55
I was a huge fan of Amazon from a usability perspective. Unmatched, I'd say!
But there are obvious reasons against it: Worker exploitation and the political situation.
Got to admit, I'm a "soft-quitter", still got my account and still order there as a last resort.
My latest purchase: Several different types of heavy-duty storage racks.
Method used for shelf type one (4x):
- Amazon for search & comparison
- tried the product on geizhals.de, but to my surprise, it omitted the manufacturer as an option!
- went directly to the manufacturer, who had a very decent paypal checkout integration (obviously, invoice / wire transfer and auto-fill form would be preferable)
Price compared to Amazon: Exactly the same.
Method used for shelf type two (2x):
- Wanted to try the new ChatGPT "Agent" to do a research based on a list of criteria. (I know, it's not European!)
- Quite happy with the results, which included a table with metrics such as "price per storage area". It did not include any Amazon results and showed mostly manufacturer stores
- It had this as the top result: https://juskys.de/products/2er-set-lagerregal-easy-160-x-80-x-40-cm#%3A%7E%3Atext=Mit+bis+zu+640+kg%2Cbelastbar
- I decided to go with a bigger version of that
- Checkout was just slightly quirky, but lost no more than 3 minutes compared to Amazon
This is what ChatGPT Agent generated. Is it a viable shopping method in general? We'll see. In any case, it's not European, and there are huge environmental issues with it. Similar results might be gained from asking online, like in a home community.
My biggest apprehension about using llms for shopping is that there is no reason to think the models won't sell priority/positive attribution as an alternative to traditional marketing. It's the only monetization scheme that I think makes the absurd valuations make any sense. Condition everyone to offload the tiresome tasks of comparison shopping by just telling the llms what they want. Once people form the impression that the top results are what they themselves would pick or at least close to it, then the llm's will take over the actual purchasing task. Then the stage is fully set to sell consumers as a product to various companies. Imagine if HP could move away from scammy ink subscriptions and instead pay OpenAI to get millions of people to buy their printers and ink anytime they ask chatgpt to find them a printer.
I recognize that this is borderline conspiratorial, but llm's smell like the next evolution of enshittification and removing that pesky rational actor of pricing models.
the transition from "actually" free access LLM/GPT to "the cost is your soul" will be really interesting, that's for sure.
It's not far-fetched at all - that's what happened with search engines. Lobotomising an LLM is not that easy, as we just saw with the strange Grok outbreaks after they tried to make it anti-woke. But they can work through training data, nudging it softly in a direction. I bet that what happened with early days SEO is already happening again: They optimise online content for influencing LLMs trained by it. When their shills and bots (also LLM driven, lol) say "shelf X is scientifically known to be very durable", that becomes a "likely thing to say", which is all an LLM is looking for.
What you add is the suspicion that the corporations behind LLMs influence this process more directly and get paid for it, either already or in the near future, and that seems likely.