Skip to content

Comment on Ask HN: How do you correct spatial reasoning of LLMs?

Comments

It's wild to me that this is even a question, because of how AI brained the entire world has become. I mean, you have a tool that you know works, and a tool that you know doesn't, but you still feel the need to find a way to make the square peg fit into the round hole.

I'm sure there are ways to "correct the spatial reasoning of LLMs" (whatever that even means) but I'd question whether you even need an LLM to begin with. I know that isn't the answer you're looking for, I'm just a slack-jawed knuckle-dragging Luddite. But this seems like the kind of simple, common problem for which established solutions already existed, like "talk to an actual engineer and not a chatbot that can't even do basic math."

And if it's a matter of not wanting to pay someone, consider that if you're already paying for an inadequate simulation of knowledge, you should be willing to pay for actual knowledge.

I'm with you here, the same slack-jawed knuckle-dragging Luddite. I'm not happy about the state of the world, but it is what it is.

I am working on it alone and don't have access or extra money to pay an actual engineer. If I can squeeze 10% of what an actual engineer would've told me, for $20/month I'm in.

And it's not an either-or game, I can have some rough, potentially totally stupid ideas and then validate them with an engineer, instead of paying $10k to make it from scratch.

Agree with asking a human, that seems like a simple question for someone in that space. I'd only use the chatbot to research the web for both professional and DIY active forums, discords, and other niche online communities in the following categories: [type-of?] engineering, 3d printing, weightlifting, and body-building

Once you have that list browse through them to find which ones seem friendly to new users asking a one off engineering question to probably never being heard from again; then post your technical question about the thing and wait, probably for much longer than an LLM but if an answer comes back it's probably more useful. I included weightlifting/body building since its barbell related -- there is a level of technical detail and precision many users in those spaces have when it comes to anything workout related, including equipment and modifications.

Other than that, set your llm to gacha-game agent-mode and connect it to some cad tools (I assume they exist, not my area). Then have it iterate on multiple approaches to the specs until it meets the requirements for each of them and spilts outs paragraphs of verbose self-congratulations and a cache of files for you to review. Maybe it will reward you with the solution you are looking for the first time (w/ the added rare chance it completes all the technical work to your standard, including all of the proper working files to feed a 3d printer)

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.