The MPL is used with non-invasive electrodes, or electromyographic sensors. There is a version where they re-innervated a guy’s chest in order to place the sensors, that had a larger freedom of movements and lower lag, but the processing still lagged a bit, and the subject and/or the system had trouble interpreting too many movements at once.

However, even more years ago, there were experiments on monkeys with a 3rd robot arm hooked up directly to a brain implant. Some seemed to get it quite quickly, ending up being able to solve problems like if they always had 3 arms.

Main problem was, the brain-electrode interface: spike electrodes can only be made so long before they risk getting bent on insertion by a neurosurgeon, and over time (months) scar tissue forms around them, limiting the sensitivity, until they become pretty much useless… at which point a test monkey can be “disposed of”, but humans not so much.

That’s why Neuralink is focusing not on the limbs themselves, which are basically a done deal at this point, but on the brain-electrode interface. If they manage to implant a large electrode array in a way that it doesn’t stop working for several years, they’ll be able to hook up any kind of limb(s), prosthetics, or excavator arms, with direct brain control.

BTW, the original DARPA research on all of this, is advertised as “helping veterans”… but in reality they’re using veterans as test subjects for their actual goals: an electromyographic sensor can detect a pilot’s intention to press the trigger, or move a stick, before the actual muscle contracts, presses the trigger, and that gets detected. That way they can shave off several ms from a pilot’s reaction time. Direct brain implants, could make that even quicker… if it wasn’t for side effect like some Neuralink’s test monkeys dying from “continuous vomiting”… something that nobody wants to happen even a few months after having paid for the training of a fighter pilot.

As an aside, all of this is possible because brain neuroplasticity works all the time. Every time we pick a tool, whether a stick, a car, or a fighter jet, after some training, our brains switch to a slightly different “body map” that includes the tool as an extension of the body.

So right now we have both the limbs (including every human-operated tool), and the ability for a brain to adapt to the “modified body”; what we’re lacking, is an effective long-term way of connecting the two.

As another aside… the brain’s ability to switch body maps, means a single implant should be able to control anything, not just limbs. The question of where is the limit for extra limbs, both in the brain’s limitations, and in how many electrodes need to be implanted, is open for now. Some speculate that a full-brain mesh of electrodes, could allow people to use multiple full bodies at once. If they were implanted at birth, it’s anyone’s guess how a developing brain would adapt to what. There are some development patterns, like stereo vision or hearing ability, that rely on brain structures that are genetically encoded, but there’s no saying where is the limit for expanding them, or everything else. Coopting the vision development to insert multi spectral data, including magnetic fields, might not be that a far fetched idea. There is also no reason to think a “brain in a jar” wouldn’t grow much bigger than when restricted by a skull, or that multiple brains couldn’t be connected and made work in parallel (brain hemispheres already do that, by default humans have “two brains” that are tightly interconnected… octopuses have 9 brains).