I recall not that long ago Russia talking about a supersonic intercontinental range self homing torpedo. Forgive me for not worrying to much. I also recall back in the 1950s the USSR tricking us into thinking they had a nuclear powered bomber with unlimited range. Before long our designers were talking about building thin walled nuclear reactors to power our aircraft. The aircraft would be based on isolated islands and manned by pilots who had already had children. President Eisenhower put a end to this talk in his famous "Military Industrial Complex" speech.

As far as Col. Brecher's assertion that human-supervised systems would be more vulnerable to hacking, I take the opposite view. A fully autonomic system could be compromised without anyone noticing---the first warning would be a major battle disastrously lost. On the other hand, a human operator would have a fair chance of noticing anything odd. That is why all industrial plants, even fully automated production lines, always have a human operator supervising things. The assertion that the presence of the operator makes the plant more hackable is preposterous. And "vulnerable" communication links would not be eliminated anyway by making the robots autonomous---human soldiers are more autonomous, and have more sophisticated neural nets, than any current or prospective robot, yet still need to communicate on the battlefield. You cannot have an army without C&C, and no future development can ever change that. <p> I also mostly agree with William---Russia regularly makes announcements like that for propaganda leverage. Still, it would be prudent to keep an eye on the advancements in the field, to be able to judge when this hype starts to turn realistic.

Why not make a nuclear plane manned by only robots? No shielding would be necessary then. It could then fly for months without landing to refuel.

I have wondered for years why we did not have a weapons system that could automatically fire back if fired upon. A system that could immediately identify the source of the hostile fire and return fire within milliseconds. Such a system would not have to be very large and would be an efficient defense against snipers by locating their position and returning fire quickly. It could also be used to defend positions that were vulnerable to attack. As far as AI goes for fully autonomous robots, I doubt they will reduce civilian casualties. When unethical enemies use civilians for cover and children to carry suicide bombs or hide in schools and hospitals, there will be no choice but to shoot. It won't matter if AI is making the decision or a human.

The scope for unintended consequences would appear limitless...

A UN Ban, that will work wonders. Gort?

The article says, "The UN has now formally convened a group of government experts..."
OK, that's it! Nothing to see here, move along.
This will be resolved in NO TIME!

Why not imagine a world without war? To those who say that will never happen, of course you are correct. If only by the unwillingness to try to accomplish it. War is big business and it is perpetuated not by man's nature, it's marketing, advertising and propaganda.
@Vincent, you are bit behind the state of the art. UAVs are remotely piloted, no need for a "futuristic robot" to manage the controls. Very shortly there will be fully autonomous UAVs, no need for a shift change.

Anyone who thinks Russia will be deterred by the UN, or "political correctness" had better think again. I know that about half the country has been converted to the idea that Russia is "OK". Let me tell those people that we have NEVER caught the Russians doing ANYTHING beneficial for the Human Race. EVER.

Queue up the skynet history lesson....
https://www.youtube.com/watch?v=4DQsG3TKQ0I