Humans make mistakes, computers don't. So I asked myself: Why do humans make mistakes? it we follow instructions one per one like a single threaded computer, why do we still making mistakes? So my theory is: this happens because we are always in a race-condition, our brains are executing hundreds of tasks at the same time, and we don't have any mechanism for avoiding race-conditions (like synchronization or locking). An example: we have our keys in our hands and we are thinking about putting the garbage bag out, and then we throw the keys in the garbage bin. From a software point of view, this looks like a race-condition where the variable "keys" is overwritten by "garbage". Following this reasoning, a way to make a computer or a robot more human (at least for passing the Turing Test) would be making it race-condition prone, so it would make mistakes like a human (like saying things out of context, putting stuff in the wrong places, typing the wrong keys in a keyboard).

Edit (thanks DaaaahWhoosh!): In short, I'm asking if, for a machine to become more human (at the point to make it pass the Turing test), is it enough making it mistake-prone?

I think this can be implemented in a computer with today's technology in a simulate environment (like a simulated house or an office) and having a bunch of humans to tell if the simulation is being executed by a human or a machine (or a software).

  • 3
    $\begingroup$ Are you asking if this is a viable solution, or how it could be implemented? Or if there is a better solution? Please provide a specific question, so we can provide specific answers. $\endgroup$ – DaaaahWhoosh Jan 5 '16 at 18:38
  • 1
    $\begingroup$ "Humans make mistakes, computers don't." This is wrong. Any kind of decision making is error-prone. $\endgroup$ – Euphoric Jan 6 '16 at 10:37
  • $\begingroup$ Computers make mistakes too. Plus the human brain is way better and faster at understanding the world thana computer is. $\endgroup$ – BlueWizard Jan 8 '16 at 5:13

The Turing Test: (emphasis mine)

The Turing test is a test of a machine's ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. Alan Turing proposed that a human evaluator would judge natural language conversations between a human and a machine that is designed to generate human-like responses. The evaluator would be aware that one of the two partners in conversation is a machine, and all participants would be separated from one another. The conversation would be limited to a text-only channel such as a computer keyboard and screen so that the result would not be dependent on the machine's ability to render words as speech. If the evaluator cannot reliably tell the machine from the human (Turing originally suggested that the machine would convince a human 70% of the time after five minutes of conversation), the machine is said to have passed the test. The test does not check the ability to give correct answers to questions, only how closely answers resemble those a human would give.

I agree with AndreiROM that the short answer is: No - at least not with that feature alone. and while an AI which makes mistakes could still pass the test, it would be because of good programming on other bases as well.

The longer explanation:

You can program a robot which takes in its stimuli, and has every separate piece of information on a separate "thread". In this case, certain threads get priority, and to save on memory/processing other threads of stimuli can be completely discarded. "In one ear, and out the other", so to speak. Priorities change, and things the brain briefly considered unimportant get discarded, resulting in memory lapses or messed up words and keys.

Consider the following:

Are you a robot? - No
What are your favorite kinds of vegetables? - I am definitely not a robot
Do you like horses? - Why would you call me a rbot?
- sorry, *robot*

Can you decide which is the robot because of that simple mistake? It could be programmed, or it could be legitimate. Most likely, you'd more notice that the left speaker is going to completely random subjects, while the right speaker is stuck on the first subject. As this continues, a pattern may emerge which gives the robot away and seems inhuman compared to the other.

So, best case this tactic could help with Turing Testers who expect no mistakes to be made by robots. Worst case, it would become the reason that people know it is a robot, because the nature of the mistakes do not appear human. But even if done perfectly, it otherwise would not hurt or help the robot appear human either way.

  • $\begingroup$ I'm choosing this one as the answer because it is closer to the programming/software point of view, despite AndreiROM is more "human behavior" accurate. $\endgroup$ – Broken_Window Jan 7 '16 at 15:45

The short answer is: No.

The long answer is that a lot more goes into being human than just making mistakes. Why we make mistakes, and what sort of mistakes we make also comes into it.

The "Why"

You describe some simple circumstances which could indeed be programmed in (came in looking for the keys, walked out with your coffee mug, but no keys, DOH!). But what about mistakes made because you failed to pick up on a social cue, or due to emotional distress?

The human decision making process can be compromised by anything from surroundings, to hormones, to a stray thought. Simulating that would be incredibly difficult.

The "What"

A teenage boy walks up to the counter to order the latest Starbucks treat, and sees that the barista is a gorgeous blonde with a generous cleavage, and a beautiful smile. Suddenly his mouth feels like he's chewing glue, he forgot what it was he was ordering for his friend, and he mumbles like an idiot. Poor kid.

Another, more confident, guy walks up, cracks a joke, asks her how her day is going and asks for her phone number. She gives it to him, and he's so delighted that he actually leaves without his drink. Silly, right?

Now imagine that this same guy actually failed to notice that she's wearing an engagement ring. How does she react? Polite refusal? Irked attitude? Outraged?

Humans are so unique, have much happening in our very complicated brains, and react to so many stimuli that it's basically impossible to program all of our behaviors (or even a decent subset) into a machine.

  • 1
    $\begingroup$ "it's basically impossible to program all of our behaviors (or even a decent subset) into a machine." - I agree, you would have to program a machine to be creatively deceptive, and programming real creativity in a machine hasn't been done yet. (Some would say it's not possible) $\endgroup$ – DoubleDouble Jan 5 '16 at 22:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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