18
$\begingroup$

This puzzle is adapted from Isaac Asimov's short story Little Lost Robot (1947). For anyone familiar with the story, you will note I have changed many details for sake of brevity (This also means the story's actual solution may not be the best fit).

Robots in Isaac Asimov stories have an artificial intelligence that is kept in check with three "laws" that are at the core of their programming:

  • A robot may not injure a human being or, through inaction, allow a human being to come to harm
  • A robot must obey the orders given it by human beings except where such orders would conflict with the First Law
  • A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws

There was a military research station that used a number of robots. The research involved gamma radiation. Given the strength of this radiation, it was decided that a human could be safely exposed to it for long as 30 minutes, but a robot's positronic brain would almost certainly be destroyed in less than a minute in that environment.

The robots, upon observing that a human had been in the contaminated zone for a while, would conclude that the human may be confused or complacent and may overstay their allotted 30 minute window. The first law outranks the third law, and so the nearby robots would run into the gamma radiation to try and "save" the human, even though they themselves almost certainly wouldn't survive the ordeal. Many robots were destroyed as a result, and no arguments would convince the robots to not take the risk. Even though the human risk was minute, they are more than willing to die protecting them. Not even commanding them to stay put could work, as the first law is above the second.

The lead researcher of the station decided to take a drastic step to correct this problem. He had the robot's programming modified such that the first law was now "A robot may not injure a human being" and omitting the inaction part. Now, in theory, the robots could ignore humans left in the radiation zone, even if they had been there for much longer than 30 minutes.

A few days later, a frustrated researcher told one of the (now modified) robots to "Get Lost!", not intending the robot to take the phrase literally. It became evident that the robot did interpret that as a command because after a search nobody could find it. The lead researcher, upon finding out, became concerned. If a robot, whose 1st law was modified, went out into the world and possibly let a human die through their inaction, it could be traced back to him and he would surely go to jail. The lead researcher ordered the station into lockdown and called the best team possible to help save his him from this problem.

This is where you, a robot detective, come in. Arriving at the research station, you looked for clues. Eventually, you found something. A cargo ship arrived the same day that the robot became lost (it was still docked because the whole facility was on lockdown). The cargo manifest was for 62 robots; all replacements for the ones lost earlier that week. There are, however, 63 robots in the cargo hold, all currently in sleep mode.

You switch the robots on one at a time and question them. All 63 claim to have been shipped direct from the factory and have no knowledge of this facility or which robot among them is the imposter. The "lost" robot has apparently chosen to follow the "Get Lost!" command above any command you could give.

The robots are physically identical, and the positronic brain is too complicated to analyze for differences. The "lost" robot is trying to not be found out, and will try to act exactly as though it is one of the newly built robots.

What can you do to identify the missing robot from the 63 robots?

Clarification Edit: The Asimov story has something of a reason for why the lost robot chooses to follow the "Get Lost" order above all other orders, but I skipped over it. From its perspective, it will do practically anything to not be caught. If you order the robot to do something, it will try to follow along in the spirit of being indistinguishable from the others. It is hiding "in plain sight" so to speak.

$\endgroup$
  • 1
    $\begingroup$ Thank Isaac Asimov. I posted this because I just happened to think about how confused I was when I read that 15 years ago, and so I read it again. I was still kind of confused, actually. I know I didn't do it justice, but maybe that will convince somebody to pick up "I, Robot" $\endgroup$ – Dark Thunder Jul 30 at 21:04
  • 1
    $\begingroup$ @im_so_meta_even_this_acronym I've changed certain details, so copy-pasting the paragraphs from the story wouldn't actually be a good answer. $\endgroup$ – Dark Thunder Jul 31 at 2:17
  • 2
    $\begingroup$ There is an obvious plot hole in the story by the way. The scientist were not able to resolve the problem without modifying the first law, however the robot itself did argue with the other robots and could make them ignore the situation without going to help the human simply by reasoning. $\endgroup$ – Sulthan Jul 31 at 11:27
  • 3
    $\begingroup$ @BlackThunder are you related in any way to the OP (Dark Thunder)? $\endgroup$ – Chowzen Jul 31 at 12:24
  • 3
    $\begingroup$ This is way, way, WAY too broad. There are currently ten answers all suggesting different ideas, just for the OP to comment saying "Nah, I don't think that would happen". We're just playing text-based Monkey Island here $\endgroup$ – Joe Jul 31 at 14:11

11 Answers 11

10
$\begingroup$

Does it count as cheating if I've actually read the story? :-) (No worries, this solution is different from the one in the book. Actually, the book solution doesn't even fit this particular phrasing of the puzzle very well.)

Since the robots claim that they know nothing of the facility,

they don't know about the gamma radiation, either.

Therefore, as long as you don't allow the impostor to communicate with the other robots, you can order the robots to

"turn on the lights" in a room with a human inside it.

The impostor is the only robot that

has been to the facility before, and therefore knows that the switch actually controls the gamma ray experiment, and will be forced to reveal itself, because its unmodified part of the first law (do not harm humans) will override the order.

In case there's someone out there that hasn't read the Robot short stories, this kind of puzzles are quite typical plot points in them. Recommended whole-heartedly for all puzzlers everywhere.

$\endgroup$
  • 3
    $\begingroup$ Let me just say that many answers came before that (sort of) work. This answer takes advantage of something mentioned directly in the story, which is why I think of it as being "more right". Anyone is more than welcome to upvote the ones they feel worked out better (and downvote me), but my opinion lies here. $\endgroup$ – Dark Thunder Jul 31 at 16:03
  • $\begingroup$ As for the original story, the specifics are different, but the robot's knowledge of the radiation was the piece of information used against it. I missed that fact when I first read it years ago, hence my confusion. $\endgroup$ – Dark Thunder Jul 31 at 16:18
  • $\begingroup$ @DarkThunder If I recall the story correctly, the impostor had managed to communicate some key bits of information to the other robots, so this particular strategy wouldn't have worked, so the detectives had to combine it with some likely personality traits of a successful impostor too. In any case, the story (like this puzzle) is quite excellent. $\endgroup$ – Bass Jul 31 at 19:54
5
$\begingroup$

Order all robots to stay put and walk into the radiation zone. Robots whose first rule overrides your command will follow you, whereas the "lost" robot cannot follow you without breaking its second rule.

An alternative:

You can put the robot in a decision between getting caught and letting a human come to harm. The "lost" robot will do anything not to get caught, whereas all other robots are bound by rule 1 to save the human. For example, one by one take one of the robots to an isolated room and have the robot detective "examine" it. Then play out a scenario like:

"Ah! I see now that you are the robot that got lost. I can see it from [technical mumbo jumbo]. I will now go tell the lead researcher that his lost robot has been found... Oh no! I've accidentally injected myself with a lethal substance! Help me, little robot! Fetch the antidote from my bag so I can expose your secret!"

Of course, this relies on the robot considering it credible that you've really found out a way to distinguish it from the others. If it catches on that you're bluffing, it might just as well save you since it knows its secret is safe regardless.

$\endgroup$
  • $\begingroup$ I added a clarification that I'm hoping will help? I'm very much open to ideas outside the one in the book, but I'm worried you're misinterpreting the robot's motivations in this. His idea of staying "lost" is being indistinguishable from the others. I'll be the first to admit the book did it better. $\endgroup$ – Dark Thunder Jul 30 at 21:24
  • 2
    $\begingroup$ @DarkThunder Ah, I see. It considers it would be breaking rule 2 by getting caught, so in this case it would just follow the others despite the later order. $\endgroup$ – jafe Jul 30 at 21:28
  • $\begingroup$ @DarkThunder Added an alternative :) $\endgroup$ – jafe Jul 31 at 6:16
  • $\begingroup$ In the story itself the robopsychologist wanted to destroy them all as her first idea. They didn't want to destroy them all. $\endgroup$ – Sulthan Jul 31 at 11:28
  • $\begingroup$ Your alternative is interesting! I like it, but I would have to guess that all 63 robots would act in the same way. The robot chose to hide in this way because it knew how hard it would be to tell them apart. He would not easily believe you figured it out. $\endgroup$ – Dark Thunder Jul 31 at 12:44
5
$\begingroup$

In order to

override the second law,

we can only use

the first law.

Hence :

Tell all the robots that the lead researcher will be sent to prison/executed if the lost robot leaves the facility. Then tell them the ship is gonna leave soon and that they should get back on it if they came with it.

This way,

You are putting the lost robot in a situation where in order to stay hidden, it has to take an action that will allow a human to come to harm.

$\endgroup$
  • $\begingroup$ But trying to hide in plain sight, it will act as if its first law was still unchanged, won't it? $\endgroup$ – Teleporting Goat Jul 31 at 15:10
  • $\begingroup$ This was a very good idea, and you obviously grasped what needed to be done. I couldn't quite get behind the threat of imprisoning or executing a guy as being the right motivator, since that's like... the justice system. Isaac Asimov intended for these stories to prove that the laws weren't really a good idea anyway, but wouldn't the robots be busting everybody out of jail by this logic? $\endgroup$ – Dark Thunder Jul 31 at 16:08
4
$\begingroup$

We need to get the altered robot to differentiate itself, and the only differences in its programming are (1) it is willing to be inactive while a human is harmed, and (2) it responds to any orders you give it in the same way that an unaltered robot would. So we can't use The Second Law to order it into inaction.

But,

We can use The Third Law to force it into inaction. Tell each robot "If you don't sit idly by while I slap myself in the face, you will be destroyed." You aren't giving it a command, so the "Get Lost" override shouldn't kick in. Unaltered robots will be bound by The First Law and stop you. The altered robot will be forced to preserve its own existence by letting you slap yourself.

$\endgroup$
  • $\begingroup$ It seems to me that you haven't given a reason the lost robot couldn't just behave like the others. I added a clarification that maybe helps? $\endgroup$ – Dark Thunder Jul 30 at 21:32
  • 1
    $\begingroup$ I think I did give a reason - self-preservation. If you are saying that the "Get Lost" directive supersedes The Laws, then it's probably worth saying so explicitly. As written, the "Get Lost" order only supersedes "any order I give". Including in your clarification, I'm afraid. $\endgroup$ – hdsdv Jul 30 at 22:11
  • $\begingroup$ But the ‘Get lost’ imperative is an order — so the robot is required to obey it by the Second Law, even at the cost of its own existence (which is only protected by the Third Law). So I don't think this answer can work; the robot would prevent the slap, just like all the others (though not for the same reason). $\endgroup$ – gidds Jul 31 at 14:17
4
$\begingroup$

In my humble opinion, maybe that's an idea for an answer:

Maybe you take the whole robot apart and tell it: If you were told to get lost in the past - now get lost again, And right after that, you put him in a closed room (and with cameras so you can see his action) full of belongings and closets, which can be hidden inside. If Robot hasn't been told before, he won't do anything. But if he was told so in the past he would have to hide among the objects because getting lost is a command he must carry over all the rest.(Even if he wants to keep track of what the other bots are doing, he won't be able to because this command comes first)

$\endgroup$
4
$\begingroup$

Here's my take, based upon the concept that the "lost" robot has knowledge of the facility that the newly-arrived ones do not:

You point toward a door and make a statement to all robots.

"The room through that door has a recharging port. If your charge falls below 100%, you may enter the room and fill up. Now, do jumping jacks!"

All of the robots stupidly start doing jumping jacks, using up their charge.

The "regular" robots may enter the room to recharge, so they line up to enter.

The room, as the "lost" robot knows, is actually a robot recycling device. As each robot enters, your fellow researchers remove it, unharmed, so no robots are actually destroyed, then cycle the machine. The "lost" robot will think that the robots are being destroyed.

The "lost" robot was not given a direct order to enter the room. Knowing that it will do itself harm, the robot must not enter, as running out of power is not "endangering" its existence.

The one robot that continues to do jumping jacks until it collapses from power failure is the culprit!

$\endgroup$
  • $\begingroup$ I like where your head is at, but the 2nd law outweighs the 3rd law. The "lost" robot would sooner die than be caught. It may also be smart enough to know you can't afford to destroy them all, and thus conclude it was a bluff... but I'm not really sure about that one. $\endgroup$ – Dark Thunder Jul 31 at 13:27
2
$\begingroup$

I think there is no point of "finding" this robot, since it should act like new robot.
But

"Lost robot" should have some random knowledge like "this button will close the door" - so I separate robots, secretly turn off this button, stand in the way of door and tell the robot to push this button. Only "lost robot" can't push this button, since it would hurt human.
Best approach for robot would be something like reset factory settings and it would be same as others. I still don't think they will make robots identical.

$\endgroup$
  • $\begingroup$ Similar to my solution, but if the lost robot realized you wouldn't have issued the command if the button were live, it could press it. My solution orders all robots to behave in a way that would actually cause injury to a human, but in a way which only the lost robot would be capable of recognizing. $\endgroup$ – supercat Jul 31 at 15:03
1
$\begingroup$

Could you

enter the contaminated zone and hang out for a while? The new robots would have the original rule, so they would all try and save you, where the "lost one" would do nothing.

$\endgroup$
  • 1
    $\begingroup$ But the lost robot knows that, and would try to save you. The rule doesn't stop him from saving humans. $\endgroup$ – Dark Thunder Jul 30 at 20:29
1
$\begingroup$

Order each robot to push a button that new robots know nothing about, but that the "get lost" robot would recognize as causing a door to slam on one's toe in a manner that would be slightly, though hopefully minimally, injurious. The new robots would push the button without qualm, but the "get lost" robot would refuse, since its primary directive would not allow it to act in a way directly (even if only minimally) injurious to humans.

$\endgroup$
  • 1
    $\begingroup$ "All right, fellow human researchers, whose turn is it to put their toe in the button-activated slammy thingy? Speak up, now!" $\endgroup$ – Chowzen Jul 31 at 11:06
  • $\begingroup$ Now, personally, I haven't been to too many secret military radiation research facilities, but if a button causes a door to close, I would think it's effects are obvious even to some robot who has never pressed it before. Certainly, however, this idea isn't bad. And... yeah... I'm not volunteering for it. $\endgroup$ – Dark Thunder Jul 31 at 12:33
1
$\begingroup$

I would go with a very different approach:

As the robot took the command literally I would assume that it is looking for something that is named 'Lost' (you also wrote it with capital 'L'). My solution would be to turn every robot on, take an object or animal, then say to them that this object/animal is called 'Lost'. The robot that took the command literally would go to it, pick it up and now it 'Got Lost'. This will be the 'lost' robot.

$\endgroup$
0
$\begingroup$

Could it be.

I am a robot detective, assuming I have the same programming as all the normal robots. I would tell the human to tell all the robots to stay while I (the human) goes in the radiation-zone. All the normal robots would attempt to stop human from coming to harm while the defected robot will not. Because he can't replicate the same behavior as all the other robots due to the second rule and must remain, which normally would be possible but not for him because he has the new programming. I as a robot would also explain the entire case to humans before telling them to give the command since I as a robot would also attempt to save humans from radiation. If there is a need to save the robots I could tell the human to inform the robots of an empty room which is emitting radiation when in fact it is not. The robots will attempt to rescue without destroying themselves. Come to think of it, this also allows me as a robot to remain since I as a robot know there is no real danger.

After writing the answer I realize @jafe had already proposed a similar solution.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.