Robotics researcher: foundation models only cut costs, performance ceiling remains

On September 16, robotics researcher Glen Berseth posted a series of tweets pouring cold water on the recent wave of "foundation models for robotics" hype. His core claim: the progress brought by foundation models lies on a single axis—cost—while performance has not genuinely broken through. The implication is that embodied AI must push the performance ceiling rather than only compress the cost curve.

Confirmed

• Berseth noted that as far back as the 1960s, Marvin Minsky's lab had a robot arm using computer vision to recognize object geometry, plan paths and assemble parts.
• He argued today's foundation models largely solve the same class of problems, just more cheaply and quickly—effectively shifting the cost curve.
• He listed the directions that truly need work: more object categories, robustness to in-the-wild environments, multi-robot and human-robot interaction, real-time control, and better hardware—noting that the human hand remains impressive.
• Citing autonomous driving, he pointed out that decades of investment have not cracked it, because such tasks require real-time closed-loop feedback rather than cheaper static perception plus planning.

Why it matters

• This gives the hot embodied-AI narrative a historical anchor and a sober perspective: cost reduction is real, but without breaking the performance bottleneck robots will struggle to leave controlled settings.
• It reminds researchers and investors to distinguish progress on the cost axis from breakthroughs on the performance axis, which carry very different implications for the field's maturity.

2026-09-16 ~ 2026-09-16 · 10 related posts

• Minsky's 1960s robot arm solved the same class of problems as today's foundation models — GlenBerseth · 2026-09-16
• Robotics researcher: AI progress is cost-only, closed-loop control remains unsolved — GlenBerseth · 2026-09-16
• Glen Berseth: foundation models made 1960s robot tasks cheap, but performance is still the bottleneck — GlenBerseth · 2026-09-16
• Berseth: Self-driving shows cost cuts aren't enough — real-time closed-loop control is the hard part — GlenBerseth · 2026-09-16
• [source] Robotics researcher: Embodied AI progress must push performance, not just cost — GlenBerseth · 2026-09-16

5 near-duplicate retellings: GlenBerseth · GlenBerseth · GlenBerseth · GlenBerseth · GlenBerseth