I think any decent company would be using LLMs through an interface, so they can swap out providers, same as any API.
That's fair, but different LLM behave differently, so you would have to redo your testing from scratch if you were to swap the model. I think that would be the primary problem.
Testing for LLMs is an evolving practice but you need to have tests even if you stick with one provider, otherwise you won't be able to swap out models safely within that provider either.
One view is that it’s not first mover, but first arriver advantage. Whoever gets to AgI (the fabled city of gold or silver?, Ag pun intended) will achieve exponential gains from that point forward, and that serves as the moat in the limit. So you can think of it as buying a delayed moat, with an option price equivalent to the investment required until you get to that point in time. Either you believe in that view or…
Exponential gains from AGI requires recursive self improvement and the compute headroom to realize them. It's unclear if current LLM architectures make either of those possible.
People need to stop talking about "exponential" gains; these models don't even have the ability to improve themselves, let alone at this or that rate. And who wants them to be able to train themselves while being connected to the Internet anyway? I sure don't. All it takes for major disruption is superhuman ability at subhuman prices.
Genuinely curious if anyone has ideas how an LLM provider could moat AI. They feel interchangeable like bandwidth providers and seem like it will be a never ending game of leapfrog. I thought perhaps we’d end up with a small set of few top players just based on the scale of investment but now I’ve also seen impressive gains from much smaller companies & models, too.
Regulatory capture. Persuade governments that AI is so dangerous that it must be regulated, then help write those regulations to ensure no one else can follow you to the ladder. That's half the point of OpenAI's game of pretending each new thing they make is too dangerous to release. It's half directed at investors to build hype, half at government officials to build fear.
If Elon Musk's pals regulate away OpenAI because they declared their technology to be too dangerous that would be an ironic turn.
Why Meta decided to open source their Llama architecture and complete models? From a strategic business perspective, I'm trying to puzzle why they would give that up, a competitive advantage (even though it might not be as good as OpenAI product), rather than directly competing with OpenAI. What's the reasoning behind this decision?
Meta's is doing classical (and brilliant) "commodify your complements".
LLMs and AI is not where Meta adds value to the market. They add value via advertising. Therefore, AI is not Meta's product, and thus Meta should aim to make AI as cheap as possible.
They did the same thing for infra. Meta doesn't sell cloud computing, so they want to make infra as cheap as possible.
>> OpenAI currently offers, by far, the best product experience of any AI chatbot assistant. Claude is noticeably better
I use Claude because the model is indeed better, but having listened to John's podcast I know that he means that the experience of using OpenAI overall is better.
This rings true to my ears: > There is no technical moat in this field, and so OpenAI is the epicenter of an investment bubble. Thus, effectively, OpenAI is to this decade’s generative-AI revolution what Netscape was to the 1990s’ internet revolution. The revolution is real, but it’s ultimately going to be a commodity technology layer, not the foundation of a defensible proprietary moat. In 1995 investors mistakenly…
Why is everyone so sure that technology can’t be the moat? Why are you convinced that all AI systems will be very similar in performance and cost, hence be interchangeable? Isn’t Google the perfect example that you can create a moat by technological leadership in a nascent space? When Google came along it was 10X better than any alternative and they grabbed the whole market. Now their brand and market position make i…
ChatGPT isn't ten times better than Gemini or Claude, is it? Even if they magically released such a model, the competition would quickly catch up. The competition has similar or better resources.
Where are you getting that Deepseek is at the level of o1? In my experience, it's not even as good as Claude.
Even v3? edit: Fair enough. I'm fishing for opinions too.
My testing has been very limited, so I don't want to opine too much. If anyone has a differing opinion based on more testing, please share, I'm interested!
If we get nuclear fusion and better GPUs as a byproduct of enabling LLMs, great. We haven't had this kind of dynamic since the heft of Windows pushed Intel to deliver faster processors.
There is no technical moat, but that doesn't mean there isn't a moat. Amazon might be a good analogy here. I'm old enough to remember when Amazon absorbed billions of VC money, making losses year over year. Every day there was some new article about how insane it was. There's no technical moat around online sales. And lots of companies sell online. But Amazon is still the biggest (by a long long way) (at least in the…
A couple other comments touch on the point I want to make, but I feel they don't nail it hard enough: if today you told me you had a website that was 100% technologically identical to Amazon, but you were willing to make slightly less money, you'd still have no products... as a user I'd have no reason to go there, but then without users there is no reason to sell there, so you have a circular bootstrapping issue. Tha…
If Walmart made a slightly better website you wouldn’t shop there? Cause that’s really all that’s holding me back
This rings true to my ears: > There is no technical moat in this field, and so OpenAI is the epicenter of an investment bubble. Thus, effectively, OpenAI is to this decade’s generative-AI revolution what Netscape was to the 1990s’ internet revolution. The revolution is real, but it’s ultimately going to be a commodity technology layer, not the foundation of a defensible proprietary moat. In 1995 investors mistakenly…
Honestly the more I use ChatGPT, the more I see of their "moat-ish" opportunities. While the LLM itself may be ultimately reproducible, the user experience needs a UI/UX wrapper for most users. The future is not user exposure to "raw" LLM interactions but rather a complex workflow whereby a question goes through several iterations of agents, deterministic computation using "traditional" processing, and figuring out when to use what and have it all be seamless. ChatGPT is already doing this but it's still far from where it needs to be. It's totally possible for a company to dominate the market if they're able to orchestrate a streamlined user experience. Whether that will ultimately be OpenAI is an open question.