Robotics gains shared tools and better ways to learn
By Nathan Benaich (Air Street Capital) · 2024 report
Robotics became a larger part of the 2024 report as researchers worked around its persistent data bottleneck. Open tools made it easier to start, human demonstrations supplied training examples, and autonomous driving provided evidence of deployment at increasing scale.
Hugging Face’s LeRobot brought pretrained models, demonstration datasets, and development tools into a shared ecosystem. Robotics had fewer open resources than language or vision research. Making them easier to find and reuse reduced the amount each team needed to build before it could begin experimenting.
HumanPlus combined a shadowing system with imitation learning. A camera let a person control a humanoid’s whole body, and demonstrations then trained autonomous skills. The report described tasks learned with up to 40 demonstrations, while the underlying control policy also relied on motion data and simulation. The small demonstration count was not the system’s entire training budget.
Waymo reported 100,000 paid trips per week in the US in August 2024, and Wayve raised a $1.05 billion Series C. These were meaningful signs of deployment and financing after years of disappointment. They represented different business models and milestones, rather than a single measure of autonomous-driving readiness.
A small demonstration count sat on top of prior learning
HumanPlus’s task demonstrations were only one part of its training story. Motion data and simulation supported the underlying controller before the highlighted skills were learned. Reporting both layers gives a clearer picture of sample efficiency than presenting the small demonstration count as the system’s entire experience.
Evidence you can use
Robotics in the 2024 report
Historical snapshot: October 2024. Dates and populations are specified per row.
Robotics in the 2024 report
Measure
Reported value
Definition and source
HumanPlus task demonstrations
Up to 40
Demonstrations for the tasks highlighted by the report, in addition to prior motion data and simulation training.2024 report, slide 78 (PDF page 79)
Demonstration counts, paid trips, and funding measure different things. The HumanPlus examples are research tasks, while Waymo’s trip volume is an operational milestone. Neither establishes unrestricted general-purpose autonomy.
Waymo expanded service and reported 100,000 paid US trips per week in August. Wayve raised a $1.05 billion Series C. The report treated these as signs of progress after a difficult period for the sector.
HumanPlus combined a human-controlled shadowing system with imitation learning, building on a low-level controller trained with motion data in simulation.
It brought pretrained models, demonstration datasets, and development tools into a shared ecosystem. The aim was to make robotics experimentation easier to begin and reproduce across teams.
Its shadowing system let a person use camera-observed movements to control a humanoid’s whole body. Those demonstrations then supported imitation learning for autonomous skills.
No. The report also described prior motion data and simulation for the control policy. The demonstration count applied to highlighted task learning, not every stage of the system’s development.
It counted paid trips per week in the US, as announced in August 2024. It was an operational volume measure at that time, not a global autonomous-driving total.
Wayve’s $1.05 billion Series C measured financing, while Waymo’s figure measured paid service use. Both were meaningful milestones but answered different questions about the businesses.
Historical snapshot published October 10, 2024. This web edition was prepared on October 10, 2026 from the online deck and original launch posts. Findings and forecasts retain their original time frame.
2024 report, slide 75 (PDF page 76)Original 2024 report. Printed slide numbers are one lower than PDF page numbers because the cover is unnumbered.
2024 report, slide 78 (PDF page 79)Original 2024 report. Printed slide numbers are one lower than PDF page numbers because the cover is unnumbered.
2024 report, slide 123 (PDF page 124)Original 2024 report. Printed slide numbers are one lower than PDF page numbers because the cover is unnumbered.
Benaich, Nathan. “Robotics gains shared tools and better ways to learn.” State of AI Report 2024. Historical report snapshot; web edition prepared 2026-10-10.