It seems the Figure 01 won't just be making coffee when it shows up to work at BMW. New video shows the humanoid getting its shiny metal butt to work, doing exactly the sort of "pick this up and put it over there" tasks it'll be doing in factories.
Figure teaches its robots new tasks through teleoperation and simulated learning. If its videos are to believed – which is not always a given in this rapidly evolving space – its humanoids are capable of 'figuring' out the success and failure states of a given task, and working out how best to get it done autonomously, complete with the ability to make real-time corrections if things appear to be going off-track.
Its early-generation body is fairly slow, weak and wobbly on its feat; it's no Atlas, in athletic terms. But in hardware terms, the 01 has enough strength and sensitivity in its arms, legs and dextrous humanoid hands to get out into the workforce and start earning its keep. It's not the brawn that's holding these things back, it's the brains.
In that context, take a look at what Figure tells us is its latest completely autonomous task.
OK, so it picks up a box, then walks over and puts it on a slide. It's certainly not sexy – but then neither was I, 25 years ago, when I showed up pimply and unkempt for my first day of warehouse work, for a job that required me to pick things up, then walk over and stick them on a pallet.
And while I personally may never have managed to get any sexier in the interim, I can tell you one thing business owners do find very sexy: An opportunity to reduce labor costs.
Figure and its competitors are beginning to test these general-purpose machines in real-world work situations, and while they may start out with very rudimentary abilities like these, they may well very soon start paying their way, with the ability to be employed round the clock, when they're not charging, for per-hour rates well below what humans cost.
The abilities won't remain rudimentary for long; each new task – indeed, each new tiny sub-task – contributes to the development of a "large behavior model" that will quickly grow to include a broad range of different tasks, motions and physical capabilities. Swarm learning will ensure that when one learns, they all learn, and the rapid rate of AI acceleration will ensure that once a certain baseline is reached, progress will be as horrifyingly fast as it currently is everywhere else in the AI world.
All of which may illuminate why Figure has reportedly drawn in a massive amount of funding, with Fortune reporting some US$675 million worth of investment that's about to be announced. OpenAI and Microsoft were in early – you can draw your own inferences as to what these companies might bring to the table – and now Jeff Bezos, Amazon, nVidia and a host of venture capital firms are in on the action.
These embodied AIs should eventually have the physical and mental capabilities to take over more or less any manual task currently handled by human workers. No holiday time, no injury risks, no arguments, no unions, no Granny's funeral to take the day off for.
The potential for this tech to reshape society is near-limitless; there'll be no labor constraints on the economy, productivity levels could rise through the roof, and maybe we'll reach an age of abundance in which there's enough of everything to satisfy everyone's needs and wants.
Or, as we put to Figure founder Brett Adcock last year, maybe this becomes a world in which a handful of robot/AI overlords control all the capital and the rest of us start looking like gaping mouths to feed with nothing to contribute. In which the value of human labor, human intelligence and even human rights might plummet to zero. In which we're no longer the top of the food chain.
So while the Figure 01 is just picking things up and putting things down right now, and fidgeting with Keurig machines, this is the beginning of something extremely significant that could go about as far as it's possible for any technology to go in terms of changing the world.
Source: Figure
