unspecified @strangersea - Tumblr Blog

I don’t have it in me to go into details right now, but I’m really struggling financially. I’m having a family crisis that is costing me a lot of money and my seasonal job ends in under two weeks. Any help would be greatly appreciated.

Paypal: [email protected]

CashApp: $AleshaKills

nerdreamer

#q

mioxenoblade

When I say “free water, free food, free shelter, free healthcare, free education for everyone” in that “everyone” I even include the people I hate. Too many people get surprised at the idea that I do wish for the people I hate to have better lives.

#q

pancakeke

google doesnt give a shit what you’re trying to search any more. it’s like “I didnt bother using half of your search terms but here’s pinterest and wikihow. enjoy”

blenderman-deactivated20230314

derinthescarletpescatarian

I don’t know anything at all abut computers but

You know how, when you first talk to a new chatbot (the kind that are designed to be trained as they go), it’s kind of limited and stilted and obviously AI? And then, once it’s been online and trained up and tweaked for awhile, it gets really convincing and talks mostly like a person? And then after a golden period it rapidly descends into incoherence, with a few phrases dominating regardless of conversation context and everything else being sentence fragments and non-sequitors?

The last few years of using google have kind of felt like slipping into that incoherence stage.

narcissus-the-sky

it pretty much is what the above person is saying. a few years ago they switched to usimg the ‘bidirectional encoder representations from transformers’ (BERT) algorhythm, which used AI to scan your quety f9rwards and backwards searchong fpr intent, rather than just using the meaning of the individual words ypu type. few problems with this. one, it’s shit. two, it is actively ignoring what you say in favour of trying to figure out what it THINKS you’re trying to say. great leap forwatd for AI, and great slip backwards for clear communication vs essentually being gaslit by your computer. three, it is trying to work out ypur meaning according to the indescribably vast pool of search data from all of history, around the world. which is great if you want to search for the same things as everyone else in history arpund the world has. it’s using previous searches to predict your intent. so if yoy want to search for 'can the measles vaccine cause blood clots’, congratulations, ypy’re going to get a load of results about the covid vaccine. you didn’t mention it but millions of other people did, so that’s probably what you wanted, right? how on earth could google have made such a niave mistake? simple. it’s no longer invested in showing you what you want; it’s trying to show you what you should want. looking for an independant crafts shop? nobpdy searches that, you’re having amazon instead. searching for a specific live journal, by name? yeah wrll it isn’t so bothered abput names now, and your intention is presumably to read a blog, so here are 50 links to the same huge blog sites owned by the same huge multimedia corporations. you asked for a biography of a local photographer but running your query against every similar one in history, turns out more people use pinterest and wikipedia than that random guy you mentioned, so obviously you meant to search for 'photographs’ on pinterest or 'photography’ on wikepedia, right? essentially, the old algorhythm was like ordering grocery delivery by saying what you want to buy and accepting that sometimes they won’t have certain items in stock and will send something similar instead. the new one is saying what ypu want to buy and it analyses your shopping list to figure out which items you’re most likely to want, according to how previous shoppers have purchased, and sending ypu that instead. and call me cynical, but i think the latter method is going to lead to an awful lot of 'you want coca cola, you just didn’t know it’. sooner or later ypu won’t even bother to search for that off-brand cola you used to drink, it never comes up anyway. what a shame. you can read abpit the model at google’s own blog at https://blog.google/products/search/search-language-understanding-bert/ which is a simple, clear explanation of the algorhythm and also the answer to the question 'what will happen when AI starts correcting humans?’. the answer is profit.

pancakeke

thank you for this explanation! but now I want to start killing people at google

#q