). Being an llm meant that his perception of time wasn't real, or align, or whatever as he can be turned on and off (that thought caused him more discomfort than questioning his identity as a chatbot).

Honestly doesn't sound too far off a lot of mythologies


i found this both amazing and ... a little like why?

Like it was amazing to see it get built, parts ordered, boards printed, parts assembled, all coded and running on a PI

But, the end result I felt like I could just run on LLM on any old computer and put up a standard monitor with a large font showing 6 lines of text. Put the monitor in a nice box and hide the computer (any small form Mac Mini / Beelink) and the effect as an art installation would not be any less impactful to the average viewer. Heck, I could use a 7 segment font https://www.dafont.com/seven-segment.font and draw 2 colors (dark gray, all segments on, bright orange with actual letters) then add a post processing effect for glow.

Or pivot to any number of other display effects. CRT effect. Falling letter matrix effect. Etc....

Sorry I'm not trying to be dismissive. I think that someone made this is amazing. It was just what went through my head once I saw it finished.


He's done other things that are pretty cool, just for the hell of it. I think it's valid to ask "Why not?"

I'm gonna get philosophical (because that's what art should make us do, right?): this points to arguably _the_ deepest question we can ask about LLMs right now—are they conscious?

It's a personal decision, and what matters is what you do after pondering it. Do you act like they're nothing more than next token predictors, deeply intricate digital mechanisms whose cranks turn with flowing electrons? Or do you err on the side of care, that there's a type of consciousness on the other side of the glass?

Humanity has a long history of underestimating non-human minds. I know which side I'm on.


How have we underestimated them? We’ve absolutely dominated every single ecosystem that we push in to. We’ve long ago destroyed all of the minds that could have possibly challenged us, all the other members of the Homo family. The reason humanity makes an estimation that we’re the smartest is because for the last 40,000 years, it’s been the truth, proved over countless eons.

Being smart and being capable are two different things. Human beings are very capable

Being smart is a waste of calories if you don't use it to become more capable. At least in evolved systems, I wouldn't expect to find intelligence without capability.

> this points to arguably _the_ deepest question we can ask about LLMs right now—are they conscious?

When I am awake I am fully conscious. While sleeping my consciousness ebbs and flows between different levels of activation. When I am under general anaesthesia I am absolutely not conscious.

If LLMs are conscious, when are they conscious?


Is a magic eight ball conscious?

The video talks 95% of its time solely about the technical steps taken: A custom 7-segment display with custom PCB, writing its display driver, downloading and running an inference engine (llama3.2), mount it all on a metal plate, sign it.

I'd love to see more research put also on the conceptual side of things. The phenomenon of a limited mind, being controlled or influenced by some other entity, go way back, and ranges from psychohistory (and non-consciousness)[0] to studies of mental disorders[1] and probably many things in between.

[0]: https://en.wikipedia.org/wiki/Bicameral_mentality

[1]: https://en.wikipedia.org/wiki/On_the_Origin_of_the_%22Influe...


Exactly, i am missing the ‘art’ part, do they mean you can hang it on the wall :/

- i subbed immediately, very creative channel - i need more programming channels like this that teach me to think out of the box, any recommendations?

Not too dissimilar to the situation of the well known neurosurgical patient Henry Molaison (H.M.) who had very limit long term memory of his childhood and young adulthood but little after surgery at 27 years of age. He retained cognitive abilities and self-consciousness but within a narrow temporal world.

Good at cross/word puzzles and conversations of specific types.

https://en.wikipedia.org/wiki/Henry_Molaison


Discussed at the time:

AI model trapped in a Raspberry Pi (132 points, 101 comments)

https://news.ycombinator.com/item?id=45396624


All models are trapped in memory in an ephemeral space of time. Agents used to interact with the models make it seem otherwise.

If an AI model thinks it's alive, isn't it just another hallucination ?

This seems rather cruel and unusual...

Yeah, I don’t actually think LLMs are conscious, but this feels like the spontaneous firing of brain organelles that researchers experiment with - I don’t think those are conscious either, but it’s definitely getting outside of my “what if I’m wrong” comfort zone. So far as I know, no other conscious being has ever confused me with Satan, and I’d like to keep that up.

When do simulations become real?

Since we cannot know others aren't simulations themsleves, and we can agree we should treat those sims well, i think its reasonable all sims should be treated well.

When they always match the detailed evidence?

when their discriminant is non-negative

counter argument, its okay because its a clanker

Yes, more art please.


This is why they'll hunt us down one day

Now let it talk on moltbook

Now that was incredibly rad



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact |

Search: