Live data from Hacker News

Google has sent internet into 'spiral of decline', claims DeepMind co-founder

telegraph.co.uk

311–320 of 390 posts

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#311
post #110

I'd like to see some experiments with Search engines that rank on things that are hard or impossible to game. Impossible in the sense that the search-engine controls this variable entirely. Like "age of the article" through "time we saw it first & how much it changed since". Or like "churn", through "how often we saw new content come and go". Anything that users or owners control, will be gamed. So maybe there are va…

I'd like to see it possible to game search engines by having the fewest ads on the page. :)

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#312
post #191

Earlier quoted context omitted.

I dreamed of building a website like the type I grew up on in the late 90s and early 2000s. The gamefaqs, mrfixitonlines and casterrealms-type websites for me were at the forefront of making this internet medium to be better than books, voice or anything else for that matter for the transmission of information. Fast forward 20 years later. I think I've spent close to 300k on the website and team to help with it. I do…

I see your point and mostly sympathise. It's relevant to the topic and provides a personal perspective. >> ran by non-native English speakers I feel this clause doesn't help promote your grievance. A big promise of the Web was that it was World-Wide. If a site in say Thailand could exceed say a USA site's quality, all the better. I get that your specific competitors didn't excel - except in ranking. But, in principle…

Listen I watched the movie Hackers as a kid and want to throw my fist up and say yeah too but in reality that's not how it works. These sites in 'Thailand' are very seldom competing on quality in my experience.

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#313

I am interested to here what Kagi users think. Kagi is searching the same web. If the web is now garbage, then I wouldn’t expect improved search to yield improved results.

It's about the ordering of the results, but also about finding rare information.

I found a link to the age of a very minor lava flow in Oregon on the third link on Kagi. The answer was nowhere on the first 3 pages of Google, which is where I stopped looking.

If you deprioritize stuff that works for Google's SEO, you'll get different (better) results.

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#314

Earlier quoted context omitted.

Nah, they likely didn't bother trying to ask GPT-4 and imagined hypothetical answers which could lead beginners to shoot themselves in the foot. ChatGPT Plus with a short custom instruction to answer in steps, respond tersely, ask clarifying questions where needed, and make recommendations wherever possible replied to "how to reverse a pushed commit in git" with: --- To reverse a pushed commit in Git, you can either…

You are correct! I was fantasizing how a bad GPT answer could look like, similar to the errors that were in the answers to my prompts to GPT-3.5 for imagemagick. These prompts where precise and in good faith, with some rounds of clarification. I am sure newer GPT versions will do better, and I think for a tool as ubiquitous as git the answers from ChatGPT would have been perfectly fine too. Just wanted to illustrate…

C:\> debug g=c800:5

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#315
Back in my day, we had to hear about an interesting website from a friend, and then bookmark it — or remember the site name (including the www dot). And if we wanted to see a cat fall off a couch, we had to watch TV on Sunday night, or wait half the night for it to download, and then click page down repeatedly to make the ascii art appear animated — uphill both ways.

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#316
All of the top search results were generated yesterday, by design. We should be receiving relevant and accurate content, not content that is automatically regenerated constantly with fake intelligence to appear like ‘something new’ in order to game the search engines. I prefer information that has withstood the test if time, not the next thing that popped into some machine’s head.

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#317
post #199

Earlier quoted context omitted.

I'm rather shocked how the quality of their search results has tanked over the years. I used to be able to find decent academic content, or programming related topics quite consistently 5 years ago. Now I'm getting only a few vague results either from stackoverflow, math stackexchange, reddit, or apple, and the rest is just utter garbage. Yandex, of all search engines provide consistently better results. Fortunately,…

The Russian search engine pops up in the comments whenever Google search quality is discussed here. I am sure this is non-organic astroturfing.

I like kagi

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#318
post #221

Earlier quoted context omitted.

> Professional YouTubers are also making content for rankings above all, but the fact that it's their own face/voice keeps some of the rot at bay. YouTube's stuck in its own algorithm hellhole. Mostly due to their insistence that creators have to post on a regular basis, creative block or lack of things to cover be damned. So if your niche has nothing to talk about right now, well you'll need to find something, anyth…

Hmm, Jesse Michels posts videos so infrequently (he just did the first one in a year) and he still got 2 million views instantly..

I think the algorithm is geared now towards higher production value videos that encourage viewers to click on other videos

That’s why there are so many collabs now

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#319
post #82
post #75

Earlier quoted context omitted.

Yes, I think the clickbait sites are done long term, people will just ask the AI on their phone. The people who were capable of writing those large detailed posts on subjects have most likely already moved to youtube or moved on. The data that's actually useful to the LLM's will exist in places like Discord, reddit, twitter. It's pretty obvious that twitter and reddit are going to sell their data to the LLM companies…

Advertisers will simply begin to pay more money to inject their content on those platforms, to have it end up inside LLM models. It could be even worse, because it will be less obvious that the content we'll be consuming through LLMs is polarized/paid for. There will be less transparency.

It’s like in Google Maps it loves to let me know there is Sponsored Business at the corner I’m passing. AI in smartphones will be gross.

Re: Google has sent internet into 'spiral of decline', claims DeepMind co-founder

#320
post #226

Earlier quoted context omitted.

I use chatgpt for such things(and I'm working hard on getting an equivalent service quality from my local models as I hate the not-at-all-OpenAI, but they have the best model now). Then again I never know if the thing it says are true recipes or hallucinations. If you think about it, it's exactly like on the general Internet.

How well does that work? I would imagine recipes iate really hard to entrust to chatgpt. At least for baking, where ratios and timing is crucial. LLMs have very bad sense of numbers generally. How would you verify it’s not all bogus hallucinations? At best it’s a source of inspiration, for ingredients you can freestyle from yourself.

The internet is full of false recipes that can't possibly work, probably increasingly LLM generated themselves. Consider: Persimmon jam. Not possible short of some secret chemical process-- heating reverses the chemical reaction that eliminates the astringency of the unripe persimmon. Yet there are now countless recipes online that will lead you to inedible muck, all effectively bogus hallucinations. Not necessarily machine ones, but also just content farms manually making every conceivable permutation of every existing recipe, the author of those persimmon jam recipes has probably never seen the fruit first hand.

At least if you LLM them on your own you know to be dubious and also you're more likely to use a 70B model (or GPT4) which is a little more likely to get them right, vs whatever garbage the spammer used.

It would probably be pretty interesting for someone to make a concerned effort to digitize every recipe in print prior to 2005 or so-- the relevant threshold time isn't when LLMs came on the scene but when google made low effort content farming so profitable.

It wasn't guaranteed that every recipe printed pre-internet had actually been cooked by anyone, but it was a lot more likely.

Post reply on HN