To do something, you need energy, and you need information. Energy to do the thing, and information to define what's being done. The mental image I have of a "robot", has inputs of both continuously. The input of energy comes as electricity, and the input of information comes in the form of digital instructions. They're both abstracted away from the sort of work being done.
A missile (some kinds at least) takes in digital information, and converts that to a location. It then flies there. But when it does it gets its energy from a specific source: fuel. It can't do anything with the fuel except fly forwards.
A motor takes in abstracted energy but not information. The energy input is electric, but the motor can be used for lots of different things. The information it takes in, however, is entirely related to energy flow.
A wrench does neither. The energy input isn't abstracted, and neither is the information input.
| Direct/Specific Information | Abstracted/General Information | |
| Direct/Specific Energy Input | Category 1: Wrench | Category 2: Missile |
| Abstracted/General Energy Input | Category 3: Motor | Category 4: Robot |
To get to generality, we need category 4. We can also define a category 0: which are non-energy-transmitting devices. These are structural objects like bridges.
The ultimate goal of nanotechnology is atom-precise, scalable production. "programmable matter". This would be the second most influential bit of technology developed this century. I've often wondered what's up with nanotech, and what needs to be done to advance the field.
To do this, it needs to be general. To be general we need to get to category 4 in the bottom right part of that diagram.
There are lots of possible ways of making very small things do things.
Here I will include various systems which are typically not atom-precise, such as nanometer-sized lipid droplets (like those used in RNA vaccines), and nanoscale materials like carbon nanotubes. These almost always fall into category 0. Some of these nanomaterials are interesting in that they can respond to stimuli (like changing volume upon pH change, or being part of a battery which charges and discharges) but this only moves them to category 1.
It might be worth mentioning the existence of systems like the ones which allow manipulation of single atoms. For various reasons these tend to rule out the creation of rigid structures which limits their usage to making nice electron micrograph pictures. The resulting nanosystems are solidly in category 0. They also scale extremely poorly, it's very slow to make lots of them.
Chemical synthesis is the most common method of construction of atom-precise structures. It's so ordinary it's not even called "nanotech" most of the time. Typical chemical synthesis often falls into category 1: it's bespoke. At each step of the synthesis we add in specific reagents which act as both an energy source, and determine what reaction occurs. We also need to carefully manipulate conditions.
Working out how to make a molecule is difficult, thousands of careers have been just this. Imagine not only designing a robot, but having to spend a year of your PhD just figuring out how to put it together.
Chemical synthesis is moving towards category 4 over time, as more and more standard reagents and reactions are created.
Chemical motors exist and are quite cool. Basically they rely on a series of reactions which can transduce energy from some source to rotation. One method is to catalyse the reaction of two high-energy small molecules. Another is to rely on input of alternating light and heat. Like regular motors they are category 3.
Biological chemical motors also exist. The most common kind interconvert between chemical energy (synthesis of ATP) and electrochemical energy (which is a combination of a concentration difference in H+ ions, and a voltage, sort of like electrical energy but with protons flowing instead of electrons). But there are some which use the energy released by ATP breakdown to "walk", and some which convert an electrochemical gradient into rotation, which powers bacterial flagellae (which are like tiny propellers).
DNA nanotech generally refers to building structures out of DNA directly, and doesn't have much to do with how DNA carries information in living systems.
DNA nanotech is based around building a final structure based on base-pairing. A base-paired DNA structure is very low in energy, so forms very reliably. Then DNA is synthesized which isn't base paired yet. Often it's designed to form a less-effectively-base-paired structure initially, which then changes to the final structure upon some sort of trigger. Heat is a common trigger.
This makes it category 0 or 1. The energy input can only be used for one purpose: being in a low-energy shape. The specifics of the base-pairing pattern must be redesigned for each structure.
Protein nanotech encompasses a lot of things. One example is enzymes. Some just act in the same way as chemical synthesis, category 1. But many are powered by ATP, and carry out a specific chemical reaction. The specific reaction is defined by the shape the protein folds into, which is intimately connected to the sequence of amino acids which makes it up in an extremely complex way. This means protein structure is not well abstracted from its function. This puts it in category 3.
A good example is the CRISPR-cas9 system. The protein is given a short stretch of RNA, which is designed to base-pair with itself forming a little loop. The RNA acts as a guide against which DNA is compared. When a match is found, the protein cuts the DNA. Cutting DNA does not require an energy input. This is one of the few nanotech areas which is well on its way to being category 2, since the protein is genuinely taking a significant information input.
Using a wrench might not seem like it's computationally expensive, but it is. You need to be observing the tension in whatever you're turning, and adjust the force you're applying to
If you want to use a motor to actually go somewhere, you need to do computational work. You need to start by increasing your energy input, then observing where you are, then decreasing it as you approach your destination.
This computational work may not be trivial. To successfully convert protein nanotech from category 3 to category 4 you need to solve both protein folding, and understand how protein structure relates to function.
One interesting question is what happens when we string together different systems from different categories. For example, imagine taking the guidance system from a missile, and attaching it to a bunch of motors. Now we have a system which takes a digital input of a location, and an electricity input, and goes there.
If we repeat this process with a different category 2 system (like an imaginary programmable but gas-burning blowtorch) and category 3 system (like an imaginary non-programmable laser-torch) we get a programmable laser-torch. Now, we can do something really magic, we can put the two systems together without having to worry about energy input. Now we have a little robot which can move around and also cut things. This is a way of getting towards category 4.
Living cells have a brilliant category 4 system which translates information from DNA into proteins. This information abstraction is what makes life so successful. The energy is derived from ATP, a general energy source for cellular processes.
But once produced, most proteins act as a category 3 system. This means if we want to control protein activity, we have to understand the relationships between protein structure and function. While our system superficially looks general in information and energy, it still has all the trappings of a category 3 system in terms of computational overhead.
We've learnt that the distinction between categories 1 and 3; and 2 and 4, is important in terms of computational work required to work backwards from desired output to required input. Even when an input looks to be abstracted physically, if it's not abstracted in terms of computation, it does us no good (DNA -> protein -> reaction requires the same amount of computation as (protein -> reaction).
The distinction between categories 1 and 2; and 3 and 4 is important in terms of combining systems together.
Lots of progress comes from generality and abstraction.
The invention of programming languages abstracted the fetch-execute cycle (and these days also things like memory allocation) from the majority of people who want to do things with computers.
RNA vaccines are brilliant because they abstract most of the delivery and safety concerns of vaccines away from the specific pathogen.
Drug delivery is poorly abstracted from drug function. Lots of molecules which are excellent at killing pathogens just can't get into the body. Delivery to the brain is even harder.
Chemical synthesis is poorly abstracted from the end product as we've seen above.
Genetic engineering is on its way to being better abstracted from the organisms being engineered, but it's still not great.
This gives a clear potential direction for work in nanotech and nanotech-adjacent fields: make systems which do the work of abstraction.
The scheme above does not extend to sensors. The archetypical example of a nanosensor is the nanopore system developed in Oxford. This takes electricity as an input, and uses it to push DNA through a pore. The DNA blocks the flow of current (in the form of moving ions) through the pore, which is measurable. Importantly, the current is blocked differently by different bases in the DNA. This allows the DNA sequence of a single molecule to be read. This is probably the closest thing to "sci-fi" nanotech that we have.