Among a strong field, this is the single most depressing comment I’ve ever read on Hacker News. Several grim components but it’s the “I don’t understand why” which seals the deal.
reply[dead]
replyhow is it depressing? that seems a tad strong. Maybe disappointment is the correct feeling
replyWhy? Calling a reasonable thing grim without any follow-up isn’t the hallmark of a good comment either.
replyIt is not remotely reasonable to ask "but why didn't he feed it to ChatGPT?". It is pretty silly to assume that ChatGPT should always be consulted.
replyIt's a good starting place. As an analogy imagine someone wanted to look up the definition of a word. If someone wanted to know the definition and they went out and crawled the entire internet, built an LLM, and then asked it for the definition you would wonder why they didn't check an existing dictionary first. I wouldn't consider it silly at all to always check a dictionary or existing LLM first when you want to know the definition of a word.
replyWanting to know the definition of a word is not an original problem. Similarly wanting to know what's in an image is not a new problem either.
I understand this is more about the process than the result, but note that: a) his result is completely wrong, he identifies a living land snail as the 100+MYa fossil; b) a conversation with Claude helps with some decent knowledge and provides a few possible candidates, that can be then double checked. c) Claude could have talked the author out of trying to identify a Jurassic fossil against a database of living species.
replyI trust a proper solution (even though I can be certain how accurate it is), which compares to a known dataset much more than just giving it an AI. For identifying current living species it is probably fine but this is something to nice for an AI to be trustable. Also this path is much more fun and you learn sonething along the way!
replybut, from my understanding what the author was really wanting was an adventure and to learn new things. he gained so much more than just learning what type of shell it is
replyMaybe he's not an idiot?
replyWho says the whole analysis isn’t AI inspired?
replyNo one. I'm pointing out there are existing AI models that can do this that the author could have tried before investing all the work to build his own.
reply> investing all the work to build his own.
replyI don't think you understand why the author did this on a fundamental level. Sometimes it isn't explicitly about getting the outcome directly, it's about putting in the work to understand how you get there.
The article says that the author wanted to have an analysis done, but ruled out the option of a human doing it. He did not rule out using a multimodal LLM. It's possible that he just is unaware that multimodal LLMs are capable of doing an analysis. Considering the approach wasn't mentioned in the article one can guess that the author just didn't know it was possible. But there is no way to know for sure.
replyThe AI would confidently give him the wrong answer, since it has no way to provide the correct answer, and doesn't know its own limitations. (Or however you wish to describe "hallucinations", which is about as accurate as my description ;))
replyAnd he would think he has the right answer, perhaps write up an essay about his findings, which later AI bots will read and learn from, propgating the mistake...
The AI would confidently give him the wrong answer
replyThere is irony here that does not sleep.
Wait, the author identified the shell as "Sphincterochila candidissima". Which is a living species of air-breathing land snail, a terrestrial pulmonate gastropod mollusk. Completely off.
replyBecause it’s much more fun that way
replyIs this example of vector search not "AI" enough?
replyGenAI is the new AI, now, unfortunately. PapersWithCode died for this.
reply