From: allin
Humanoid robots are making significant strides, with some timelines for deployment accelerating beyond previous expectations. Brett Adcock, CEO and founder of Figure, announced in 2024 that his company is set to begin beta testing robots in homes by the middle to end of the year [00:14:26]. This accelerated timeline is a notable development in technology and AI for 2024 [00:14:28].
Recent Breakthroughs and Timelines
Beyond Figure, there are approximately a dozen companies developing credible humanoid robots [00:14:55]. The big question for widespread adoption is when these robots will become affordable for middle-class households. Estimates suggest it could be as soon as five years from now, or potentially up to ten years [00:15:08]. The anticipated cost for such a robot could be around $20,000 [00:15:00].
Technical Challenges and Progress
The development of functional, general-purpose robots faces two primary challenges:
General Purpose AI and Models
The current state of generalized AI is not yet considered sufficient for broad application [00:15:16]. Brett Adcock notably canceled a public deal with OpenAI to develop his own internal AI model, or to iterate on an open-source model [00:15:22].
Physical Dexterity and Actuators
While robot actuators are “good,” they are not yet “great,” which limits the physical dexterity of these machines [00:15:53]. A demonstration showed robots sorting groceries completely unsupervised, an “incredibly cool demo” highlighting the model’s power [00:16:13]. This demo also revealed a “master-slave” dynamic where one model performs most computation, feeding into a second robot [00:16:05]. Despite the progress, the physical dexterity needed for tasks like sorting groceries without crushing items, making food, doing laundry, or mowing lawns is not yet fully possible [00:16:26]. A key breakthrough involves two robots communicating semantically to collaborate, as seen when one robot moves a fruit bowl for the other to place an apple [00:16:53]. This semantic awareness and understanding between models working independently is a significant step forward [00:17:34].
Diverse Applications of Automation
The potential applications for autonomous robots are vast:
- Household and Ranch Work: Robots could perform tasks like using a weed whacker, trimming bushes, getting wood, and collecting chicken eggs on a ranch [00:17:47]. Their ability to operate 24 hours a day means slow operation wouldn’t be a major issue [00:18:02].
- Specialized Robots: The market is seeing purpose-driven robots, such as Segway-made lawnmowers costing around $1,000, which are being observed in use in Austin [00:19:14]. This follows the trend set by the Roomba [00:19:27].
- Broader Category of Autonomous Systems: This includes not just dexterous automation but also drones and autonomous vehicles [00:18:27]. These systems combine mechanical response with machine vision [00:18:37].
- Dangerous Situations: Remote-controlled bulldozers with AI are being developed for tasks like creating fire paths in smoke-dense areas, where human operators would be at risk [00:20:14].
While the “year of robots” has been anticipated for 30 years, it now genuinely feels like the time has arrived [00:18:14]. The ambition of general-purpose humanoid robots presents a technically challenging roadmap, but progress in specialized automation is more straightforward [00:20:03].