Via @s8mb on Twitter.
- The key point is that there was no programming or learning involved to create the behaviors. The connectome of the worm was mapped and implemented as a software system and the behaviors emerge.
Also, Homepage for The Open Worm Project: http://www.openworm.org/
By far the most interesting and significant part.The key point is that there was no programming or learning involved to create the behaviors. The connectome of the worm was mapped and implemented as a software system and the behaviors emerge.
Also by far the most amazing achievement in terms of sheer work by Sydney Brenner's group way back in the day. They SERIAL SECTIONED a few worms and imaged with electron microscopy to build a detailed anatomical map of every cell, with special attention to neuronal connections. This is an insane amount of incredibly tedious work. You can see the fruits of their labors at the Worm Atlas.
I had never looked at the protocol for that experiment before. Jesus christ. "how many grad students does it take to scan a nematode?" etc.most of the reconstructions were done by hand from a total of about 8000 prints.
It would be cool if version 2.0 utilised wireless charging, and the "food sensor" was stimulated when the wormbot neared a wireless charging station and obtained charge. Then maybe you could add obstacles and introduce a Darwinian element where only conectomes that could successfully "feed" themselves made it through to the next iteration. You could add random mutations to provide variation and weed out the bad ones by testing them irl with the bot.Stimulating the food sensor made the robot move forward.
http://zenpencils.com/comic/33-edgar-mitchell-a-global-consciousness/ (I have the print version of this over my desk in my bedroom)
The crazy thing, is how easily it will be to scale this. I am a believer that this is intelligence, and that this LEGOtode has a consciousness within a similar order as a nematode. It's always fascinating to see patterns emerge from software like this. There can be little doubt that our brains act similarly in our youth. What is interesting, is that the results of one nematode's LEGOtode's learning process could be mapped upon another. LEGOtodes can skip childhood.
Scale it to what? Like, use this process with more neurally complex organisms? Implementing this was rate limited by the available technology, but mapping that motherfucker? mechanosm on my comment pointed me to the protocol (granted that was in 1986) but, while automation is our friend nowadays, it's not any less tedious. My experience is limited to mapping neurons using confocal in 50 micron sections, but that was not terribly efficient. Just physically linking the neurons is bitch in and of itself, but then you have to confirm each neuron type, response, etc. It's not as simple as "draw the roads on the map, and now you drive across the country." bad metaphor but whatever.
It really blows my mind. Basically they made Google Earth for a worm and the mapped structure took it from there. Here he is in your browser too: http://browser.openworm.org/
But if they skip "childhood," doesn't that eliminate the possibility for any potential variances that might occur? Any anomalies that might veer them in a new direction developmentally?
That is both amazing and scary at the same time. What happens when(if) we manage to simulate a larger mind, and a larger one, up to human level. What happens if we get a computer that can simulate hundreds of minds and construct them in a way that they work in a group to solve problems? We can capture hundreds of years of evolutionary problem solving without the need to program all of it from scratch. More importantly, what does that mean ethically? Do these simulated minds count as humans? What if we do not simulate their ability to feel pain or suffering? Does a computer count as the person, or does the simulated space count as the person? Does turning off the computer count as murder? I get the feeling that should these projects get to higher scales, these are going to be the next generations, or the next-next-next generation's abortion issue.
My assumption is that currently, you are the arrangement of particles and matter in your mind. If you create a second copy of you, that is not you, it is a copy. It isn't made of the same matter, and both machines can exist at the same time. The only way to really "transition" is to do the same thing that we do in real life as our minds develop. Slowly replace and reduce the cells in the mind with artificial ones, and have a slow transition that lasts over years. That way there can never be a "copy" of you, and "you" are defined as the half-half being during the transition period.
These are good problems to have and good questions to ask. It means we are advancing...to the future!
Damn...they ported a nematode to Raspberry Pi. http://themindi.blogspot.com/2007/02/chapter-26-conversation-with-einsteins.html I'm always reminded of this Tortoise & Achilles bit when I hear about machine learning.
Anyone know how this compares to IBM's work? This is far outside of my area of expertise, but it seems like IBM has already gone far beyond this in terms of raw simulation power.
this experiment isn't about computational power, this is a emulation of a natural connectome. They're different in application. IBM is running a bunch of artificial neurons at the size of a cat or mouse brain, but they do not have the connective features particular to an entire connectome. They took existing data about how exactly each of the neurons physically and electrochemically interact, imported that into an artificial neural network, and hit "run". Out came a synthetic version of a nematode.