Great panel, really insightful! The knowledge gaps in training sets is a bit of a problem with the current frontier LLM's but i think with tuned SLM's like Phi, we should be able to bridge those gaps, especially with agents in the workflows. Katja is absolutely correct with her outlook on the human + AI collaboration, its something that we really need to get right, and if done well, the benefits to humanity would be immeasurable. Thanks for sharing MSR, and to all the researchers, engineers and staff working on these projects- well done!
can't wait till we get generative systems that can interact with symbolic ones, solve a problem once, and from then one it's reliable and cheap. obviously it's not a solve all, but the low hanging fruit is bountiful 😁
Great panel, really insightful!
The knowledge gaps in training sets is a bit of a problem with the current frontier LLM's but i think with tuned SLM's like Phi, we should be able to bridge those gaps, especially with agents in the workflows. Katja is absolutely correct with her outlook on the human + AI collaboration, its something that we really need to get right, and if done well, the benefits to humanity would be immeasurable.
Thanks for sharing MSR, and to all the researchers, engineers and staff working on these projects- well done!
can't wait till we get generative systems that can interact with symbolic ones, solve a problem once, and from then one it's reliable and cheap. obviously it's not a solve all, but the low hanging fruit is bountiful 😁
grounding 🤌