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

How will OpenAI compete?

ben-evans.com

Post reply on HN
Print view View original

681–690 of 693 posts

Re: How will OpenAI compete?

#681

Post by lelanthran »

energy123 wrote: post #675
lelanthran wrote: post #665

Earlier quoted context omitted.

> I made 2 posts in this thread regarding why I think they have a moat. Was there anything ambiguous or that you disagreed with? I'm afraid I don't see those posts; I see 2x posts from you asserting they have a moat, but not why you think they have a moat. I distinguish between "They have a moat." and "This is why $FOO, $BAR and $BAZ forms a moat." Maybe you think brand recognition is a moat, but that didn't work out…

It was kind of buried in my second post: > They have internal scale and scope economies as the breadth of synthetic data expands. These frontier labs will have a hundred or a thousand teams of people+AI working in parallel generating synthetic data to solve different niches. A few teams solve computer use. A few teams solve math. A few teams solve various games. So the org is basically a big machine that mints data,…

I disagree that the model is a moat; distillation of models is going to happen, and even without it all the current players have models that are virtually indistinguishable for the use-case.

Model capbilities have converged over time, and I don't see this trend reversing. OpenAI owns only the model.

The provider who does have a moat is Google - they own the entire vertical, from the hardware, to the training data, they have it all.

OpenAI has to buy GPUs, Google makes them.

OpenAI has to rent data centers. Google owns them.

OpenAI has to scrape the web for all training data. Google's collection of user emails (not counting their Android data harvesting, ad data harvesting user-tracking, etc) alone gives them a ton of training data which will never be available to scrapers.

Google has billions of signed-in users, OpenAI has to market to and attract users (800m user count last I checked, but also last I checked that growth was asymptotic and flattening out).

Thats what a moat looks like. Better technology and/or results has never been, in my memory, a moat.

Re: How will OpenAI compete?

#682

Post by gck1 »

gverrilla wrote: post #472
gck1 wrote: post #319

Earlier quoted context omitted.

Every proprietary harness is just proprietary junk without ability to extend it without polluting context. This includes claude-code, gemini-cli, codex etc. They have tools which hardcode the behavior that is impossible to modify, they add tools you may not need that pollute context, they inject an entire textbook's worth of words into the system prompt which pollutes context, they provide zero observability into wha…

Great reply, thank you! I can see all the problems you mention, but I haven't started playing with the harnesses myself yet. Will read that repo when I get there. Before reading you reply, I was under the vague impression that a harness really needed a lot of bells and whistles, and that it would be hard for FOSS to compete at pace with Claude Code or similar because of that only. But I see now there's a different pa…

It becomes much clearer when you realize that these harnesses are basically JSON-line parsers under the hood that forward SSE data to commands and contain human readable instructions like "make no mistakes".

Re: How will OpenAI compete?

#683

Post by energy123 »

lelanthran wrote: post #681
energy123 wrote: post #675

Earlier quoted context omitted.

It was kind of buried in my second post: > They have internal scale and scope economies as the breadth of synthetic data expands. These frontier labs will have a hundred or a thousand teams of people+AI working in parallel generating synthetic data to solve different niches. A few teams solve computer use. A few teams solve math. A few teams solve various games. So the org is basically a big machine that mints data,…

I disagree that the model is a moat; distillation of models is going to happen, and even without it all the current players have models that are virtually indistinguishable for the use-case. Model capbilities have converged over time, and I don't see this trend reversing. OpenAI owns only the model. The provider who does have a moat is Google - they own the entire vertical, from the hardware, to the training data, th…

Good points about Google.

I think where I don't agree is about the model. You're mostly correct right now, and your view is supported by how close everyone is.

Where I am more optimistic about the 2-4 biggest labs (not just OpenAI) is what the next 2 years looks like.

I expect this to happen:

- Synthetic data goes from 30% of training data to 90-97%+ of training data.

- Synthetic data becomes hugely varied, and the production of it is factory-like and parallelized.

The moat here is the data factory, and the scale/scope economies behind it.

Thoughts?

Re: How will OpenAI compete?

#684

Post by codethief »

harambae wrote: post #655
jcbrand wrote: post #447

Earlier quoted context omitted.

I still use perplexity. Which tool is better currently?

I’m also unclear on what’s better than perplexity if you want accurate information (and not just to write Harry Potter fan fiction or whatever) I finally switched off ChatGPT premium when I asked a simple question (“which terminal is this airline”) and it was so confidently wrong. Perplexity referencing sources and trying to double check accuracy is great IMO.

Weird. I have used Perplexity various times over the years and every single time it was confidently wrong about a good 50% of what it was saying. In particular, it would cite references that said the exact opposite of what it claimed they said, or references that had nothing to do with the topic at hand and were only tangentially related, etc. My coworkers have reported the same, so it's definitely not just me.

In short, I really don't know where Perplexity's reputation of "being accurate" comes from. It's anything but.

Re: How will OpenAI compete?

#685

Post by neya »

biztos wrote: post #560
neya wrote: post #382

Earlier quoted context omitted.

Yes, but still, targeting is done even in billboards based on the location's demographics based on census data. It's not random. Some countries in Asia (like Singapore, Malaysia) have digital bill boards to target certain demographics based on the time of the day or the estimated crowd demographic at a given bus stop. And a few of them even track eyeballs to count "views" of the ad.

I admit this is a factor I hadn't much considered. I'm sure at some point, if not already, the data collected by your phone will enable the equivalent of a tracking pixel on your physical location, so you can get personalized ads when you step into the subway car: the system will quickly evaluate which rider is most likely to spend money based on ads, and on what, and then an auction will be run in two nanoseconds an…

Amen to that!

Re: How will OpenAI compete?

#686

Post by lelanthran »

energy123 wrote: post #683
lelanthran wrote: post #681

Earlier quoted context omitted.

I disagree that the model is a moat; distillation of models is going to happen, and even without it all the current players have models that are virtually indistinguishable for the use-case. Model capbilities have converged over time, and I don't see this trend reversing. OpenAI owns only the model. The provider who does have a moat is Google - they own the entire vertical, from the hardware, to the training data, th…

Good points about Google. I think where I don't agree is about the model. You're mostly correct right now, and your view is supported by how close everyone is. Where I am more optimistic about the 2-4 biggest labs (not just OpenAI) is what the next 2 years looks like. I expect this to happen: - Synthetic data goes from 30% of training data to 90-97%+ of training data. - Synthetic data becomes hugely varied, and the p…

Look, I'm upvoting your posts in this thread because you make some good points, but I'm not really convinced that a) synthetic data will result in good models, nor that b) quality synthetic data can be generated by labs outside of those orgs that have a ton of user-info.

This is why I say that OpenAI has no moat - even if synthetic data (however it is generated) is 90% of training data, there are still only two possibilities:

1. Orgs like Google, Microsoft and Amazon have a ton of user-data with which to produce synthetic data (after all, it's not produced out of thin air).

and

2. You don't need a ton of real data to seed the synthetic generation.

In the first case, yes, that looks like a moat, but not for OpenAI, more like for Google, etc al.

In the second case, what's to stop an upstart from producing their own synthetic training data?

In either case, companies who provide only tokens (OpenAI, Anthropic, etc) don't have a moat. The moat is still the same as it was in the 90s - companies deeply embedded into users' workflows.

In my memory, like I said, I struggle to think of even a few successful moats that were technology. The moat is always something else.

Re: How will OpenAI compete?

#687

Post by sigmoid10 »

SlinkyOnStairs wrote: post #407
sigmoid10 wrote: post #219

Earlier quoted context omitted.

>Is she paying for it? That is the only question that matters in the end. Don't underestimate advertising. Noone pays for Facebook or Google search. Yet the ad business with a couple billion users seems profitable enough to fund frontier LLM research and inference infrastructure as a side-gig in these companies. Google only rushed out AI overview because they saw ChatGPT eating their market share in information retri…

> Don't underestimate advertising. OpenAI is talking out of their ass with their advertising plans. Meta and Google are an advertising duopoly, extremely anti-competitive, and basically defrauding their own customers. OpenAI can't just replicate that. Worse still is that OpenAI has no competitive edge. All the hype around their advertising plans is based on the idea that they can blend the ads right into the response…

If that were true Meta and Google wouldn't be so desperate to get in the game. And don't think that other nations would step in against abusive marketing practices. The EU has been battling uphill for decades and the only ones who had some moderate success for user rights are private groups like NYOB. There is no law that will save the old tech companies and they know it.

Re: How will OpenAI compete?

#688

Post by danielparsons »

danielparsons wrote: post #155

sammy boy needs to pull a rockefeller and buy up all the competitors. Maybe that's what all these backroom deals about datacentre investment will amount to...

This was poorly worded but indeed rockefeller was the richest man ever by selling a commodity. But he did it by merging with all his competitors. Not sure Altman could pull that off here because most of the other models are attached to massive tech co's that can use AI as a loss leader for other profitable services.

Re: How will OpenAI compete?

#689

Post by cowhax »

kergonath wrote: post #462
graemep wrote: post #457

Earlier quoted context omitted.

It have worked for Google for years, and that was without even the barrier of download in app, just going to a different URL.

Google was clearly superior fo a long time. They got close to 90% before enshitification started in earnest. We are not at that stage yet with AI chatbots. Also, Google benefited from being the default on mainstream OSes. When people have to download an application, getting one or the other does not take more effort. Yes, OpenAI being tightly integrated within Windows, Android, and iOS would be a moat. That’s not the…

We are at a point though where when average people think of "asking AI", they instinctively think of ChatGPT. That's a big thing.

All OpenAI has to do is not fall behind too much to the point where an alternative can generate enough hype to take the crown (see AltaVista and Google)

Re: How will OpenAI compete?

#690

Post by lern_too_spel »

mercer wrote: post #277
oblio wrote: post #133

Earlier quoted context omitted.

> Being a first mover doesn't guarantee getting to the golden goose, remember MySpace. MySpace, ICQ, Altavista, Dropbox, Yahoo, BlackBerry, Xerox Alto, Altair 8800, CP/M, WordStar, VisiCalc, the list is very long.

Hotmail is a good example too. I remember it being pretty ubiquitous, at least for the 'personal email' crowd, and it seemed implausible that people would give up on what was often their main email 'location' for another offering without being able to transfer their often important and personal stuff. then gmail came along.

This took 8 years. HoTMaiL had a long time to become a competitive product, but it just blew it over the course of a decade. https://venturebeat.com/business/gmail-hotmail-yahoo-email-u...
Post reply on HN