# What are good motivating contexts of use for telerobotics? # What are good motivating contexts of use for telerobotics Abstractly, it’s important for the context of use to prevent overfitting, kind of like regularization in a machine learning model. This means that the motivating context of use needs to require a wide array of environments, force, dexterity, complexity, and time-sensitivity.[^1] This requirement is actually tricky because in almost any given context it’s a better idea to build a specialized system and push towards autonomy asap.  My hypotheses is that most of the value of general-purpose telerobotics will come from the long tail of activities that on their own are not valuable enough to build specialized systems for. (Training an autonomous robot absolutely counts as system building.)  One could imagine trying to prevent overspecialization by going after a bundle of unrelated tasks (pick up this rock, fill this cup, hammer this nail, move that puppy). But [[Tools need a serious context of use]] and more importantly [[Enabling technologies must be developed while doing serious work]]. A set of unrelated tasks is not serious work and it would lead to a weirdly incoherent system. Tensions everywhere! So we need to think about serious, varied, but coherent contexts of use — not easy task. Abstractly, telerobotics is useful in situations that are bad for both people and autonomous robots *or* could enable superpowers that neither of them. Manipulating the environment is a core piece of “general purpose” telerobotics. Situations that The oil and gas industry is already poking at telerobotics because work on oil platforms is dangerous and getting people out there is incredibly expensive. Nothing is in active use (to my knowledge). As far as I can tell, [[Woodside Energy]] is the only oil/gas company who is putting resources towards it; they are both doing internal work ([[Mark Micire]]) and funding at least one lab ([[Nuclear and Applied Robotics Group at UT Austin]]). The publicly available work still feels pretty hacky — see [[TeMoto: Intuitive Multi-Range Telerobotic System with Natural Gestural and Verbal Instruction Interface]] and [[Reducing the Teleoperator’s Cognitive Burden for Complex Contact Tasks Using Affordance Primitives]]. Most of the work seems to revolve around inspecting pipes and turning valves, but I’m sure there are many more jobs that need to be done on an oil rig. It’s probably worth enumerating them. Another advantage to oil and gas is that the cultural overhead might be the lowest to making the transition to actually building in the context of serious work — I wonder how much damage you could do as long as you made sure you had an E stop ready to go. One could imagine that asteroid mining (and extravehicular space activity) is a powerful motivating environment. It’s especially interesting in the sense that baseline human ability is hampered, so there’s less of a gap for telerobots to bridge. Space suits limit range of motion, dexterity, and touch sensations. At the same time, it’s impossible to actually develop this while doing serious work — nobody is doing asteroid mining Asteroid mining does raise the question of terrestrial mining. How much of it is now done by hand vs. with big ass specialized machines? Growing food (specifically gardening) is actually another area that requires a wide range of actions. The trick is that most farming has been adapted to the tools available rather than the other way around, resulting in large homogenous farms in most cases. As a result, most existing farms a specialized autonomous tool (like a strawberry picking robot) makes much more sense. However, [[Gardening maximizes output per acre but requires a fuckton of labor]] because you need to be managing several kinds of plants on top of each other. This raises several possible lines of thought: could gardening be a serious context of use without being a serious *commercial* context of use? Are there any crops that are still grown garden-style? What about vertical farms? Expanding on the last point, vertical farming seems like a promising future direction that (I am guessing) puts a very high premium on output per hectare, which is what gardening maximizes. It seems like vertical farms are likely to go down the optimize for tools and automation monoculture route but if you reduced the labor costs perhaps they could become more like gardens? Nuclear power was (and continues to be?) The original application of telerobotics. “Home maintenance” feels a bit too disjoint to be a motivating context of use. But perhaps one could frame it as “all the jobs that you would pay someone to come into your house to do”: plumber, electrician, house cleaner, handyman, … Doing science could actually be a power ### Related * [[You can’t track people and resources on oil rigs]] [^1]: Is this secretly the ontology of telerobotics tasks? (See [[Serkan Cabi convo 6 Oct 2020]]) [Web URL for this note](http://notes.benjaminreinhardt.com/What+are+good+motivating+contexts+of+use+for+telerobotics) [Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/What+are+good+motivating+contexts+of+use+for+telerobotics)