Getting results from a LLM with no guarantee of correctness is like computational photography. It looks nice, but that thing is not real.

Instead, I'm using Kagi, and feeding it old school search queries. Which provides me results I want in the first three slots, all the time.

Why fix something ain't broken?

Search wasn't broken Google needed a business plan and chose web ads. That is what broke search. The bifurcation of interests between selling ads and serving search results.

Then don't use Google? There are alternative and promising search engines now. Google is not the only game in town.

I don't. I was simply pointing out the business model's effect on Google. I participate with a search that is uses a subscription model service.

Who cares what's real if it's useful? We use all the times fictions that are not "real" but are useful, such as law codes, corporations, religions, etc.

I hope Harari appreciators will one day realise that social constructs are real too

When you're doing scientific research, or looking for public transport; the difference can decide your career's future or likelihood of you going home at night.

careers are just social constructs; and I've not seen yet Google maps fail to find me a public transport

> careers are just social constructs;

Well, too bad that that social construct supplies you the tokens that allows you to buy food.

Or do you imply that food is not real, because everything leading to it is also unreal, even if it's physical (e.g.: money). If that's the case, try to survive without it for a couple of months.

> and I've not seen yet Google maps fail to find me a public transport

This is because Google Maps doesn't use LLMs, but standard and deterministic indexing methods working on hard data provided by governments.

Obvious answer: because it is too expensive to offer it for "free".

Even if you had the GPUs, compute and electricity cost money:

https://medium.com/@zodhyatech/how-much-energy-does-chatgpt-...

YouTube made a loss for many years because the storage and data costs were vastly higher than their revenue, despite offering the product for free.

Those were very different market conditions though

Even if there's enough GPUs they also needs to be available at a cost that makes it financially viable, both in terms of purchasing price, but also power and cooling. Environmentally it might not look to good to offer LLM answers to a wider audience, there are already concerns about the water consumption of ChatGPT in some regions, see: https://fortune.com/2023/09/09/ai-chatgpt-usage-fuels-spike-...

Microsoft does offer "Bing Chat" (derived from ChatGPT) integrated in their Edge browser, effectively free.