Now that we are getting some clearer signals from the labor market about how AI may be affecting employment of different groups, @ide_enrique provides a terrific thread tying together the key results with his new paper on AI in the Knowledge Economy with Eduard Talamas.
MACHINE LEARNING
-
Voyager SDK 1.4.0 Release: Enhanced Pipeline Support and YOLOv10 Integration
By
–
Still on Voyager SDK 1.3? You might want to peek at v1.4.0 smoother cascade pipelines
more operator & metadata options
easier Model Zoo customisation
plug-in custom evaluators
QoL polish across debugging & benchmarking
support for YOLOv10 [n,s,b], FastSAM, -
torch.export maturing for inference graphs in PyTorch
By
–
it comes with its own temporary friction, torch.export is still maturing for backward/training graphs and all effort over the past two years was put on inference maturity
-
torch.export: Compile-time autotuning for training without runtime JIT
By
–
if you want to avoid guards and do autotuning at compile-time, have no jit at runtime, consider torch.export; thats what you'd want.
in fact, @ezyang has been pushing it for some training workloads that dont want any jitting at runtime and ones that need a bunch of sharding info -
Controlling Graph Reasoning Difficulty: Size and Operator Complexity
By
–
We control difficulty with: Graph size: more nodes/edges = more elements to reason over Operator complexity: different ops stress different reasoning skills (counting, paths, optimization)
-
Snorkel Graph Benchmark Shows Significant Room for Advancement
By
–
As the leaderboard shows, this benchmark is far from saturation ATM: https://
leaderboard.snorkel.ai/category/snork
elgraph
… We’re looking forward to seeing continued advances in this skill, particularly in open models. -
SnorkelGraph: Graph Reasoning Benchmark for LLMs
By
–
Operations in graph theory are fundamental to analyzing complex systems, and provide a multi-hop, long-context reasoning challenge for LLMs. That’s why we built SnorkelGraph: a benchmark for evaluating graph reasoning with verifiable ground truth. https://
leaderboard.snorkel.ai/category/snork
elgraph
… -
Modular Models and Forgetting in Multi-Modal AI Systems
By
–
Q: "what about forgetting as you add modalities/languages" A: optimistic on modular models (me: like Apple i think but not sure what other research is relavant. see also
-
Francine Bennett Uses Data Science To Make AI More Responsible
By
–
Francine Bennett Uses Data Science To Make AI More Responsible | http://
ow.ly/SucM30sOQyJ
#AI #AIio #AIInnovation #ML #DataScience #Futureofwork @demishassabis @Ronald_vanLoon @TamaraMcCleary @geoffreyhinton @goodfellow_ian @jeffdean @erikbryn -
Long-Horizon Tasks and Embodied AI Scaling Laws
By
–
(10/N) Together, let’s explore:
How close are we to solving long-horizon, complex, human-centric tasks?
How to efficiently combine low-level control and high-level planning?
What are the generalization limits of current models? Are there scaling laws for embodied AI?