Hacker News Reader

Open AI doesn’t have any meaningful “products” they have good technology and they charge $20 a month for a subscription to a technology demo that doesn’t come close to covering their costs.

You think the employees at “Open” AI care about anything more than their large paychecks no matter what they say?

It is implied in OP's reply too, right? No word about 'hey we're giving powerful tech to a hyper-monopolist', it's only about your friends and the money

I’d be curious if $20/mo is or is not enough to cover their costs. It’s opaque what their inference setup is, or what the final economics of LLMs will be. With 30-100 Billion param models becoming available which are in spitting distance of gpt-4, and dedicated ASIC implementations… openAI might be charging too much.

Are you taking into account they are also paying far below market costs for compute or even what a large customer could negotiate?

So because their costs are lower than that of competitors, customers should pay more? Ok.

There costs are lower…for now. Do you think Microsoft is going to subsidize them forever?

GPUs will be cheaper later. GPUs have still been tracking Moore's law, enterprise GPUs went from a niche product to being the hottest enterprise hardware overnight. Margins on H200's are massive from what I understand. I would bet that we'll see massive competition in the hardware space from AMD, NVDA, MSFT, AMZN, and GOOG + startups.

We're likely to see an 8-16x reduction in hardware costs over the next decade. That's about enough to shrink the current GPT-4 inference endpoints to a laptop/desktop sized hardware footprint.

Can you provide a source for this?

Since it is MS providing the compute for equity, they would not have an incentive to discount.

* this *

notice how the one OpenAI employee posted he'd "consider it" if the comp was right. Ie, it's about the money.

Yep. Someone is going to get the money, I’d like it to be the employees.

> MS is still evil

All big corporations are "evil", as in, their decisionmaking is scaled and institutionalized enough to effectively implement the goal of "maximizing shareholder value" above what's good for you or society.

No, not all companies try to exploit their strength in one area to make me use some other product that I have no interest in.

I'm not talking about "hey, we have this product that you might like", I'm talking about "if you want to use this product, you must also use this product". There's also no technical reason for it, sometimes they will let me use it.

Not O.K.

> No, not all companies try to exploit their strength in one area to make me use some other product that I have no interest in.

I'm not sure of any large one that doesn't. The incentives are aligned such that it's hard for them not to. It's easy to find examples of similar (sometimes almost identical) behavior from most large companies.

Okay, what other company does it? You may say Apple & App Store but that's not the same.

Apple/iMessage/Safari

Google/Android/Chrome/YouTube

Meta/Messenger/Facebook/WhatsApp

...just to name a few.

I don't know if Safari is that important to Apple. I have the enhanced data security feature of iCloud turned on, and I can still access my stuff in Chrome on the iCloud web interface (via the auth flow on my phone). It's actually exceedingly well-thought out and a nice bone to throw to Windows users that otherwise use Apple stuff! They could have been Safari-only or Mac-only, but aren't, even for super niche features.

You don’t need to use any of Facebook’s other platforms to use WhatsApp. I’ve not had a Facebook account for nigh on a decade but use WhatsApp as my main messaging app.

Apple with iMessage onthe iPhone is about as similar as you can get, and was actually at the forefront of my mind when I wrong my comment. It's a chat platform that you have to use their product to be able to utilize.

Name one company of any size that isn’t doing the same thing?

Valve's a great example. They have overwhelming market share and could have long since tried to tighten the screws to make a buck, and haven't done anything of the sort. They've also launched numerous 1st party systems and hardware that they could have tried to shove on people (Steamdeck, Index, controllers, etc) but, again, have done nothing of the sort.

Valve is worth around 7.7 billion and from the best I can find, it doesn’t have a quarter of the gaming market. How much power could they yield

That figure is quite ridiculous. Even if one completely ignores Steam and all of their games/products, even CS:GO alone is worth billions by itself, with millions of concurrent players (and growing) more than a decade after its launch!

All that said, I would expect Valve's total immediate-term 'worth' to be much less than it could be, which is the entire point of this topic. The reason companies treat their customers poorly is because it's profitable, but it also has consequences. You see large short to mid term growth in exchange for long term failure. That's why I think it's more typical to see public companies treat their customers poorly than it is for private companies. Generally the "employees" (including board and CEO) of a public company are evaluated on their short to mid-term performance.

By contrast private companies, especially ones that remain private long beyond the point of being able to 'cash out' with a public offering, tend to be more focused on the long term viability. So it's expected that their 'value', whatever it may be, would be less than it would be if they acted in a hostile fashion to their users.

https://www.bloomberg.com/billionaires/profiles/gabe-newell/....

What organizations are interested are interested in “what is good for you or society”? Even if they start out that way, eventually corruption sets in. I’d rather have companies have a clear profit motive instead of a series of fake motives and increased regulatory capture.

I don't think any of that really affects the day-to-day at OpenAI if they got bought. There are plenty of counter-examples for forcing you to use Edge or ads in the Start menu. They made Typescript. They made VS Code. Those teams are just doing their thing, and it's very likely that OpenAI would be like that. Microsoft is past the world of "everyone must use our browser and OS", they really just want to sell any of their products if you'll buy them. Like, I use Outlook on my iPhone because my organization uses Office 365. That would have been unheard of a couple decades ago; Outlook and Windows was their moneymaker and if you wanted one you got both. But now, it's just teams doing their own thing.

The likely effect on day-to-day for OpenAI would have been a different promotion process, their stock being worth $0, and having to use a different videoconferencing solution. I worked at a startup that got bought by a huge company, and that's basically all that changed for us. Our team continues to work on our product, and I'm only vaguely aware of what the rest of the company even does. (I think they make servers? Like solder together chips and put them in the box.) Day-to-day doesn't change much; it's the same people, the same code, and the same goals. I didn't get bought by Microsoft, but I can't imagine that these strategic acquisitions are much different across the Fortune 500.

> Microsoft is past the world of "everyone must use our browser and OS"

OS maybe but browser? They're going to great lengths to push Edge. Including harassing customers trying to download Chrome or Firefox.

Yeah I don’t think that’s true at all. It’s a minor nightmare as well in an enterprise setting trying to lock down their attempts to change defaults, push copilot, lock down office smart services etc.

Yes it's a game of whack-a-mole.

Even though they are in fact considering the enterprise market as a key customer. I used to manage Apple devices (iPhone but also Mac which was much more work) for a long time and that was much harder.

For example, for a federated Apple ID you can still, after 5 years since release, not use federated Apple IDs if your Azure AD UPN and E-mail are not the same. In like 90% of enterprises this is not the case but they just won't fix this. Apple thinks we should just change our entire org for them. Well with Macs taking up 0,5% of our installed base that's just not going to happen (and this way it's not going to be much higher either).

Not to mention all the terrible Mac ports of windows software like VPN, Security etc that don't follow Mac concepts or installers. You have to pile workaround upon workaround.

I suspect many developers, including OpenAI engineers are using VSCode, maybe TypeScript for frontend engineers, GitHub for sure which had no real features like pull request approvers before Microsoft money came in. Less at OpenAI most likely, but C# and .NET are going strong with bleeding edge tech like .NET chisel containers support. And I've seen great support from their engineers working on OSS, best of FAANG.

Google created the meme that a company could be evil or not but we're past that age. Let's focus on the experiences, and while not forgetting the past (yes the 90s do not paint MS well), be forgiving. MS seems to continue to develop and even innovate on a lot of this tech, painting them as an evil dinosaur seems frankly ridiculous. Yes as with any large company, there will be parts that are good and parts that are not.

Disclaimer: I have never worked at Microsoft and own no stock directly though so have ETFs. And through OSS I have some friends there that I strongly respect and think are awesome and always think of them any time I see hate talk towards MS, which is unfortunately common...

Most of OpenAIs tech stack is AI toolage, but basically everything you interact with is written in Python. Nobody uses C# for anything serious, because why would you?

It’s not a bad language, it’s just not a very good language either. Need efficiency? Rust/C++. Need an all round language? Node or Python, which are less efficient, but still powerful enough to power Instagram and well OpenAI as far as Python goes and LEGO as far as Typescript goes. Realistically you’re looking to choose between C#, Java and Go. Both Java and Go are miles ahead of C# in terms of concurrency, I mean, C# is still stuck with “await” after all, and while I guess you can have a lengthy debate on C# vs Java, 9 gazillion large companies use Java while 0 use C#.

It’s not that C# is bad, like I said. It’s never really been better than it is now, it’s more a question of why would you ever use it? Even the C# developers at Microsoft admit to prototyping using Python because it’s just so much faster to build things with it, and while they do move things to C#, you have to wonder if they would if they weren’t working for Microsoft.

> painting them as an evil dinosaur seems frankly ridiculous.

Hard to understand this point of view with:

- vscode being "open source", conveniently leaving out flagship features like live share and Python extension

- Increase of "in your face, fuck you" dark patterns like forcing an account for Windows setup, forcing ms authenticator, etc

Not sure I quite follow the point, Microsoft maintains PyLance and Pyright, tools that have had a large positive impact on Python development, including AI. Not sure about live share, but if there are feature gaps with paid IDEs I would say they're not so large - the functionality available for free in VSCode or its ecosystem (enabled by Microsoft with the help of many OSS developers) has helped democratize software engineering in a great way.

Yes they have some problems in hiding their attempts for lockin on the Windows side but ignoring all the good and just labeling them terrible for that is troubling.

> Not sure I quite follow the point, Microsoft maintains PyLance and Pyright, tools that have had a large positive impact on Python development,

I think GP's point is that PyLance and Pyright, while free, are still proprietary and not open-source.