Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google's live from Paris event private/deleted immediately

news.ycombinator.com

Post reply on HN
Print view View original

351–360 of 391 posts

Re: Google's live from Paris event private/deleted immediately

#351

Post by DOsinga »

vineyardmike wrote: post #311
sho_hn wrote: post #245

Earlier quoted context omitted.

I'd like to have more discussion about the attribution problem, specifically. We now have a couple of the players saying they're working on, or having demos. But from what I can see, all of these don't _actually_ attribute the source. They're just able to find _a_ source that fits to the output, working backward from it. Often the attributed source fits, but is actually disagreeing with the original output in specifi…

> But from what I can see, all of these don't _actually_ attribute the source. They're just able to find _a_ source that fits to the output, working backward from it I feel like it should be possible to build attribution "simply" by finding good results via traditional search (or even using LLM embeddings), then asking a LLM to summarize the sources. Then you can show the attribution. This would (to me) be much more…

I think this what the demo's with sources do. But they might still run into the same problem - the LLM summarization might be a bit wrong and so now it is saying one thing and the source has a slightly different opinion.

Re: Google's live from Paris event private/deleted immediately

#352

Post by amelius »

CyberDildonics wrote: post #343
amelius wrote: post #205

Earlier quoted context omitted.

Yeah, but it will cost them a lot more to run it.

Cannibalize means customers migrate to one thing and leave the other behind.

I think the old functionality of Google Search will still be available.

Re: Google's live from Paris event private/deleted immediately

#353

Post by immibis »

algorias wrote: post #178
treeman79 wrote: post #134

Earlier quoted context omitted.

Decades ago intel put out a press about the amazing new tech their CPUs would have. 3 of 5 CPU manufactures gave up. By the time the tech actually arrived, AMD had it as well. The others would have been fine. Sometimes a press release can make all the difference. (Sorry can’t find story that old)

Agreed. There's a lot of hype around LLMs right now, but they have not proven their value yet.

I think they have proven their value for creating spam and propaganda. There's a lot of money in convincing people the system isn't broken and they just need to work harder. The people who receive the profits those workers make can pay a lot for the propaganda that keeps them working.

Re: Google's live from Paris event private/deleted immediately

#355

Post by jjkaczor »

PaulDavisThe1st wrote: post #301
tarsinge wrote: post #284

Earlier quoted context omitted.

That thread is mostly about Google completely messing up their response to what Microsoft announced yesterday. I don’t see that unanimous acceptance, there are a lot of interesting discussions about the shortcomings. There are many comparisons between Google results for examples, with also potential solutions to problems you cite like https://news.ycombinator.com/item?id=34709883 It’s not about replacing search overn…

If it takes even 10% of search volume (in its current technological form) then those of us working in this domain have failed, miserably. I have no idea what Microsoft is thinking. It's like a state hiring a known liar to be its star witness in some notable number of court cases. [ EDIT: OK, so one thing they might be thinking is to just this sort of technology to generate readable/listenable summaries of search resu…

...It's like a state hiring a known liar...

Even worse than a star witness, the US let a known liar run their entire country for 4 years... So, frankly... people will love ChatGPT and other similar clones - people love bullshit.

Re: Google's live from Paris event private/deleted immediately

#356

Post by msikora »

patapong wrote: post #313
Wowfunhappy wrote: post #135

Earlier quoted context omitted.

> I feel like Google is going to bleed credibility for every month that they don't have a lamda/llm enabled search integrated into the home page. You can talk all you want about the foundational models and how advanced they are, but the search product itself will require a ton of fine-tuning on real-world training data that they aren't collecting yet. I love ChatGPT as a source of inspiration, but I don't think I'd w…

I disagree. The other day, I was debugging an issue in a web application. Only on iPhones, dragging an element around left odd residue on the screen. I googled the issue but was completely unable to find a solution, even though I consider myself a seasoned google user. I asked the question to ChatGPT, which immediately gave me the way to fix the issue. Having a ChatGPT answer next to the google result would have save…

Exactly same experience. ChatGPT is now my go to source for programming problems. It is usually correct (or correct enough) and when it is correct enough it saves a ton of time vs Google search and then going through dozens of StackOverflow or GitHub pages that may or may not have an answer.

Re: Google's live from Paris event private/deleted immediately

#357

Post by eric_cc »

PaulDavisThe1st wrote: post #294
pasquinelli wrote: post #264

Earlier quoted context omitted.

> If there's a crowd anywhere in the world large enough to leave 200+ comments on a post like this, you'd think thatg the one at HN would understand these models' currently unsolved problem with lying and fabulation. if i'm being frank, that doesn't match my experience with HN at all. chatgpt is the hot new thing, of course people here are preoccupied imagining a future powered with it. there are a bunch of different…

in general, i often post similar comments on HN, disputing the idea that there is such a thing as "HN groupthink". what puzzles me here is the the technical limitations of language models (i refuse to call them "large" because the sizes are going to change dramatically) seem so obvious and so well known, that i just find it hard to believe that there are more than a handful of people on HN who uncritically accept the…

> disputing the idea that there is such a thing as "HN groupthink".

There 100% is HN groupthink. Any doubts about it should have been erased by observing HN’s position on bitcoin and cryptocurrencies.

Re: Google's live from Paris event private/deleted immediately

#358

Post by jononor »

onion2k wrote: post #342
rchaud wrote: post #199

Earlier quoted context omitted.

> Over the next 20 years the war between Google and SEO spammers meant finding things got a bit harder. Along with that Google's proliferation of adverts on their search result pages meant it actually became harder to even find the website link I wanted in the page itself. Yes, SEO spam has largely made the search experience a lot worse and ripe for change. But ChatGPT crawls these terrible links too! Many of the lin…

So why would you trust GPT to provide an authoritative answer more than your own instincts? I wouldn't right now, but I believe the accuracy of an LLM is a function of its size and how much you can reduce the loss function. I think OpenAI can improve those things faster than Google based on the things that have been shown so far. I could be completely wrong though. I also suspect there's no benefit to corrupting a GP…

For ChatGPT and similar LLMs there is the loss function does not reward accuracy/truthfulness/correctness. But instead answers that seem correct to a superficial human check. These rewards making something plausible up instead of saying "I do not know", over-simplifying things, and passing off things with high confidence regardless of how much evidence backs the statement. Any successful deceit will bring down the loss. The overconfidence and lying is not due to model size beng too low - it is due to misalignment. If not corrected it will just get worse as size goes up. A recent Computerphile video with Robert Miles discussed this.

Re: Google's live from Paris event private/deleted immediately

#360

Post by schleck8 »

mlboss wrote: post #359

ChatGPT will kill Google in 2 main ways * Creating a better search experience without wading through the plethora of BS articles. * By enabling marketers to write BS articles faster which will further ruin Google results.

No it won't. Google's Bard (aka Lambda with search results input) performs almost identically to it from what I've heard and it's already being tested internally.
Post reply on HN