An example: We didn't start to learn about electricity until we started messing with it. That's because there was very little to observe passively. Touching electric fish, rubbing amber to make static electricity, flying a kite into a thunderstorm--these sorts of actions produced the early observations that started us down the path toward the knowledge we have today.
Your example was observing USB traces. In order to do that, you had to plug a well-understood sensor into a well-understood interface. You did something, and observed the results. You just think of it as passive observation because you're taking for granted the enormous amount of prior knowledge that made it possible.
Reasonable observations would consist of using electrodes to monitor left leg movements vs right leg movements. Running vs walking, joint extension etc.
Build a waveform generator and compare the generated waveforms vs real waveforms and iterate.
(The scientists did actually do this.)
Once you've got a fair degree of confidence then you can try apply your work.
Paralysing a monkey just to replay captured data when we already know we can interface with the nervous system is totally unnecessary.
Edit: It should really go without saying that when it does come time to apply your work you do everything within your power to avoid harming a non-consenting creature. Does applying the work even require paralysing a monkey? Can it be temporary? What about sedation and then interfacing with the nervous system whilst sedated?
(Source: doi:10.1038/nature20118)
Is that kind of what you mean?
You need to debug this software without a keyboard, mouse, monitor. The language is an assembly language written by aliens with no documentation that uses quantum states in protons to manage the machine. Oh by the way, it's encrypted and obfuscated into nothing but MOV instructions. And there's 100 billion MOV instructions.
That's the kind of complexity we are talking about here. So get off your highhorse and let the experts who just made a paralyzed monkey walk do their jobs.
There are parts of this study that seem quite legit (another poster referred to them above).
However, the paralysis, surgery, and replay of data into the nervous system, just to be the first to "claim" they made a monkey walk was not one of them.
I already mentioned software is very different to biology, my issue is with reckless experimentation that has caused a lot of harm to a creature and provides no real gain. What were they actually hoping to achieve with this part of the experiment? It's already well understood that we can interface with nervous system, this is nothing new.
P.S. concern for the well being of other creatures isn't a particularly high horse. It's called empathy.
Here's a practical application for this research. Someone with a disease will eventually suffer from paralysis due to the degen. nature of the disease. Applying the technique from the research, we could record their body movements before they become paralyzed and then restore their ability to walk later.
I think that's a pretty real gain. Just because it's in a fledgling stage of research doesn't mean it's worthless. Medical research is a much longer timeline than even complex software. You make incredibly small baby steps precisely because you don't have tools like a debugger, logic analyzer, etc, etc.
Ok, let's not do anything then because actually putting this understanding to use involves touching actual brain matter and seeing what works.
Of course they use technology to observe and record normal and pathological "use". Do you think brain implants are made without using scanner, IRM...?
Research of this nature requires reading the electrical spike trains of small sets of neurons (often even individual cells).
EEG will give you a measurement of the combined electric field of every single neuron that is active in the entire brain. You can use multiple electrodes and run machine learning algorithms to get a qualitative measure of a macroscopic chunk of brain activity. So, terrible spatial precision.
MRI gives you anatomical information.
fMRI gives an idea of blood flow in the brain (about millimeter precision which is also not good enough) which is a time-lagged (a few seconds) _response_ to sustained neural spiking in a region.
Think of it like this: The brain is a computational organ where there is very little hardware-software separation. We're not going to be able to patch the software (say, to treat paralysis) without patching the hardware (say, with implants)
1. Humans do volunteer and it does happen, but even with full consent, regulations will always be stricter on human experimentation than on animals.
2. We don't understand the brain very well and it is easy to make mistakes when working with neural tissue. The consequences of causing irreversible damage/death are much higher when the brain belongs to a human being. On the other hand, most people eat animals for pleasure.
Cellular level investigations are obviously important but getting a wire in somewhere is not going to happen at a level this small
It does happen, there exist small electrode array implants that can be placed in neural tissue to both induce (for example, cause sensations that one can feel) and read-out activity. Such implants are in fact used in some human subjects (there are issues such as development of scar-tissue however).
A decent MR DTI can help when you want to know where the fibres go, but as you say, the hardware/software division isn't quite so clear with live things.
Yeah, for neural interfacing a connectome is not enough, you need to work at the neural code level. Reading out each spike with full spatial resolution from a distance will probably never happen.
Believe me, neuroscientists do not enjoy the fact that invasive experimentation has to be performed on blameless animals; unfortunately it has to be done, unless people are willing to live with the level of understanding/treatments we have now and never move forward.
"Cellular level investigations are obviously important but getting a wire in somewhere is not going to happen at a level this small It does happen, there exist small electrode array implants that can be placed in neural tissue to both induce (for example, cause sensations that one can feel) and read-out activity. Such implants are in fact used in some human subjects (there are issues such as development of scar-tissue however)."
What I mean - in a human subject the placement of any implant is not going to be guided by anything approaching a cellular level of imaging resolution. It will mostly use conventional radiological image guidance to get close and then measure electrical signals once close. Or is this in fact what you are saying?
Sorry about my formatting - phone.
Neural lace looks promising. But wouldn't be the first to volunteer for one :)
http://www.nature.com/nmeth/journal/v13/n10/full/nmeth.3969....
Sidenote: I feel slightly uneasy about all this stuff.
On one level it is progressing our knowledge, but both the methods of the research (animal testing) and potential implications of deep human+machine symbiosis, combined with CRISPR on animals and humans, feel instinctively bad. It won't stop at curing disease.
well, yes, but in a lot of neuroscience research, mainly the fundametal research, the initial observation stage led to the knowlegde that you first need controlled experimentation in order to perform any further valid/interesting observation at all
Honestly, if this was software, I'd consider this sort of approach as nothing short of pathetic. When I first reverse engineered firmware flashing software for mobile phones, experimentation was the absolute last thing I did - and all that experimentation could do was damage my phone, not another living creature.
Instead I recorded observations of software operating under normal use (USB traces) and identified patterns in this. Only once I'd spent a long time determining patterns and built a model did I decide it was safe to perform some experimentation.
To be fair, biological systems are much more complicated than computer systems written by other humans. However, observation should always come first! This study seems crude to me.