I actually meet DeMarse in Florida. I made it a point to try and speak with him about his "brain in a dish". He was clearly aggravated with me. He told me about all the other amazing things his lab was doing, and how "dumb" his "brain in a dish" actually was. It was obvious to me that he felt the attention his "brain in a dish" had received was unwarranted.
Thank you for sharing this. The attention he received for it was absolutely unwarranted. Anyone who reads the paper and understands what was written would also understand that. Unfortunately, not many people take the time to do that. I've not heard of anyone else attempt it -- it took me several hours of intensive thought to finally figure out what he'd done. At the end of it I felt so disgusted that it was being presented as "rat brain flies plane" that over the years I've often been tempted to do a write-up chastising the author for advancing his scientific career by tricking people rather than discover something new or build something innovative.
This story has been reposted to HN a few times over the years. My reaction to this particular repost was to feel intensely guilty that so many people were still being tricked by the same old snakeoil. Then I looked up what the author had done in the subsequent years, and it's a pleasant surprise: DeMarse has worked on some wonderfully interesting projects which are quite unique.
I'm glad that I never tried to expose this "brain in a dish" as the lie that it is. Everyone deserves to make a mistake once in awhile, and his, I think, was merely to be flattered that reporters were interested in his work at all, which is quite a natural reaction. I'm sure he regrets that he wasn't as careful as he should've been with correcting the reporters' assumptions.
Yeah, DeMarse is a real scientist (or engineer?). I speculate that he did what he did to raise money for his research. I think DARPA money started pouring in to fund his lab after that paper. I just wish more money flowed through the NSF/NIH instead :-)
There's also the possibility that the grad student who'd been working on something a lot cooler (but never got it to work) wanted to graduate, and so pushed out what s/he had.
I'm wrestling with the moral implications of a scientist riding a wave of unfounded hype to raise public money for his own lab. Also, I don't know if that's an accurate description of what happened; it seems that way to me, but I'm worried I'm wrong. I'd be grateful to get your thoughts (and anyone else's).
I don't know how to reply to the old post you put up. So I will reply here. I think DeMarse accomplished something a bit more complicated than what is described in the comments of the old Hacker news post. The protocol used by DeMarse actually modified the strength of the synaptic connections between the neurons in the petri dish. It seems (though not proven) that LTP/LTD is occurring.
Now to answer what you asked me about. Do I think what DeMarse did is morally right or wrong? I do not want to pass judgement because I am still in school and don't have to worry about funding my own research. A professor I respect for his scientific integrity told me every scientist has a skeleton in their closet (referring to research projects). I think the fact that DeMarse's paper caught on like wildfire just goes to show you that a majority of Humans (computer programmers included) can be relatively stupid.
It looks like they're using a combination of high and low frequency pulses as a reward/punishment mechanism, though I don't fully understand how that influences the decisions being made. Would love if someone could explain it in more detail.
On the scale of human brains, the type of flying described in the article is actually pretty easy. I've taught children as young as six years old to do it [in the pods shown at http://www.museumofflight.org/programs/aviation-learning-cen... ], and even younger kids show promise but didn't have the physical ability to use the control setup (the stick was too big, the chairs weren't tall enough, etc.) Honestly, walking is a considerably more elaborate task than flying a jet on a PC simulator.
I'm still impressed that an artificial brain was able to do it.
They've evolved a neural network (in this case, made of real neurons instead of computer simulation) which presumably emits certain signals when it does not detect the horizon in the correct position. It's impressive, but I don't think they were having it take off from an aircraft carrier and land at JFK.
Controlling pitch and roll while at altitude is not actually very impressive, no. If you were writing code to do this, you'd probably just slap a PID controller on each axis, tune the coefficients, and be done. You might break 100 lines of code if you like being verbose.
Every time I see a Wikipedia list such as this I can't help to marvel at the extent at which people contribute.
Yet as a developer I still have the uneerie feeling that all this data ought to be in a database and such a page content should be autogenerated with a single query.
Flying straight and level is More or less the me dificulty of driving a pedal cart in your street (with an added dimension up and down). Old time jets were very tricky to flight due to their high inestability. But modern flybywire combat jets must be easy to fly( even if they are impossible to fly without computers due to their crazy inestability), because the hard part is outside in the combat theater.
Also like in sea combat, the most important part is not being a skill beast (aerobatic champion), is more about knowing how to obtain the best positions, that gives you advantage against your enemies.
Any way I don't see rat brains landing in a n aircraft carrier anytime soon!
Yeah, you'd think it should have ushered-in a few more serious applications but given that, as others mention, the article is from 2004, it seems that indeed "more elaborate applications are a long way off".
Regarding the other comment that was asking about how flying a jet might be insufficiently advanced: basically it's just wired up to Microsoft Flight Simulator, and the neural outputs are hooked up to the essential inputs and controls of the simulator. Hooking up an actual tissue culture with an electrode array tends to be more difficult (or at least more work) than wiring up keyboard bindings to a weight-summer network thing.
This is a fundamental question when doing rat studies, you would like the regions of the brain your are affecting in the rat to be similar to the regions you would find in a human brain.
Odd, this is from 2004, so it's 8 years old, which is like a century in tech years. I'm surprised I have not heard of more advancements in this (creepy!) field.
That because basic science research is a painstaking slow process. It is very rare to see a field of research just explode at an exponential pace like software development has.
precisely. So, then replace 'consciousness' with 'sentience' in my original proposition. I am not saying response to an environment is sentience; but, when using living brain tissue, there may be a chance that these computers will eventually display sentience.
I saw this documentary yesterday (which had a VERY interesting sequence about those rat neurons flying planes, interviewing one of their "fathers" ^^). It is about the future, AI and transhumanism, called TechnoCalyps, which despite the title is actually rather good. It's on youtube, too.. so if you have 3 hours to spare.. it is sure to make you think about a whole range of subjects.
IMHO these things cannot be considered in isolation. We as society should really start talking about our progress and where we want to go, instead of a few people being into it, the rest feeling threatened. We can do better, and we surely have to the tools for it.
You cut open the skull of the rat, pick out the ripest looking neurons, and place them on a block of cheese.
Okay, here is what actually happens. Neurons don't usually differentiate. So instead, cancerous neuronal cells are commonly used (neuroblastoma cells). Basically, these neuronal cells have decided to divide rapidly. You simply place these cells in a serum of growth medium consisting of protein, sugars, and salts and watch them divide over the course of a few days.
It would be nice to use stem cells instead or cancerous cells... but that is another thread entirely.
> Presumably someone, somewhere has a bathtub of human neurons and is probably using them to try and predict the stock exchange.
That's nothing. I've seen -- with my own eyes -- places where they have entire office floors or even buildings full of whole human bodies stationed trying to do the same thing.
