Hacker News Reader

Comment by hsjqllzlfkf | Hacker News Reader

This is Google's inability to innovate.

Google stated goal was the "organize the world's information". Nevertheless, they didn't come up with Wikipedia, the highest quality curated human-readable information repository. They also didn't come up with ChatGPT, arguably the first LLM good enough that can perform non trivial tasks of data recall and organization.

Google had early success with search. Then decided to cement their lead by just throwing money at "all the smartest people"[1] so that their competitors would have difficulty hiring. This put Google in the situation where it's better to spend its resources fighting to keep the advantage that they have (by controlling the Internet infrastructure; android, chrome, email, etc) than innovating.

Google is dead man walking, unless something very substantial changes.

[1] "all the smartest people", read, "academically successful but rule-obeying, unimaginative, and risk-averse people".

View on HN

> They also didn't come up with ChatGPT, arguably the first LLM good enough that can perform non trivial tasks of data recall and organization.

Not only did Google make the main breakthroughs that all LLMs still rely on today (mainly attention architecture, and large scale training of DNNs), but they had a chatbot similar to ChatGPT (Meena [1]) way before ChatGPT was a thing.

Granted: they didn't release it to the public, so there is a failure of innovation there, just not on the technical/machine learning side.

[1] https://ai.googleblog.com/2020/01/towards-conversational-age...

Xerox Parc also invented the laser printer, the transistor was invented in Bell Labs yet Intel (et al) take advantage of it. Google seems more like history now regarding search. Also, it doesn't seem good that the company founders are not there anymore. The last auccessful companies were run by the founders until a more mature point.

Innovating is "easy", getting it in production is though. If the Google models are so genius and perfect, and nobody can use or see them, it's like if they don't exist.

That should be very strong quotation marks. I would say that it is the other way around. We should be glad that not everything is immediately turned into a product.

Bingo. Getting it into production is a non-concern for non-entities with no reputation, but Google has a lot more to lose. The big guys watch and wait, and use their sizeable wallets when the time comes. I have no doubts that Apple is going to eat Meta's lunch soon, for example — a lunch Meta didn't even realize was there, for all their innovation and go-to-market.

Google puts all kinds of things into production, then takes them down again in a year. The reputation is already that google puts out experiments

They are terrified of another “people labeled as gorillas” incident. Not of later canceling the product, they don’t care about that.

Also, regarding search, there are simply worthy competitors now. This is a bigger deal than I think people credit for. There truly was a time, maybe even ~ 2 years ago where even though it was broken Google was still the best. Kagi, while costing money to use, it fantastic.

YouTube is just bollocks now too, it contains a stunning amount of junk content and clickbait.

If someone tells me they learned something from YouTube, I actively question their authority on the matter at all. I didn't feel this way 2 years ago. YT seemed to be a generally credible source of information. Words cannot describe how polluted my suggestions are. I actually find it disturbing how much fake, incorrect and bias information is being suggest to me all the time. If I took it all on face value I'd be an idiot.

I've noticed recently that even people I thought were pretty credible have upped the clickbait severely in recent times. I guess this is what it takes to get seen but it's at the point where I just don't open YouTube unless I already know what I'm looking for so I can avoid it.

I never, ever log in on YT, and I'll sometimes clear my cookies as well. I find that if I stick with channels I like and carefully consider the first two or three recommendations, I don't get too much junk. But yes, the clickbaitification of even good channels is getting pretty grating.

Yeah I'm sure they had a really good LLM, they just decided not to show it to anyone.

Google had LLMs before OpenAI. It also produced the core techniques for making it happen.

My take is that it was taken by its own weight. The government fines & incorporation that started to fear those fines slowed down the progress. How difficult is to launch a change in that environment? Same for the baseline -- people react to changes. You consider the baseline to be the Golden standard. How difficult is to launch a radical change considering that?

IMO, the only way out is to shrink the product. Split, shard, whatever. Be small enough compared to the rest so you are not the default target (Coca Cola, Procter & Gamble, etc. do it right).

> IMO, the only way out is to shrink the product. Split, shard, whatever

Microsoft and OpenAI nailed this in my opinion. The funding deal between the two was genius. OpenAI gets billions of dollars from Microsoft that it then spends on Azure.

Microsoft got priority access to ChatGPT, equity in OpenAI, high-scale testing of Azure, plus some margin on OpenAI's usage for training. At the same time, it _didn't_ get the legal liability of running ChatGPT. Maybe it could eventually be hit for allowing training models with copyright data on its servers, but that feels unlikely.

It got all of the access to cutting edge productized LLMs that it could have wanted, including financial upside, without the bad parts.

Are… are you blaming the government for Google’s enshittification? Because that’s the only factor related to its size you cite as causally connected to its decline.

Two parts. The government & organic size issues. Both contributing xx%.

Any high churn change is difficult to get out at that scale. Now add all corp-fear because you are under lense of all governments. You can't keep the startup mentality with that kind of environment.

> Wikipedia, the highest quality curated human-readable information repository

I have some doubts about the quality of Wikipedia content. Everything that can be politicised is politicised , everything that can be played in order for someone to advance his/her plans or point of view is played.

Apart from bad faith, some article lack relevant info or contain errors due the volunteer nature of it.

I know what you mean but if we consider that Wikipedia is the only such existing platform of that scope, it still technically is the one with the highest quality :)

Being of the "highest quality", in this case, is comparative to the landscape we have available. With that said I believe it may be the highest quality open information repository.

And if it's not - then what's better?

Your local university library and its ILL department.

Apples and oranges.

I don't see a way to access my local university library for that data in the same way Wikipedia is available.

I think it’s possible for Wikipedia to have certain flaws and for Wikipedia to be the highest quality curated human-readable information repository available.

Highest quality !== great quality

Wikipedia suffers from massive bias and subjectivity issues, and also from accuracy problems. Yet, all the attempts to do better fizzled out so far, and most non-niche knowledge and information repositories have the same or worse issues.

I think some subjects are just not fit to be learned by reading a biased encyclopedia article, such as political subjects. These subjects are better learned by instead reading forum debates, where several biased sides duke it out. A wiki article will just have the bias from one side.

If we look back at history, many of the earliest written works were published in this way, as a philosophical debate between different fictional characters.

However it seems increasingly difficult to find good online debate with several sides of an issue. People tend to silo themselves into their own bubbles, and denounce any platform who even dares to let an opponent speak his mind. I remember just a decade ago where you'd find all opinions on the spectrum on the same message boards. Now, people prefer to shelter themselves instead.

> Google is dead man walking, unless something very substantial changes.

Beyond search, they are controlling the Android ecosystem. Your predictions may be immature.

Android being open source means that any competitor can take it and run with it. Tbh, the main issue with that is the staffing required in terms of software engineers, but we’re Google to make the wrong move, I imagine Samsung would waste little time.

> Android being open source means that any competitor can take it and run with it.

Oh yeah we are really seeing all those successful forks taking off. /s

On a more serious note, the key parts are not open source, that's why all those alternative ROMs are having a hard time and nobody can really run Android without official builds outside a niche of enthusiasts who are fine with sacrificing certain functionality.

Android isn't open source. AOSP is open source, but isn't a complete operating system.

The main reason competitors don't take off (e.g. Tizen) is the chicken/egg problem of the Google Play Store (which can't be shipped without Android).

A few days ago, some people were discussing Schumpeter's idea that (very) large companies drive innovation. I'm curious what they would make of Google Search.

The core of Schumpeter's position here is that innovation requires making large long-term capital investments in risky projects, which small firms can't afford.

Things are somewhat different today then when Schumpeter was writing. On the one hand, governments and venture capital firms are now major sources of R&D funding. On the other hand, the growth of the financial sector has to some degree crowded out higher-risk investments within large firms.

The financialization of like, literally everything, makes most of that “classic” work of political economics people like to cite wobble badly. Hell, so does advertising if you look at the classic Austrian Econ stuff. It’s invisible in Mises’ Human Action (because it barely existed at the time! no fault of his I guess), and it causes massive tensions if you try to confront his framework with it.

Google's money is in ads, not search.

I'm still curious how they're going to properly monetise ChatGPT at scale. I can't wait for the AI ads...

And the vast majority of their ad money, comes from ads that show on searches. So if they lose search, they lose the ads too. So their money is in search.

This is conjecture, but I don't think it's a long shot: The rich are going to pay them to manipulate responses, exactly like happened with google. Whether it's ads or political influence, with the current state of wealth those who deep down only want to make money (and OpenAI has shown their colors, in my opinion) need to look upwards for those cash injections, not towards their consumers.

Right now they are producing value for consumers, in order to attract investors. Investors will want a return: the product will be the consumers.

You ask ChatGPT to write a poem and it replies with a poem about butter puff pastry biscuits made with layers of light, crisp puff pastry and expertly baked with nutty Gouda and mild Edam cheese which you can order with a click but in the style of Geoffrey Chaucer.

>Nevertheless, they didn't come up with Wikipedia, the highest quality curated human-readable information repository.

Wikipedia was good before the culture wars got to it.

Could you elaborate or provide examples? I have never seen what you're describing on Wikipedia.

The RFK Jr. article. Here’s the talk section:

https://en.m.wikipedia.org/wiki/Talk:Robert_F._Kennedy_Jr.

I find that talk page fascinating. I have only rarely looked into how the sausage is made over at Wikipedia, but I'm intrigued every time I do.

Not that anyone asked, but I think he is absolutely guilty of spreading vaccine propaganda - however, I think one editor has a very good point about how this should be approached. Here is what they said:

>I too would avoid the pejorative word "propaganda" on WP:TONE grounds. If they must be included, terms like "propaganda" and "conspiracy theories" should not be stated in wikivoice. Instead, it should be stated dispassionately what the sources say. See Deepak Chopra for some examples of more appropriate wording: "His discussions of quantum healing have been characterised as technobabble"; "The ideas Chopra promotes have regularly been criticized by medical and scientific professionals as pseudoscience." (emphasis mine). HappyWanderer15 (talk) 12:46, 17 June 2023 (UTC)

Note that the URL needs a period on the end. HN strips it. I was very confused by the blank page with no history.

Reply on news.ycombinator.com