Is it common to use Internet search like that??? You're typing in literal questions to a search box rather than keywords, the name of the site you're looking for, or topics you want to read about. Maybe I'm just too old school, from the time where internet searches were essentially keyword searches, but it would have never occurred to me to type an actual english question as a full sentence into a search box.
If that's how most people use search engines these days, then I guess the transition into "type a prompt" will be smoother than I would have thought.
I’m quite sure it was common, because Google optimized for that over time, that’s why they switched to a semantic search instead of actual “contains” (remember they had a few questions and answers at the top way before ChatGPT).
Also if you type a few words on Google, it’ll “autocomplete” with the most common searches. Or you can just go to trends.google.com and explore search trends in real time.
I think those are examples of AI prompts, not search queries. Searching sometimes requires effort even for simple questions. For example, if you're trying to find the word for an object, you might need to consider what sort of website might talk about that, how to find that website in a sea of SEO spam, and then read through the article manually to find the specific information you are looking for. Using an AI, you can just ask "what is xyz called" and get a quick answer.
You can find the answer this way if the query is simple enough, but in general, if you are asking for something specific or trying to retrieve a piece of information based on keywords it's not usually indexed by, AI will do better. For example, "how are large concrete piers supporting a roadway constructed on a 45 degree slope?" Claude gave me an answer immediately, most Google results for my first two queries weren't specific enough/didn't include all details. I'm sure Google could find the answer, but asking Claude is just easier.
yes, but was it a good answer? were any sources backing up the answer credible (or even present)?
I don't know why we are suddenly so confortable trading speed for accuracy. Rule 0 of optiization involves making sure the probalem is actually being solved.
I'm not a highway engineer. I only care that the broad strokes are correct because I'm satisfying casual curiosity.
> I don't know why we are suddenly so confortable trading speed for accuracy.
Googling something or looking it up on Wikipedia is already trading speed for accuracy because you'll be reading a summary/reinterpretation. If I really cared about accuracy, I'd be reading highway design and civil engineering journals.
It's been common for the last decade. It's been a great way of finding forum/blog posts where the question is answered, even if phrased slightly different.
The questions above would be changed up for a Google search. The point is that LLMs can answer those questions pretty accurately now. I'm using LLMs to write technical cheat sheets for Linux sysadmin stuff, and to write a hobby website. I'm using search far less than before.
I have been using computers since the early 2000s, and I honestly don't remember the last time I searched Google for an answer to a specific question. It's incredibly inefficient compared to the even the most basic AI tool.
If that's how most people use search engines these days, then I guess the transition into "type a prompt" will be smoother than I would have thought.