#google

#Google putting wrong medical advice in their #AI Overviews.

This is not funny. It is not acceptable. Today I saw someone who had asked a health-related question to Google Search. The AI Overview presented mixed-up, wrong information. I urged the person to ignore the AI Overview and use the regular site links.

They didn't know there were any regular site links, because the AI Overview filled essentially their entire window and they didn't know to scroll down. This is a common situation with busy, nontechie users.

They have depended on Google to point them at accurate information for so many years, and now Google Search spews out convincing looking AI garbage. This is not an anomaly.

Google's AI Overviews are full of wrong, partially wrong (even more dangerous!), and just plain misinformation. Answers that are completely reversed from supposed source pages because the AI didn't understand the wording. Measurements wrong. Math wrong.

It doesn't matter how often AI Overviews are correct, because you NEVER KNOW when they're going to be wrong, either completely or partly (again, mixing true with false -- like contaminating a well).

And now Google is trying to convince users to use "AI Search" instead -- "Hey Ma', no more list of blue links!" -- making it even harder to see that so many of their answers are, if you'll excuse the expression, bulls*it, sometimes dangerous as well.

This is unconscionable. Frankly, whether Google understands this or not, this behavior is uncaring and evil. Apparently Google's leadership no longer feels any shame at all. Disgusting.

Okay, Back of the napkin math:
- There are probably 100 million sites and 1.5 billion pages worth indexing in a #search engine
- It takes about 1TB to #index 30 million pages.
- We only care about text on a page.

I define a page as worth indexing if:
- It is not a FAANG site
- It has at least one referrer (no DD Web)
- It's active

So, this means we need 40TB of fast data to make a good index for the internet. That's not "runs locally" sized, but it is nonprofit sized.

My size assumptions are basically as follows:
- #URL
- #TFIDF information
- Text #Embeddings
- Snippet

We can store an index for 30kb. So, for 40TB we can store an full internet index. That's about $500 in storage.

Access time becomes a problem. TFIDF for the whole internet can easily fit in ram. Even with #quantized embeddings, you can only fit 2 million per GB in ram.

Assuming you had enough RAM it could be fast: TF-IDF to get 100 million candidated, #FAISS to sort those, load snippets dynamically, potentially modify rank by referers etc.

6 128 MG #Framework #desktops each with 5tb HDs (plus one raspberry pi to sort the final condidates from the six machines) is enough to replace #Google. That's about $15k.

In two to three years this will be doable on a single machine for around $3k.

By the end of the decade it should be able to be run as an app on a powerful desktop

Three years after that it can run on a #laptop.

Three years after that it can run on a #cellphone.

By #2040 it's a background process on your cellphone.