Speaker unverified: belief

Transcript context

…Yeah, so we talked a lot about the transformer architecture and the autoregressive transformer architecture specifically, like GPT. And it doesn’t mean no one else is working on anything else. People are always on the lookout for the next big thing, because I think it would be almost stupid not to. Sure, right now the transformer architecture is the thing and it works best, but it’s always a good idea to not put all your eggs into one basket. People are developing alternatives to the autoregressive transformer. One of them would be, for example, text diffusion models. And listeners may know diffusion models from image generation, like Stable Diffusion popularized it. Back then, people used GANs, Generative Adversarial Networks. And then there was this diffusion process where you iteratively de-noise an image, and that resulted in really good quality images over time. Other companies build their own diffusion models. And now people are like, “Okay, can we try this also for text?” It doesn’t make intuitive sense yet because it feels like it’s not something continuous like a pixel that we can differentiate. It’s discrete text, so how do we implement that de-noising process? But it’s kind of similar to the BERT models by Google. When you go back to the original transformer, there were the encoder and the decoder. The decoder is what we are using right now in GPT and so forth. The encoder is more like a parallel technique where you have multiple tokens that you fill in in parallel. GPT models do autoregressive completion one token at a time. In BERT models, you have a sentence that has gaps—you mask them out—and then one iteration is filling in those gaps. And text diffusion is kind of like that, where you are starting with some random text, and then you are filling in the missing parts or refining them iteratively over multiple iterations. The cool thing here is that this can do multiple tokens at the same time, so it has the promise of being more efficient. Now, the trade-off is, of course, how good is the quality? It might be faster, but the more de-noising steps you do, the better the text becomes. People are trying to see if that is a valid alternative to the autoregressive model in terms of giving you the same quality for less compute. Right now, there are papers that suggest if you want to get the same quality, you have to crank up the de-noising steps and then you end up spending the same compute you would spend on an autoregressive model. The other downside is that while it’s parallel, some tasks are not. For reasoning tasks or tool use where you have to ask a code interpreter to give you an intermediate result, it is kind of tricky with diffusion models. So there are some hybrids. But the main idea is how can we parallelize it. It’s an interesting avenue. I think right now there are mostly research models out there, like LaMDA and some other ones. ls. So there are some hybrids. But the main idea is how can we parallelize it. It’s an interesting avenue. I think right now there are mostly research models out there, like LaMDA and some other ones. I saw some by startups, some deployed models, but there is no big diffusion model at scale yet on the level of Gemini or ChatGPT. But there was an announcement by Google where they said they are launching Gemini Diffusion, and they put it into context of their Nano 2 model. They said for the same quality on most benchmarks, we can generate things much faster. I don’t think the text diffusion model is going to replace autoregressive LLMs, but it will be something for quick, cheap, at-scale tasks. Maybe the free tier in the future will be something like that. I think there are a couple of examples where it’s actually started to be used. To paint an example of why this is so much better: when a model like GPT-5 takes time to respond, it’s generating one token at a time. This diffusion idea is essentially generating all of those tokens in the completion in one batch, which is why it could be way faster. The startups I’m hearing are code startups where you have a codebase and somebody is effectively vibe coding. They say, “Make this change,” and a code diff is essentially a huge reply from the model. It doesn’t have to have that much external context, and you can get it really fast by using these diffusion models. They use text diffusion to generate really long diffs because doing it with an autoregressive model would take minutes, and that time causes a lot of churn for a user-facing product. Every second, you lose users. So I think that it’s going to be this thing where it’s going to- -grow and have some applications, but I actually thought that different types of models were going to be used for different things sooner than they have been. I think the tool use point is the one that’s stopping them from being most general purpose because, with something like Claude Code or ChatGPT with search, the autoregressive chain is interrupted with an external tool, and I don’t know how to do that with the diffusion setup.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.