Skip to playerSkip to main content
  • 2 days ago
Transcript
00:00The story with Meta typically is they hire, you know, ex-AI expert from Apple. They've hired up
00:06so much talent. You'd think that they have an insane bench and that things are going smoothly.
00:10Does this suggest that things are not going so smoothly with Meta and their development of
00:15their AI offering? Well, in the first instance, when the story hit, there was a kind of strange
00:20algo spike and then drop in the shares, but we're now down one and a half percent. So it's difficult
00:25to interpret the market's read on Meta shifting to closed model. What Kurt Wagner and Riley
00:31Griffin, our colleagues, are reporting is a lot of detail about the specific model, Avocado,
00:37which had originally been planned as an open model and, according to sources, may launch as a closed
00:42model. It's interesting in the technology development side because what Meta did was take models from
00:48others, including, you know, Alibaba's Gwen model, for example, and use that existing work to distill
00:54model and prepare their own model. That's something that, you know, Frontier Labs do all
00:59the time. But it's not new, the idea that Meta might go from a very clear open source approach
01:05to close. Mark Zuckerberg talked about it in July's earnings call, basically saying at some
01:10point in an open source world, the models are so big and the resource requirements are
01:15so high. What's the point? Academics don't have the funds to work at that scale. And really,
01:22it's only relevant to private industry if you're working on a model that has that level of parameter
01:28and requirement.
Be the first to comment
Add your comment

Recommended