372 | Erik Winfree on DNA Computers and Robots
What this covers
Sean Carroll talks with Erik Winfree, a Caltech computer scientist and bioengineer, about building computers and robots from DNA. They start from a broad view of computation as information manipulated in physical matter, then move to the bench. A synthesized order arrives as powder holding roughly 10^16 copies of one designed strand, and single-stranded DNA finds its partners by stochastic search. Winfree explains DNA origami: a roughly 7-kilobase scaffold folded by 200 to 300 staple strands into an 80-nanometer smiley face. It is too small for optical imaging, so atomic force microscopy confirms it.
The middle of the conversation covers the field's history and projects. Len Adleman's hope that chemical parallelism would beat silicon on NP-complete problems faded after about five years, and attention turned to programming molecular behavior. Winfree describes work with Lulu Qian, including molecules that compute the discrete square root of a four-bit number and a DNA robot that walks, picks up cargo, and drops it at a set location. Tile self-assembly, following Hao Wang's tilings, can simulate Turing machines and cellular automata. Winfree also covers error control, including cooperative proofreading, and growth near melting temperature at almost no free-energy cost, which connects to Charles Bennett on the thermodynamics of computation.
The talk turns more contested on life. Winfree treats information as central to biology and asks whether nucleic acids alone, without proteins, could support living functions if RNA came first. Carroll argues that entropy increase drives biology and complexity. They also discuss Blaise Aguera y Arcas's result that self-replicators emerge in a simple programming language, and a possible DNA version of it.
Winfree argues that DNA is a useful minimal substrate for information-based chemistry, where programmable molecules realize computation directly in matter. The payoff is not competing with silicon on speed but building systems from few uniform parts, which tames complexity and clarifies computation, thermodynamics, and the origin of life.
- DNA both stores and acts on information, and its four-letter alphabet is a reduced instruction set that makes molecular design systematic
- DNA computing abandoned competing with silicon on speed after about five years, shifting to specifying and programming molecular behaviors
- Programmable self-assembly, tiling, and heat-powered systems let molecular computation probe thermodynamics, error correction, and self-replication
Computation is physical information processing, not confined to silicon, and biology already realizes it.
- Computation is the manipulation of information as instantiated in physical matter, so silicon electronics is an especially effective but not the only way to realize it; biology already performs information processing in DNA and RNA long before humans formalized computation.
“The idea of manipulating information in a certain way as instantiated in matter, right?”
- Carroll suggests that the gap between a computer simulation of self-replicating programs and embodiment in real physical matter may be smaller than one might worry, and that this sheds light on long-stalled origin-of-life questions.
“what your work is suggesting is that the gap between his computer program, literal computer simulation of that, and embodying it in real physical matter might be smaller than you might worry.”
- The tiling result is analogous to discovering that computation can occur in simple two-dimensional cellular automata, except that here the computation is carried out in physical matter.
“But now it's physical matter rather than other things, which is very cool.”
Entropy and simple parts drive complexity, making energy-driven DNA research a worthwhile direction.
- Academic work that appears to be fun or impractical, such as making smiley faces at the nanoscale, often teaches unanticipated and valuable lessons because showing how to do something specific in the real world reveals questions one did not know to ask.
“so often just showing and figuring out how you can do something specific in the real world teaches you a hugely valuable lesson you didn't even know you were wondering about.”
- Carroll holds that the increase of entropy, equivalently the using up of free energy, is the driving force behind biology, thinking, metabolism, and the development of complexity in the universe, making energy-driven DNA systems an obvious direction for research.
“I think that increasing entropy, or if you like, using up free energy, is the driving force behind all biology, all thinking, all metabolism, et cetera, the development of complexity in the universe.”
- A car is complicated because each part serves one purpose, whereas a complex emergent system has multipurpose parts organized hierarchically that can do many things, and simple-part cellular automata like the Game of Life show that trivial parts can yield enormous complexity.
“a car is complicated because it's made of all these parts, but the parts are all doing something very purpose-built for just to do that thing. Whereas a complex emergent system has all these parts that are kind of multipurpose”