Of course anecdotal but 2 months ago I tried to do that for a missing bolt in a popular scooter (roughly 7 years old, bought 2nd hand so didn't know exact specs). I fed it various images that I took + internet images. It found a bunch of local shops but always found the wrong part (but very similar looking). I double-checked everything multiple times, via multiple contexts, even different pre-prompts from various sources, and asking plenty of questions. We chatted in thinking mode for what felt like ages, and according to paragraphs explaining why it HAD to be the right one, multiple times, with evidence that it sometimes gladly generated in image form (often completely garbage imagery e.g. with half-drawn bolts extruding from multiple surfaces). Eventually it found something very plausible which I ordered. It was the wrong part.

I had to get someone on the phone to help me find and order the right part (which was on their website, for many years according to waybackmachien).

I love LLMs but it's still totally hit/miss what you get. I'd rather not give it write-access to my bank account just yet.

Conversely, I did something similar, but I took a picture of the model number, copy/pasted the text from the image into a search engine, and that was that. Either way you're taking a picture and performing a manual step (query an LLM, query a search engine)

I would never trust an LLM to accurately identify and purchase something for me based on a picture, a prompt, and a prayer.

ChatGPT doesn’t know what the best-converting dog food in Scranton, PA is. Amazon does.

Anything that starts with chat history presumes the theoretical limit for ad effectiveness is higher than it is now, and chat is a better way of getting there than actual purchase history. I have a feeling it’s not.

Imagine a person who shops on Amazon for basically everything. So theoretically Amazon should know a ton about them, more than enough to put together a profile on that person.

To say OpenAI could do a better job of selling products is to say they can do better than Amazon already does if you scroll through their personalized product recommendations. There is some better feed out there, or some better way of presenting the feed that Amazon hasn’t thought of.

I don’t doubt it can be marginally better.

I do doubt whether it can be double or triple digits better that can justify trillion dollar valuations. And I do doubt whether a model trained on Internet text rather than user interactions can do better.