zlacker

While Google certainly has its faults and I am frustrated with its trajectory, I would bet money that this is an article pushing the nonsensical notion that AI language models will replace search engines.

Edit: based on the subtitle, it seems like I am correct.

>>chias+(OP)
How is that nonsensical? Seems quite likely to me, especially if they can ground them better in a factual knowledge base. Most search engine queries are looking for information.

>>rileyp+I
The problem is that this literally does not exist today. ChatGPT is a chatbot, it has no factual knowledge base. It's optimized for dialogue.

That's why it's absolutely nonsensical to see all these articles, because they are talking about a chatbot as if it's trivial to turn it into a search machine. That's an entirely different problem.

>>phailh+T4
ChatGPT is pretty much the definition of a troll. It's only mission in life is to keep you talking to it. To do that it has no problem with lying, confusing, and various other deceitful methods you wouldn't accept from a human.

>>chias+(OP)
Is it completely normal on HN to comment on articles you haven't read and to freely admit so?

>>rileyp+I
Most of the information people are looking for isn't really the sort of thing an LLM is set up to cover though. If I want information on anything since 2021, or to look into stuff near me, or access a price, a download, or information in visual form, or anything that's too niche for its corpus, ChatGPT is not good. When ChatGPT can offer an answer it's fast, but about as reliable as a search engine's spammiest result

And answers as text wrapped around a factual knowledge base just sounds like the sort of queries Google Home already handles in natural language form. That can certainly be improved upon, but it isn't obvious that the LLM engineers will beat the search engineers to it...