fMRI Lets Israeli Student Control Robot In France With His Mind

MrSeb writes "An Israeli student has become the first person to meld his mind and movements with a robot surrogate, or avatar. Situated inside an fMRI scanner in Israel, Tirosh Shapira has controlled a humanoid robot some 2000 kilometers (1250 miles) away, at the Béziers Technology Institute in France, using just his mind. The system must be trained so that a particular "thought" (fMRI blood flow pattern) equates to a certain command. In this case, when Shapira thinks about moving forward or backward, the robot moves forward or backward; when Shapira thinks about moving one of his hands, the robot surrogate turns in that direction. To complete the loop, the robot has a camera on its head, with the image being displayed in front of Shapira. Speaking to New Scientist, it sounds like Shapira really became one with the robot: 'It was mind-blowing. I really felt like I was there, moving around,' he says. 'At one point the connection failed. One of the researchers picked the robot up to see what the problem was and I was like, "Oi, put me down!"'"

Comments

Re:Old news

[NettiWelho]

Israel has been mind-controlling America for a long time.

But now with computers!

what's next

[ozduo]

mind melding with your partners sex toy

Re:what's next

[theshowmecanuck]

I can't help but thinking the first person to do this will be thinking, "It's dark in here!"

Re:what's next

[tehcyder]

I can't help but thinking the first person to do this will be thinking, "It's dark in here!"

You'll just have to employ the Doom 3 duck tape mod and wrap a torch to it.

Next Step

[Phrogman]

is to combine this technology with remotely piloted drones, spy-copters, and eventually combat robots. Then I can imagine a military formation formed up to receive orders, being told they were going to war, and then told to go to it - and no one needs to move :P

Re:Next Step

[Anonymous Coward]

If such technology existed, why would that be better than a keyboard and mouse or a gamecontroller as an interface? And using my extensive experience in an environment where they have been used for "combative" purposes I ask: why is a huge latency desirable? And how is a human better than an aimbot?

Human lives are cheaper than advanced combat robots (with or without AI). The notion that no sacrifice in blood has to be made in war might be alluring to democratic politicians and their constituencies, but the sooner democracies in the world stop initiating wars the better the world will be off. Thus I am happy to say that due to economics drone armies will not conquer the world*, either under the pretext of humane intervention or blatant imperialism.**

* Conquest is long lived occupation (with repression) not mere military victory. The technological saturation point of destruction has already been reached, no further advancement is required.
** If you read between the lines of this comment a "Fuck the USA" can be spotted.

Re:Next Step

[superwiz]

The only reason you think of a military formation is that modern society is so far removed from necessary physical interactions with nature. Why military? This could be farmers controlling robots which gather crops.... computer connections are cheaper than transporting people in actual physical space.

Re:Next Step

[theshowmecanuck]

The first money maker will be a Dutch sex-bot, remotely operated by a hooker from Amsterdam.

Interesting

[jd]

Most of the comments so far seem to confuse EEG-based interfaces with fMRI-based ones, or local with telerobotics, but no great surprise there. The politics is sad, but again only to be expected.

We still can't do true mind-control of robots (there's no way to read minds yet, we can only say "pattern X equals action Y", which is not the same thing) but this is an interesting development to say the least. Think in terms of medicine. Robotic interfaces in surgery are typically data gloves or joystick, plus goggles, but muscles have poor granularity of control, data gloves and joysticks reduce this further, and goggles are incredibly low-res. If they get to the point where surgeons are limited only by the precision of their mind, you're looking at a major revolution.

Re:Interesting

[Baloroth]

The bigger issue, it seems to be, is feedback. Sure, you can train the machine to "read" certain patterns with attempts to move the arm, and potentially create very advanced interfaces, but the interface is purely one-way: there is no way to tell the human he has "touched" something. Cameras work to some extent to provide visual feedback, but more advanced and more delicate control requires something beyond just that. We need to find a way to provide neural feedback to replicate the sense of touch, at the very least. Sight can be provided easily (without requiring a neural interface), as can hearing, and smell is largely unneeded, but for an arm, touch feedback is essential.

Re:Interesting

[jd]

Agreed. Force-feedback is a start, albeit a crude one, but it's not enough. It might be possible to electrically stimulate individual pressure nerves to give a sense of touch, since nerves are electrical by nature, but you're talking an amazing number of electrodes to get any detail and some major technological problems to get it to stimulate the right nerves.

For something that is compressible/expandable to some degree along only certain directions, you can simulate that with pneumatics. It's essentially the same as force-feedback (you apply pressure in one direction, something applies force in the opposite direction) but instead of having one or two motors, you can have a crude surface where each point applies different feedback. Mechanical devices of this kind aren't complex, require no new technology to be invented, and would be in the price range of a decent facility - I assume you don't hear of them because there's simply no scientific or industrial application outside of perhaps telerobotic pottery-making and there's not really a huge market for that, and the increase in the number of variables that could be fed back to the user is still going to be extremely small - an increase from 2 to 12 sounds reasonable - but the cost would be substantially more than 6x that of a joystick. The cost/benefit isn't there.

Re:Interesting

[superwiz]

This interface doesn't read the neural impulses. It does the detection by determining which parts of the brain receive more blood.

Re:Interesting

[theshowmecanuck]

(there's no way to read minds yet, we can only say "pattern X equals action Y",

But isn't this what you are doing when you learn how to do something physical like play a guitar or pianno, or tennis, etc? Your mind equates how you move your fingers or body to a result or external action on your environment and you say to yourself, that's good or that's bad and you store the information. So now you know that if your fingers do X, you get Y note from the instrument, or you hit the ball. As far as I can tell, and I'm not a neurobiologist or whatever discipline studies this, but as far as I can tell, that is what creating muscle memory is (in terms of being able to easily repeat something you practice that is). The more you do X to get Y the easier it is for you to retrieve and use that data and required action.

Don't need thought control to feel you are there

[caseih]

Ask any good backhoe operator about how he operates the machine and you'll find he doesn't think about the mechanics of his arms and feet interacting with the control levers. His brain abstracts all that and treats the hoe as an extension of his body. Once you've been trained how to move the controls, you stop thinking about it. You just dig.

A similar feeling could be generated simply by video goggles and a joystick. In fact when I fly my airplane using a video downlink, it feels like "I'm there." Seeing yourself on the ground is a bit weird! I can look down at something, turn the plane to look at something all without really thinking about what my hands need to do, since they've been trained and my brain just does that automatically in response to what I want to do. This is true of normal RC airplane flying as well. People often ask me how I can remember to move my fingers in the opposite direction as the plane is flying towards me but the truth is I don't think about it at all very much. I just move the airplane where I want it to be.

The exciting goal of thought control, though, is obviously to enable people who don't have the use of limbs or fingers to control and interact with robotics, such as an artificial limb, as if it is part of the body. And as the test subject can attest, that's pretty much what happens with training. Now if they can just get the sensor equipment to weight less than a few tons and not draw metal objects towards it...

Passport

[jamesh]

If you are in Israel with a physical (but virtual) presence device in France, do you need a passport?

Brain vs. mind

[colinrichardday]

Shouldn't that be brain instead of mind?

Re:Brain vs. mind

[Anonymous Coward]

No. Using the word 'mind' simply indicates that the brain activity which controls the robot is conscious. So it specifies rather than confuses.

and everyone but me hated Caprica!

[dicobalt]

Now you have real school kids controlling real robots in real life with their real minds.

fMRI

[guttentag]

For a moment I thought this was going to be an article about how doctors can now get a clear picture of your brain by analyzing your Facebook profile, reducing the need for a traditional MRI, but somehow costing more.

2000km?

[hcs_$reboot]

I'm much less impressed by the 2k distance than the ability to control a device from the mind. The robot could be close to the student or - even better - on him [iron-man like], the achievement would be just as great.

Yes, but ....

[PPH]

... you must think in French.

The other direction would be more interesting ...

[PolygamousRanchKid]

. . . how about a robot controlling a human with its mind . . . ? Now that would be definitely more interesting, and would foster more vivacious cocktail party talk.

Temporal resolution

[Harvey Manfrenjenson]

I don't think an fMRI interface is going to be very useful for controlling robots, because of the issue of temporal resolution. I think you can only acquire an fMRI image once every couple of seconds (at most). The above poster referred to the "granularity" issue with data gloves and joysticks, but it's a thousandfold worse with fMRI and probably always will be.

A better choice might be magnetoencephalography (MEG). Nearly instantaneous "image" acquisition, and as a side benefit, there are no health risks to the user (fMRI bombards you with intense magnetic fields and no one really knows if that's safe).