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

How ChatGPT serves ads

buchodi.com

Post reply on HN
Print view View original

361–370 of 383 posts

Re: How ChatGPT serves ads

#361

Post by andai »

Hasz wrote: post #312

Ads is v1 of how-do-I-make-money. I wrote about this a while ago privately, but IMO LLMs are about to be on par with the printed word for distributing low-cost, high-impact propaganda. It has never been cheaper or easier to influence millions of people, either deniably-subtly (though omission, selective results, "hallucinations" etc) or via sock puppetting. If I am a government, there is nothing more valuable to me t…

>IMO, a key development will be having an "overlay" of weights to apply on top of a "clean" world model that is tailored to whatever interests can pay for it. Being able to serve that overlay dynamically, or atleast per-user is the killer app.

You mean LoRA?

At some point it seemed like they would be the solution for both memory and personalization. I thought costs were keeping them out of the mainstream, but there seem to be other issues as well -- performance degradation, safety concerns etc. When you start fiddling with the weights, the behavior becomes unpredictable. (The fine tuning endpoints appear to be powered by LoRA.)

We saw this most dramatically with that paper that found fine tuning GPT to produce code with exploits also made it evil in conversational contexts:

https://news.ycombinator.com/item?id=43176553

Re: How ChatGPT serves ads

#362

Post by andai »

programjames wrote: post #68

Less than two years ago, Sam Altman said > I kind of think of ads as a last resort for us for a business model. I would do it if it meant that was the only way to get everybody in the world access to great services, but if we can find something that doesn't do that, I'd prefer that. So, is this OpenAI announcing they're strapped for cash?

Well, they want to give everyone access for free. That's very explicitly their mission.

We don't seem to have invented a way of doing that which isn't ads.

Hence, every other online platform.

...Except this one, which is funded by... benevolence? :) Come to think of it, Archive.org and Wikipedia also seem to have found a way.

I don't think that model scales to "free LLM for everyone" though, at least not for another decade or two.

Re: How ChatGPT serves ads

#363

Post by idiotsecant »

fluoridation wrote: post #349
idiotsecant wrote: post #342

Earlier quoted context omitted.

There are lots of soldiers that don't rape and pillage when afforded the option to. There are plenty of good leaders who aren't sociopaths, it's just a career limiting feature. There are, in fact, a substantial proportion of us that aren't doing horrible things because they are comfortable enough that risking that comfort is worse than what they would gain.

>There are lots of soldiers that don't rape and pillage when afforded the option to. Sure, but you don't get stuff like the rape of Nanking from just a few handfuls of lunatics. It can't be simply explained as "oh, armies are just manned by 80% psychopaths, even after drafts". There's something about the extremeness of the situation that pushes an otherwise normal person towards abnormal behavior, even while some of…

>It's easy to say that without having gone through those experiences (either as a soldier or as a CEO).

I'm not sure what part of what I said is even remotely controversial. We see it literally every time the guardrails of society are relaxed and the typical social contract breaks down.

We are, as a species, riding the ragged edge of shit-slinging simian collapse. Humans were designed to exist in tribes of between 7 and 100 or so people. Any more than that relies of abstractions and heirarchy. The further up that heirarchy you go the less your world looks like the only expected human experience that our brains were designed for.

Re: How ChatGPT serves ads

#364

Post by keyle »

phist_mcgee wrote: post #203
Arkhaine_kupo wrote: post #198

Earlier quoted context omitted.

Its like fussion power, except there we half the funding every year instead of doubling it

Fusion power is proven to be possible. AGI is not.

That's ok, that's when you change the definition of AGI and claim success!

Re: How ChatGPT serves ads

#365

Post by fluoridation »

idiotsecant wrote: post #363
fluoridation wrote: post #349

Earlier quoted context omitted.

>There are lots of soldiers that don't rape and pillage when afforded the option to. Sure, but you don't get stuff like the rape of Nanking from just a few handfuls of lunatics. It can't be simply explained as "oh, armies are just manned by 80% psychopaths, even after drafts". There's something about the extremeness of the situation that pushes an otherwise normal person towards abnormal behavior, even while some of…

>It's easy to say that without having gone through those experiences (either as a soldier or as a CEO). I'm not sure what part of what I said is even remotely controversial. We see it literally every time the guardrails of society are relaxed and the typical social contract breaks down. We are, as a species, riding the ragged edge of shit-slinging simian collapse. Humans were designed to exist in tribes of between 7…

Ah, reading it again, I realize I misunderstood your meaning. Disregard my previous response to that sentence. Let me try that again:

>There are, in fact, a substantial proportion of us that aren't doing horrible things because they are comfortable enough that risking that comfort is worse than what they would gain.

That sounds like you're saying that most people don't "do horrible things" out of a utilitarian calculus (which, to some extent, I would agree with, depending what we include on that "horrible things" set), which would mean CEOs are acting just like normal people, except put in an unusual situation. But how do you reconcile that with your earlier statement that CEOs are sociopaths who are more dissimilar from normal folk than giant squids? Or did I change your mind already?

Re: How ChatGPT serves ads

#366

Post by dnnddidiej »

tombert wrote: post #333
dnnddidiej wrote: post #156

Earlier quoted context omitted.

Even as a not for profit they would need cashflow.

Yes but they would only need enough to keep the lights on and pay the engineers. When you're a for-profit company, especially a public one (which I believe they're looking to be soon), you can't just maintain homeostasis. Your investors want growth every quarter. Conceivably if they stayed non-profit then they could charge just enough to maintain the project, and they wouldn't necessarily have to have ads.

The lights being billions in hardware and plant investment, possibly power generation and operations and maintenance, attracting and retaining top 0.01% of engineers.

In addition if you don't keep up with SOTA +/- 10% you instantly lose all customers. There is zero stickiness.

Re: How ChatGPT serves ads

#367

Post by davidatbu »

masfuerte wrote: post #294
torben-friis wrote: post #62

These are the less worrying kind of ads in our future. Seeing how google has been fighting SEO for ages, what's going to happen when companies figure out how to inject ads into the model? We haven't yet seen the problem of adversarial content in play, I think.

> Seeing how google has been fighting SEO for ages I wish people would stop repeating this canard. Google gave up fighting SEO in about 2020. Emails that came out during antitrust discovery revealed that Google had decided to include advert-laden SEO trash in search results because it made them more money. This is why search quality has drastically declined in the last several years.

I'd love to see a link to these emails, if you have one handy!

Re: How ChatGPT serves ads

#368

Post by otabdeveloper4 »

cj wrote: post #351
otabdeveloper4 wrote: post #210

Earlier quoted context omitted.

> Doesn't history show us you just get both? No. "Opaque ads" are usually heavily regulated out of existence by government legislation.

Product placement in TV shows / movies is a $30 billion industry. They're opaque, and not regulated out of existence. They're so opaque that I'd wager 50%+ of people aren't aware it's happening. (Not fact checked) My favorite is Apple's "no villain" rule, where protagonists are allowed to use iPhone in movies, while antagonists are not.

> Product placement in TV shows / movies

...is a big exception in the advertising industry, not the norm.

Re: How ChatGPT serves ads

#370

Post by TZubiri »

otabdeveloper4 wrote: post #368
cj wrote: post #351

Earlier quoted context omitted.

Product placement in TV shows / movies is a $30 billion industry. They're opaque, and not regulated out of existence. They're so opaque that I'd wager 50%+ of people aren't aware it's happening. (Not fact checked) My favorite is Apple's "no villain" rule, where protagonists are allowed to use iPhone in movies, while antagonists are not.

> Product placement in TV shows / movies ...is a big exception in the advertising industry, not the norm.

Very common in netflix shows btw. They know they will be pirated, so they do product placement to monetize anyways.

So pirates get their wish of no ads granted, and they get propaganda instead.

Post reply on HN