Claude 3.7 Sonnet, BeeAI agents, Granite 3.2, and emergent misalignment

Granite 3.2 is officially here! In episode 44 of Mixture of Experts,
host Tim Hwang is joined by Kate Soule, Maya Murad and Kaoutar El
Maghraoui to debrief a few big AI announcements. Last week we covered
small vision-language models (VLMs), and this week Granite 3.2 dropped
with new VLMs, enhanced reasoning capabilities, and more! Kate takes us
under the hood to understand the new features and how they were
created. Next, Anthropic dropped a new intelligence model, Claude 3.7
Sonnet, and a new agentic coding tool, Claude Code. Why did Anthropic
release these separately? Then, as we cannot have an episode without
covering agents, Maya takes us through the new BeeAI agents! Finally,
can fine tuning on a malicious task lead to much broader misalignment?
Our experts analyze a new paper released on ‘Emergent misalignment.’ All
that and more on this week's episode!

00:01 – Intro

00:41 – Claude 3.7 Sonnet

11:58 – BeeAI agents

20:11– Granite 3.2

29:23 – Emergent misalignment

The
opinions expressed in this podcast are solely those of the participants
and do not necessarily reflect the views of IBM or any other
organization or entity.