Alessio Fanelli: belief

Transcript context

…Workhorses. Yes. And it is doing better than Amazon Textract and all the other state-of-the-art. And it's a tiny, tiny model that does this. And it's really interesting. It's a combination of Omar Khattab's retrieval approach on top of a vision model, which I was severely underestimating PolyGemma when it came out, but it continues to come up. There's a lot of trends. And again, this is making a lot of progress here just in terms of their applications in real-world use cases. These are small models, but they're very, very capable. And they're a very good basis to build things like CopolyGemma. Yeah, no, Google has been doing great. I think maybe a lot of people initially wrote them off, but between some of the Gemini Nano stuff, like Gemma 2, PolyGemma, we'll talk about some of the KV cache and context caching. Yeah, yeah, that's a rag horse. There's a lot to like. And our friend Logan is over there now. He's excited about everything they got going on. I think there's a little bit of a fight between AI Studio and Vertex. And what Logan represents is, so he's moved from DevRel to PM, and he was PM for the Gemma 2 launch. Vertex has this reputation of being extremely hard to use. It's one reason why GCP has kind of fallen behind a little bit. And so AI Studio represents like the developer-friendly version of this, like the Netlify or Vercel to the AWS, right? And I think it's Google's chance to reinvent itself for this audience, for the AI engineering audience that doesn't want like five levels of off IDs and org IDs and policy permissions just to get something going. True, true.…

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