Hacker News Reader

Comment by sterlind | Hacker News Reader

I wonder if OpenAI knew they didn't have a moat, and that's why they've been moving so fast and opening ChatGPT publicly - making the most of their lead in the short time they have left.

I find it incredibly cathartic to see these massive tech companies and their gatekeepers get their lunch eaten by OSS.

View on HN

Taking the "no moat" argument at face value, I think it's important to remember that some of the largest players in AI are lobbying for regulation too.

Yep, regulating the ladder behind you is a classic monopolist move

If past performance is any indication, it's pretty safe to lobby for regulations in the US...

OpenAI doesn't need a moat and it's fine that they don't have one. From their charter:

> We will attempt to directly build safe and beneficial AGI, but will also consider our mission fulfilled if our work aids others to achieve this outcome.

This was from 2018 and they've taken large strides away from their originally stated mission. Overall, though, they should be happy to have made progress in what they set out to do.

That's all well and good. I suspect their investors have a pretty different idea about their positioning tho.

Yes...

> "we will attempt to directly build safe and beneficial AGI, but will also consider our mission fulfilled if our work aids others to achieve this outcome"

... has big "don't be evil" energy.

I believe the next step was "we can do little a evil, as a treat."

Oh yeah, sure. I'm not sure I care much about that though. MSFT had the opportunity to think through all of this before they invested, OpenAI itself has incredible sums of money, and employees get to work on things they care about.

If MSFT doesn't make anything on this investment—which it still might, given that a big chunk of its investment will likely go into Azure—then…okay.

Note that this implies that if anyone tries to build AGI that is not "safe and beneficial" by OpenAI standards, it's fair game to suppress.

That is the nature of the singularity. Progress moves faster than any one person or any one company can keep up with.

What happens when the society as a whole cannot keep up with progress? That's a scary thought.

Sorry for the dumb question. But in the context of the AI space, what is moat?

That is the million/billion+ dollar question. If find it and get there fast enough you can own the moat, and thus become rich.

Note that I am not in any way implying that a moat even exists. There may be some reason AI becomes a winner takes all scheme and nobody else should bother playing, but it is also possible that there is no way to make your product better than anyone else. Only time will tell.

Moat is a business term coined by Warren Buffett. It's a competitive advantage that isn't easily overcome and allows a company to earn high margins.

I don't think there are any examples in the context of AI. As the post says, no one in the AI space has a moat right now.

Historically, datasets have been a moat. Google had a massive head start from having a massive search index and user data. Then access to compute became the moat - fully training a trillion-parameter language model has only been in reach for megacorps. But now, there's a ton of publicly-available datasets, and LLaMA showed that you don't need massive numbers of parameters.

Reply on news.ycombinator.com