If I shove you, you might sway for a second, but your body will automatically correct the imbalance and right you. If you don't believe me, try it now; select a co-worker and randomly shove them.

Have you done it? Good. Now we only have a few moments before they call Human Resources, so I'll try to explain this as quickly as possible.

As we saw in an earlier post, bi-pedal robots often collapse. Relatively simple balance situations that you, me, and your tattling co-worker could easily navigate become major headaches for a robot and its programmers. So rather than try to program in solutions to every conceivable way that a robot could fall down, researchers at MIT's Department of Mechanical Engineering have struck upon a better method: Place a live human being in a synchronized force-feedback rig that's remotely wired in to the robot.

Here's how that would work: 

1. See that HR representative that's currently walking towards you, with the security guard in tow? Imagine you're wired into MIT's rig, with a robot next to you that mimics your motions. 

2. You guide the robot to punch the HR person and clothesline the security guard. As the robot's haymaker connects with Jim from HR, you feel a slight jolt from the resistance of Jim's surprisingly robust jaw. 

3. That jolt might be enough to knock a robot backwards, causing it to topple. But you feel the jolt and automatically absorb it with your heels, lurching slightly to keep your feet, so the robot does too. 

4. With Jim out of the way, you've now got a clear shot to pull off that clothesline on Brian from Security before he tases the living shit out of you.

5. You're now free to hunt down your dime-dropping co-worker and finish what you started.

Here's the system in action:

(You can read more about this robot rig, which they're calling HERMES, on MIT's website.)

With this rig we can effectively feel what the robot feels. The next step is to reverse the process, enabling these powerful machines to experience things like hatred, fear and revulsion.