There is still the option to select ‘Web’ from the top menu in the search results, but I am not sure if the indexing behind the scenes is what it once was.
The web crawling bit is still largely the same. The ranking (which was always the strong feature of Google’s search) is using Gemini now.
Yeah, the ranking has gone to pot. Things that used to be easy to search have become really hard to find. And pages and pages of results have been limited to a page or two.
Yes, i used to use Google Assistant from time to time. I didn’t use it at all when it first came out, because the ratio of creepy-to-useful was too high. But then i started using it for setting alarms and initiating navigation. And then they replaced it with Gemini, and it got too hard to do those basic tasks, and i type then into my alarm app and Google maps again. It’s just easier to open the app than to get into a conversation with a bot that asks a string of questions.
I often now can’t get Google to find something even when I know exactly what it looks like - like if I remember the tagline or slogan for some company, but not their name - it just doesn’t show up in the regular search and the Gemini-assisted thing just gaslights me by saying it doesn’t exist, or the words I typed in actually mean something different (and often completely false and stupid).
Can you explain this more, please? For example, do you mean they re-rank the results using Gemini?
(Not arguing, just asking)
I’m actually finding conflicting information about this. Google used to rank web pages based on a variety of factors such as the number of other web pages that linked into them (plus factors that are probably trade secrets to prevent competition and to try to prevent people gaming that system to try to promote their own websites in the ranking - which is a whole thing).
This used to be a fairly effective way of making sure the more relevant results for your google searches would appear near the top, so you wouldn’t have to scroll down too far to find what you were looking for.
Some sources are asserting that nothing has changed and Google still does this.
Other sources claim that Google is now using it’s AI model (Gemini) to decide how relevant the search results are, bypassing the ranking system or modifying it.
My own experience of how less effective Google search results have become, suggests the latter may be the truth. Or maybe it’s just that the whole web is choked with generated slop now, so even a traditional page ranking system would fail.
No, sorry, I didn’t explain with enough detail. There still needs to be AI processing on the edge device: speech recognition, translation, intent classification, and local command handlers, but after that if the user’s request needs to be handled by the Internet, then get it to the cloud as quickly as possible and execute the orchestration of the AI agents in the cloud.
If the user wants to plan a vacation, or research products, summarize docs, ask about a news item; a local LLM can do it, but not as well as a cloud solution with it’s thinking steps and access to web sites. I think users will expect the best results for these tasks and they’ll sit through ads or give away their data.
A local LLM is fine for use-cases like “give the user 3 suggested replies to this text message”. It’s good enough for translation between a subset of languages. Other use-cases like spam detection, intent classification, scene recognition, etc. don’t need an LLM.
I have to brag that Moto X shipped in 2013 and it was the first phone or device to have a wake word (e.g. OK Google) and it had local voice commands. It even recognized the names in your address book.
Thanks. I had assumed that web search was unchanged and they just pasted in an ‘AI mode’ tab, but it makes sense that at some point (possibly already) they are using Gemini to ‘refine’ the results.
Come to think of it, that might explain the behavior I noticed a few weeks ago (which is still going on):