I remember reading about OpenWorm a few years ago and thinking it was ridiculously cool. I'm glad to see the project has persisted. They do a bad job, however, of selling themselves.
1. What have they accomplished to date? I see little more than an animated worm. Tell me about how you're trying to model the worm, and how these approaches may accurately capture its behavior.
2. Why should I care about having an animated worm in my browser? Why is this an appropriate medium to deliver the simulation? If I want to do any kind of science, how will this help me? What I've seen to date looks scarcely more useful than Bonzi Buddy. The most interesting part seems the Academy, but I must donate at least $250 to gain access. This seems counter to the "open" part of "OpenWorm."
3. What academic affiliations does the project have? If the project is useful and has experienced success to date, surely they can recruit students/postdocs/whatever to work on it full-time, with well-established labs making major contributions. Are they computational people? Biology people?
4. What are the bona fides of the people involved? If they can't typeset or capitalize the species' name properly ("C. Elegans") in their video, that doesn't lend much faith to their expertise. The gentleman in the video marvels over the mere "1000 cells" in the worm, but does nothing to put this number in context (with, say, the 10 trillion cells of humans).
I'd love to see this project succeed, and I admire its attempt to recruit funding through a novel means, but the pricing seems too steep, and the overall quality of the pitch is regrettably poor.
This project is huge. I'm glad to see it has come this far. It's the first ever simulation of a multi-cellular organism at a really useful detail, presented and made available to the masses.
It's work like this that is going to help explode the use of citizen scientist work. Imagine being able to run your own experiments on a simulation first without having to buy and breed your own worms. So many more experiments can be carried out, and in parallel too.
It's not an exact model yet, but it's getting closer. The end goal is to get the model to the point where if you run an experiment on the virtual worm, you can be certain you'll get the same results on the real worm.
I completely agree, while at the same time I wonder if they're not being too ambitious--are there simulations like this but for simpler organisms, preferably that is open source as well? I would love to be able to just inspect and toy around with a cell and its organelles, protein production, etc, in a visual as well as programmatic way. I think that's what's missing (as in I haven't come across it) in getting me into "diy biology" as a hobbyist.
The problem is you need to know about tons of stuff before doing anything useful with it. For example, you need to know the biology of M. genitalium (or even what the hell it is) and you need to know at least a little bit about how the biology works (i.e. what is the metabolome? etc.). Then you need to know enough about software to be able to install and get everything running (often with poor documentation). Sometimes you need access to expensive software like Matlab.
So I guess that means it's coming, but very slowly and only for people who are in the field, not DIY biologists.
This is actually something I'm interested in working in (once I get to grad school).
I'm sorry, but this sounds like a fantasy land. I am extremely skeptical that you could get a detailed and accurate enough simulation to reveal any useful novel results.
The list of "unknown unknowns" that we don't even know about, yet alone are able to attempt to simulate, is impossibly large.
This reminds me of the old Star Trek episodes where they would simulate the effect of the new-fangled quantum-slipstream-tachion-whatever-warp-drives in their Holodeck and then obtain meaningful results.
Of course, that was nonsensical fiction. I don't see how this could be different.
Simulations and models need to be incredibly focused. The art of modeling is figuring out what to exclude from the model, not what to include. Any attempt to make a really general purpose model/simulation is almost bound to overreach and fail.
This is the mindset that has prevented people from even attempting such projects, it is poisonous and anti-intellectual.
One thing that we almost NEVER do is with models is force them to interact with each other. This means that we are completely blind to an entire aspect of isolated models: namely that inputs may not have the structure we expect, or that their outputs don't actually have the anticipated affect on downstream models. Most current model validation is purely speculative or based only on a finite dataset. If you take two models that you think effectively represent how the world works and put them together and suddenly they no longer work, then you reveal a huge gap in our knowledge. In a sense this is the ultimate form of model validation and until you do it any interpretation you make about how models might or might not work together is complete bullshit.
I spend a lot of time working on models. This is not a "poisonous and anti-intellectual" statement, it is simple a statement grounded in experience.
There are dozens of failed projects to make different models work together. It would be such a beautiful concept if we could "snap" different models people built together like legos and watch how they interact and develop together. It is also a concept very inline with the hacker mindset where we all observe the phenomenal success of the UNIX building block approach to getting things done.
It is a great idea in theory, but in practice making models work together is incredibly difficulty. You have massive issues of scale (specifically in regards to temporal scale with process going on at very different rates) and context. This has been tried again and again. There are many software packages designed to make this easy. As a general rule, they simply do not work outside of narrowly focused domains.
What is "poisonous" to science (and science funding) is to overpromise and underdeliver. What is also "poisonous" is to ignore a long history of modeling and simulation work and the hard-earned lessons and failures gained from that work.
The only reason I say that mindset is problematic is because the way you frame your statement is "this can't be done" when what is actually the case is that we have never really tried because there are many other easier problems to solve. I don't think anyone who has spent even a limited amount of time modelling expects to be able to stick a few equations together and have it all work. It is still an open question as to whether we can capture say 95% of the variance in a system using a collection of simplified models or whether we have to go full monte carlo and study models that have the full complexity of the system they represent because you really do need every last single part and can't simplify anything.
There will be novel results no matter what happens, maybe not about c elegans, but certainly about the models themselves.
This model is definitely excluding a lot of details. For instance, the cells are modeled on an abstract functional basis. There isn't any cellular division occuring and it's going to take a lot more work to ever get to that level.
For purely simulating behavior and understanding how the insides of the worm "function together", this model is a fantastic breakthrough.
I perhaps was being hyperbolic with my original description of the worm being a catch-all for any experimentation whatsoever, but you have to admit that it has vast implications.
(You also don't deserve the downvote you got. We always need somebody anchoring us to reality. )
Outside of the germline, C. elegans does not undergo cellular division in the adult stage. There is a fixed developmental pattern, with exactly 959 cells in the adult hermaphrodite. Also, the connectivity map of the entire nervous system is known.
I do believe that this is a permanent showstopper for modeling C. Elegans reproduction, correct?
I would be very curious to hear if there have been any plans to one day re-visit the model to allow reproduction, and if any thought has gone into how that process could be allowed!
> Imagine being able to run your own experiments on a simulation first without having to buy and breed your own worms. So many more experiments can be carried out, and in parallel too.
I'm trying to better understand the significance of the project. Can you explain this more? Wouldn't running experiments require similar level detail simulators for whatever you are trying to test?
Usually when you do a genetic experiment, you have a model of how things work in your head, then you do the experiment and test that model on living organisms.
It's entirely possible that your model is way off and you get results that don't mean anything. With computational models, we can test hypotheses (the models in our heads) much faster (don't have to wait for the worms to grow up) and much cheaper (we can do it on the computers all of us already own).
I will definitely be donating. I met Stephen about a year and a half ago when this was just getting off the ground and he was (and still is) incredibly enthusiastic about going after such a challenging problem. When I mentioned that we had had relatively little success building such complex models he pointed me to the work modelling Mycoplasma genitalium [1]. The problem is not that we don't have the mathematical or computational tools or even the data to do it, it is that the social and practical aspects of organizing and integrating such a major engineering project are usually only available at companies with massive amounts of capital. Serious attempts to completely model complex systems are also usually beyond the scope of the least publishable unit. Hopefully open science will be able to bridge the gap. Good luck to the whole OW team!
I met John who worked on their (paid) iPhone app for the earlier prototype. There's one thing you can say it's that they are passionate about their project. The old prototype is here: (paid, is it OK to link this on HN?) https://itunes.apple.com/us/app/openworm-browser/id595581306...
The aim is you grow worms and see what happens. Here is the 'must read' review:
Is it possible to become emotionally attached to a kill-crazy cannibalistic worm that looks like a facehugger from the Alien movies and spends most of its time attempting to eat its siblings?
Having recently shed a tear while burying my last one - Chompy - in the back garden, I'd say yes. These little beasties inspired fear and disgust in my girlfriend, but to me they were true friends.
How could I forget the way Chompy used to play with his smaller brothers, chasing them around their tank for hours on end? Every couple of days, one of the brothers would vanish completely, and Chompy would do an extra-long poo to show how much he missed them.
After about 30 days, Chompy disposed of the final, equally large brother - Ripley - by eating him from the tail up. I caught the two of them playing on the bottom of the tank - Chompy had Ripley's face in his mouth, and was munching away without a care in the world. The rest of Ripley was nowhere to be seen.
Heartbroken, Chompy only lasted another week after that. For a while, he ate his fish pellets and bits of carrot as normal, but a triops is only half a triops without his playmates. Eventually Chompy turned green, and the end was nigh.
Would I repeat the experience? Maybe, but next time I would have to steel myself for the inevitable tragic end. Triops might not live long, but they've got personality. And they eat Sea Monkeys for breakfast.
So, if this open source 'digital organism' can evolve to be as cool as Chompy then there could be quite some appeal.
Worm Tangent: If you have children and enjoy gardening you should look into setting up a worm compost bin in your basement. I do not have kids but a lot of the kids in the neighborhood enjoy looking at the worms in the compost bin and seeing how they move up a new layer once they have eaten all of the food in the bottom layer. The biggest hit with the kids is the moving pink brillo pad that is pile of baby worms. When I got the worm farm setup my only goal was cheaper food for my roses. The worm casing/compost is great for my roses and the neighborhood diplomacy is a great side benefit.
It would be much more exciting if they could get a SINGLE cell (of whichever organism) simulated with virtual subatomic particles. Has anything like that been attempted? Do we even have the data?
Agreed, this would be much more exciting, but we decidedly do NOT have the data and do NOT have the processing power to do this in general. Nor will we for the forseeable future, even if Moore's law holds and we figure out quantum computers.
Still, thanks for asking: it's an important question that deserves a thorough answer. There are four separate "simulation domains" lying between simulations of subatomic particles and simulations of whole cells. In each domain, one might expect to answer a single targeted question through years of painstaking accumulation and application of expert knowledge.
(0) Let's start by ignoring the fifth domain (that I didn't count): simulating the behavior of the atomic nucleus. Nature has granted us a reprieve here in that biological activity is independent of what goes on at this scale.
(1) The first domain happens when you try to simulate an atom (not a molecule, a single dinky little atom) starting from a point-nucleus and an electron cloud. Time-stepping a Newtonian Mechanics simulation is typically between O(N) and O(N^2) in space and time. Unfortunately, newtonian mechanics and classical E&M can't tell you anything remotely connected to reality about an atom. The wheels come completely off: they predict that electrons radiate away their potential energy and crash into the nucleus, saying nothing about the "orbitals" we observe them stacking into. Quantum Mechanics is necessary, and the full equations are O(M^N) in both time and space, where M is the resolution along each axis of your simulation (every possible "universe" (set of particle locations) has a complex probability associated with it). There are decent approximations, but none of them suffice to answer all the relevant questions at once. Expert knowledge and experimental verification must be used to select the appropriate approximations.
(2) The second domain happens when you try to simulate a molecule from atoms. Nature grants us a huge reprieve in that we can safely ignore most atomic behavior: only the valence orbitals have interesting interactions, and nuclei are much heavier than electrons, so we can model the electrons' behavior independently. Still, the full problem is O(M^N) in time and space, and no single approximation works in all cases. A grad student with several years of training in quantum mechanics might spend months finding the appropriate set of assumptions to simplify and simulate a single chemical reaction.
(3) The third domain happens when you try to go from the scale of small molecules (dozens of atoms at most) to the scale of biological macromolecules (proteins, with tens of thousands to millions of atoms -- completely impractical even for heavily simplified QM models). Nature grants us a huge reprieve in that Netwonian mechanics becomes relevant at this scale. We can ignore most of the quantum mechanics most of the time and model it using struts, springs, and repulsive forces (at the price of ignoring chemical reactions, which we must separately account for if necessary). The difficulty here is the timescale: individual "wiggles" and collisions happen every femtosecond or so, while meaningful reactions often happen on the scale of milliseconds. If you let the width of a large pencil lead (1mm) stand for the time of a typical "wiggle," a single second of simulated time stretches from the Earth to the Sun. Cutting edge custom-ASIC supercomputers can simulate single small proteins for 1.5 milliseconds: http://en.wikipedia.org/wiki/Anton_(computer) . Last year's Noble Prize in chemistry went to people who spent their lives whittling away at the problem of integrating (2) with (3).
(4) The fourth domain happens when you try to go from the scale of a single molecule or macromolecule to the scale of a cell. Nature grants us a huge reprieve in that typical macromolecules usually only have a small number of functions and often only act in a statistical sense (they can be well characterized by their concentration and the concentration of their substrates in different compartments). Unfortunately, this data is very difficult to collect with certainty. How do you know that a given protein interacts with a given substrate in a given way? There are many ways to guess and many ways to measure, but for the most part we have to tackle proteins one-at-a-time. This is what biologists do, and after hundreds of years, billions of dollars per year, and countless underpaid PhDs plunking away at the task of characterizing individual pathways, there is enough of a picture to perform a very rough simulation of the whole thing by specifying a chemical kinetic ODE between various species in various compartments.
Note the pattern at each scale: A full simulation within the domain is impossible, but nature grants us a reprieve which lets experts answer specific questions with tremendous expenditure of effort and a quantity of data that grows dramatically in the number of the layer.
In a theoretical sense, we do have all the data we need: the only thing stopping us from going from 1-4 is computational power, however the gulf is so large it will almost certainly never be spanned by a single simulation. In a practical sense, where we want to ask high-level questions on the scale of (4) or higher, there is a great deal of data missing, since data plays the role of simplifying the lower layers of the simulation, or of allowing us to skip (4) altogether and ask a question on the scale of (3) or (2).
I sometimes like to think about the ethical consequences of having that type of data and computation power.
For example it's very likely for it to be used to perform experiments in biology.
But who's to say that an organism simulated in that low level is any different from the real thing? If there even such a thing as "the real thing".
And it gets even weirder when simulating human beings. Is a simulated person any different from us? Is he really conscious or does he merely "behave" conscious? Is it ethical to use it for experiments? What about entertainment? And also, it raises the possibility that we ourselves might be simulated in one level or another.
I think that in some point humanity will have to face these questions. Though from what I understand from you, we still have a few centuries to get there... Man, that's something I'd love to see.
The Emperor's New Mind: Concerning Computers, Minds and The Laws of Physics is a 1989 book by mathematical physicist Sir Roger Penrose.
Penrose presents the argument that human consciousness is non-algorithmic, and thus is not capable of being modeled by a conventional Turing machine-type of digital computer. Penrose hypothesizes that quantum mechanics plays an essential role in the understanding of human consciousness. The collapse of the quantum wavefunction is seen as playing an important role in brain function.
No. The simulation hasn't modeled the reproductive system as of yet. I'm not sure it ever will. A lot of the power of the reproductive system is in the DNA, and this simulation does not model DNA.
> ...I'm not sure it ever will. A lot of the power of the reproductive system is in the DNA, and this simulation does not model DNA.
I'm not sure what you mean by the last sentence. Not incorporating a full simulation of genetic expression does not preclude the authors from incorporating a simulation of C. Elegans reproduction or the reproductive system in some capacity.
DNA is just a container of letters that ultimately form the words that are amino acids. These amino acids are formed into sentences and phrases that are proteins. Should these proteins be arranged in a functional grammar, their structure and actions collectively express the language of life.
DNA is just the machine code for the emergent system that is life.
A biological system, such as the reproductive system, is comprised of some components that are several layers of abstraction above the DNA; as each system is an emergent property of the symphony of molecular machinery and interaction that constitutes a living organism.
Since many authors were already able to simulate abstractions over other structures and systems, there is no reason why the C. Elegans simulation could not extend their model further. As far as I know, they were not simulating gene expression at all.
Saying that a particular biological abstraction cannot be simulated because it does not contain a lower level simulation of genetic expression is a bit like saying you cannot simulate a ball bouncing because you do not include an atomic resolution molecular force simulation of rubber molecules. Or, that you cannot program a football simulation, like FIFA, because you do not include lower level simulations of aerobic and anaerobic respiration.
Obviously, molecular force dynamics and respiration are integral phenomena that enable bouncing balls and football matches, respectively. (I'm unaware of existing respiring balls.) But those phenomena can be closed over by abstraction, in order to create sufficiently educational simulations for particular scopes of understanding.
Of course, it would be computationally difficult to simulate each level of abstraction, all at once, at atomic "resolution" (or molecular or macromolecular etc. "resolution"). It would be impractical to simulate molecular dynamics, or transcription and translation, or protein folding; when all you care about is the general concept of reproduction and the more abstract structures involved.
You could certainly produce a sufficiently useful naive simulation that is faithful to the spirit of an organism's reproductive system.
There is always a hard boundary on the "resolution" of a biological system simulation, limited by computational power, but there is no logical limitation preventing some simulation of the reproductive system at some educationally valuable level.
As much as I understood, this project isn't so much about AI, as it is about a better understanding of biology.
I don't know if the project still lives, or are there other many such works, but Polyworld seemed like an interesting idea (developing AI through evolution): https://www.youtube.com/watch?v=_m97_kL4ox0
I may have misunderstood you, but anyways: Developing AI through evolution is pretty common in AI nowadays. Using genetic algorithms (GA) to evolve parameters for other AI solutions, for instance the weights in an artificial neural network (ANN).
Just finished a project where we did exactly that. A simple ANN with feedback (memory) where the different weights between nodes, gain, bias etc. were trained through GA. After a few generations intelligent behavior started to emerge.
They call it "open", then turn around and charge ~$50 to access the web-browser version of the sim for a year? Doesn't feel very open to me.
I recognize just how important C. elegans is to neural/bio research, and how ambitious the project probably is. I just think this would be so much cooler if they were offering it to all curious minds free of charge.
I actually felt that the video was frank as can be, without the usual kickstarteresque shameless jazzy sales and marketing schmutz.
Openworm is genuine out to improve biology. Not the money, they just need it to continue working on the project. Many kick starter campaigns, can easily be deduced as shameless money grubbing through fake passion pleas. Not this one. And yes, you may call it poor marketing. But hell, that's at being genuine appears to be. We're not used to genuine kick starters.
1. What have they accomplished to date? I see little more than an animated worm. Tell me about how you're trying to model the worm, and how these approaches may accurately capture its behavior.
2. Why should I care about having an animated worm in my browser? Why is this an appropriate medium to deliver the simulation? If I want to do any kind of science, how will this help me? What I've seen to date looks scarcely more useful than Bonzi Buddy. The most interesting part seems the Academy, but I must donate at least $250 to gain access. This seems counter to the "open" part of "OpenWorm."
3. What academic affiliations does the project have? If the project is useful and has experienced success to date, surely they can recruit students/postdocs/whatever to work on it full-time, with well-established labs making major contributions. Are they computational people? Biology people?
4. What are the bona fides of the people involved? If they can't typeset or capitalize the species' name properly ("C. Elegans") in their video, that doesn't lend much faith to their expertise. The gentleman in the video marvels over the mere "1000 cells" in the worm, but does nothing to put this number in context (with, say, the 10 trillion cells of humans).
I'd love to see this project succeed, and I admire its attempt to recruit funding through a novel means, but the pricing seems too steep, and the overall quality of the pitch is regrettably poor.