DeepMind’s Gemini AI to Power Boston Dynamics’ New Atlas Humanoids

Boston Dynamics and Google DeepMind formed a partnership to bring Gemini Robotics AI to the new Atlas humanoid robots. The effort focuses on creating robust visual‑language‑action models for broad industrial tasks, beginning with automotive manufacturing. Joint research will start soon at both companies, emphasizing safety, scalability, and real-world deployment.

Key Points

• Boston Dynamics and Google DeepMind will integrate Gemini Robotics models with the new Atlas humanoids.
• The partnership targets industrial use cases, starting with automotive manufacturing, to transform factory work.
• Joint research begins this year at both organizations, focusing on visual‑language‑action models for safe, scalable deployment.
• Boston Dynamics’ hardware and ‘athletic intelligence’ will be paired with DeepMind’s foundational, multimodal AI.
• DeepMind’s Gemini Robotics aims to help robots perceive, reason, use tools, and interact with humans in the real world.

Sentiment

The Hacker News community is predominantly skeptical of the partnership and humanoid robots in general. While thoughtful pro-humanoid arguments exist, the majority of commenters question whether the humanoid form factor is practical, whether Google can execute, and whether the technology is anywhere near commercial readiness. The tone is analytical and sometimes dismissive rather than hostile.

In Agreement

• The partnership combines BD's best-in-class hardware with DeepMind's AI capabilities, playing to each company's strengths rather than Google trying to build robots alone
• Humanoid form factors are uniquely valuable because the world is built for humans — they can navigate stairs, use human tools, and operate in spaces designed for people
• BD made a prescient strategic decision to perfect hardware while AI matured, giving them a head start now that foundation models can be integrated
• General-purpose humanoid robots are more economically viable than specialized ones for varied, low-volume manufacturing tasks that can't justify dedicated automation
• China's rapid robotics progress demonstrates that hardware challenges are more surmountable than skeptics believe

Opposed

• Humanoid robots are impractical for industrial use — specialized robots already exceed human capabilities in precision, strength, and reliability, making humanoid form a dead end
• Hardware robotics has been a bottomless money pit for decades with no proven commercial viability; every attempt at lifelike autonomous hardware has failed to deliver
• The sim-to-real gap remains a fundamental unsolved challenge, and real-world deployment involves far higher stakes and regulatory barriers than lab demonstrations suggest
• Google has a poor track record shipping successful products, and DeepMind's internal robotics team reportedly struggles with execution and retention
• Household robots will only serve the wealthy, widening inequality rather than benefiting ordinary people
• The humanoid form factor adds nothing — wheeled platforms and paired robot arms can accomplish the same tasks without the complexity of bipedal locomotion