Goldman Sachs report: AI robots need 10 million hours data
According to a new Goldman Sachs report, AI-powered robots might need a massive 10 million hours of high-quality interaction data before they're ready for real-world jobs.
That's a huge hurdle, and the report stresses how crucial accurate, diverse data is if we want robots that can handle everyday tasks smoothly.
Robots need better perception and controls
The report points out that current robots struggle with things like understanding 3-D spaces and reacting quickly, two things you really need for reliable performance.
To move forward, the report suggests dexterous Universal Manipulation Interface (UMI) systems, supported by high-resolution tactile sensing.
Interestingly, the report suggests that collaborative and industrial robots (think factory bots) could become useful much sooner than human-like robots.
Getting the hardware and real-time controls right will be key if we want to see these bots outside research labs any time soon.