How do you know?

[Solmyr]

Suppose you have a telephone conversation with “someone”. You don’t know if the other party is a human or a sophisticated AI. The question is:

How do you know if the other party UNDERSTOOD your questions and propositions? In other words, how does UNDERSTANDING manifest itself?

 

[Bradski]

Understanding something is more than just knowing the facts.

It’s easy to memorise ‘the square on the hypotenuse etc’ but understanding it is another matter. If you are shown the step-by-step process by which it is proved, then you gain an understanding of the concept.

I think that if you were talking to someone and they gave you logical answers to questions, that wouldn’t show an understanding of the question. And if they simply gave you the impression that they were human (Turing test), that wouldn’t indicate understanding either.

But if the person/computer could extrapolate from the information given and ask you questions on the same subject, then that might exhibit understanding.

 

[fisherman_carl]

If its a computer you automatically know they didn’t understand. If its a human then you know they at least have the potential to understand. And they could communicate their understanding to you. Ultimately, you can’t really know unless they somehow demonstrate that understanding in a way that reveals it. Just repeating back what you said doesn’t necessarily mean they understand.

If a puppet looks human and looks like it is walking, does that mean it is walking? The computer is just a sophisticated puppet that does what the programmer programs it to do. It doesn’t understand anything anymore than the puppet.

 

[Tomdstone]

Suppose you have a telephone conversation with “someone”. You don’t know if the other party is a human or a sophisticated AI. The question is:

How do you know if the other party UNDERSTOOD your questions and propositions? In other words, how does UNDERSTANDING manifest itself?

In some cases you don’t know if the computer or if the person understood what you have said. The only way to be sure is to ask probing questions.

 

[po18guy]

How do you know if the other party UNDERSTOOD your questions and propositions? In other words, how does UNDERSTANDING manifest itself?

In that other party expressing the understanding to you, or to others. It is demonstrated via communication in its purest sense.

 

[inocente]

Suppose you have a telephone conversation with “someone”. You don’t know if the other party is a human or a sophisticated AI. The question is:

How do you know if the other party UNDERSTOOD your questions and propositions? In other words, how does UNDERSTANDING manifest itself?

You ask the other party to tell you their perception of your meaning, in their own words. It is, of course, a Turing test, which is a great way to avoid any possibility of prejudice.

 

[Solmyr]

Bradski, Tomdstone, po18guy and inocente all got the answer right. If the other party gives logical explanations and / or asks meaningful questions then we must accept that there is an understanding. In simple words, if the other party acts as if it were understanding - it exhibits the duck principle - then the question is decided, it understood the problem.

Fisherman carl gave the usual, dogmatic answer - and to the question which was not even asked. The OP did not ask to decide what kind of entity is on the other end of the line, only method to decide if there was an understanding.

 

[o_mlly]

Suppose you have a telephone conversation with “someone”. You don’t know if the other party is a human or a sophisticated AI. The question is:

How do you know if the other party UNDERSTOOD your questions and propositions? In other words, how does UNDERSTANDING manifest itself?

Buckley and Adler discuss the notion in a film clip at youtube.com/watch?v=wW-PeayslSA

 

[Aloysium]

Bradski, Tomdstone, po18guy and inocente all got the answer right. If the other party gives logical explanations and / or asks meaningful questions then we must accept that there is an understanding. In simple words, if the other party acts as if it were understanding - it exhibits the duck principle - then the question is decided, it understood the problem.

Fisherman carl gave the usual, dogmatic answer - and to the question which was not even asked. The OP did not ask to decide what kind of entity is on the other end of the line, only method to decide if there was an understanding.

Included in your view, would be mirrors which demonstrate a simple “understanding” in their reflecting perfectly our movements. Every subtly is captured and revealed to us; move one way or another it will always show that it “knows” what we do. The problem is that we may anthropomorphize an encounter, projecting our capacity to understand onto basic material processes, as complex as they might be.
Understanding involves a being who connects; that being can be known only through love.
You may have a different understanding of understanding which I seem to not fully understand. Understand?

 

[ffg]

Suppose you have a telephone conversation with “someone”. You don’t know if the other party is a human or a sophisticated AI. The question is:

How do you know if the other party UNDERSTOOD your questions and propositions? In other words, how does UNDERSTANDING manifest itself?

It seems understanding includes a time element and an experience element -

I know my younger kids don’t understand me because I repeat myself often for them to do a similar act as the previous day. Day after day.

My older kids understand because they’ve acted previously and don’t want to hear the message.

With someone over the phone who can’t see me pointing to the clothes on the floor, wondering when that moment will come where they stop walking over the clothes and start picking them up before I see it…

Good luck understanding without being in my house day after day.

The nice thing though is I gain in understanding too, by seeing the older kids ‘getting it’, I know there will come a time for the younger ones.

Ah patience.

Take care,

Mike

 

[Solmyr]

It seems understanding includes a time element and an experience element

We are talking about a conversation, nothing more.

 

[ffg]

We are talking about a conversation, nothing more.

I’ll play.

So we start here -

With someone over the phone…

Good luck understanding without being in my house day after day.

Then ask -

Are we assuming I am trustworthy to the ‘someone’?

 

[Solmyr]

I’ll play.

So we start here -

With someone over the phone…

Good luck understanding without being in my house day after day.

I am sorry, but I have no idea what your point is… in other words, I don’t understand you. The reason for this is simple - not enough information. What is the topic that would benefit from my being in your house?

Then ask -

Are we assuming I am trustworthy to the ‘someone’?

Such questions cannot be answered without the details of the conversation. So if WE are the two “someone”-s who are engaged in the conversation, then I don’t understand you. Trustworthiness has nothing to do with it. For the times being you did not give me any information. (The internet is almost as good as a phone. We have no extra information, except what we choose to give.)

 

[Bob_Crowley]

Suppose you have a telephone conversation with “someone”. You don’t know if the other party is a human or a sophisticated AI. The question is:

How do you know if the other party UNDERSTOOD your questions and propositions? In other words, how does UNDERSTANDING manifest itself?

I suppose one way would be to slip in nonsense questions. If you’re having a discussion about the weather, ask them if “Do you think the number 64 is the meaning of all existence?”

Chances are that unless the AI has been very thoroughly programmed, it’s going to respond in a repetitive way to that sort of nonsense eg. “I don’t understand what you’re saying.” So you ask it again, and get a similar response.

Whereas a human would change the response, or just hang up on you.

 

[Bob_Crowley]

To add to my post above - crack a corny joke, and see what the reponse is viz. "(Cautious note on behalf of the AI) - “Is that a joke?” “Yes” “Oh … (tinny laugh) ha, ha, ha…”

 

[inocente]

I suppose one way would be to slip in nonsense questions. If you’re having a discussion about the weather, ask them if “Do you think the number 64 is the meaning of all existence?”

Chances are that unless the AI has been very thoroughly programmed, it’s going to respond in a repetitive way to that sort of nonsense eg. “I don’t understand what you’re saying.” So you ask it again, and get a similar response.

Whereas a human would change the response, or just hang up on you.

You might be surprised that this is easy to program. Just have the program change its response a little if it would otherwise be the same as one it gave earlier.

Although it shouldn’t always do this, as humans sometimes do respond repetitively, for instance to push home a point or deliberately to annoy (known as “broken record”).

It’s probably too difficult to program the level of AI needed to have a genuine conversation. Instead it would have to self learning, along the lines of the techniques used to teach the recent Go champion computer, but more advanced.

 

[ffg]

I am sorry, but I have no idea what your point is… in other words, I don’t understand you. The reason for this is simple - not enough information. What is the topic that would benefit from my being in your house?

Such questions cannot be answered without the details of the conversation. So if WE are the two “someone”-s who are engaged in the conversation, then I don’t understand you. Trustworthiness has nothing to do with it. For the times being you did not give me any information. (The internet is almost as good as a phone. We have no extra information, except what we choose to give.)

Perfect. Excellent post.

me: Good luck understanding (without spending a little time in my shoes)

you: "I have no idea what your point is… in other words, I don’t understand you. "

Could we have a more perfect example that time and experience are needed for understanding?

We’ve known each other for what, 2 or 3 posts? The little time and detail that has been presented is causing an ‘I have no idea what your point is’ moment.

How might we solve? Well, conversing honestly and directly might help, along with a little time. But that’s no promise of understanding occurring.

How?

Well, was there a follow-up question? Was it answered directly?

If not, we’re not working toward understanding, rather avoiding the ability to understand.

Absolutely trust matters in understanding, even if folks don’t agree.

If I don’t trust you, I hope you wouldn’t want to converse with me. Especially if you knew I didn’t trust you, it would be a waste of your time (unless you get far talking to walls).

Take care,

Mike

 

[Solmyr]

Could we have a more perfect example that time and experience are needed for understanding?

Why should I disagree? The longer a conversation lasts, the better the chance of mutual understanding. But that is not all I am talking about. Without a mutually agreed upon vocabulary it is next to impossible to conduct a conversation. For example, you say that God is omnipotent and omniscient. My first answer is that I have no idea what YOU mean by omnipotent and omniscient. The same word does not mean the same thing to different people. That is just an example. We don’t need to explore those words.

Well, was there a follow-up question? Was it answered directly?

Of course I agree with this. You just described the trial-and-error and/or successive approximation method.

Absolutely trust matters in understanding, even if folks don’t agree.

Not that simple. For example, in a criminal investigations the inquirer does NOT trust the honesty of the suspect, but that is not an impediment to “understand” the answers.

 

[ffg]

Why should I disagree? The longer a conversation lasts, the better the chance of mutual understanding. But that is not all I am talking about. Without a mutually agreed upon vocabulary it is next to impossible to conduct a conversation. For example, you say that God is omnipotent and omniscient. My first answer is that I have no idea what YOU mean by omnipotent and omniscient. The same word does not mean the same thing to different people. That is just an example. We don’t need to explore those words.

Of course I agree with this. You just described the trial-and-error and/or successive approximation method.

Not that simple. For example, in a criminal investigations the inquirer does NOT trust the honesty of the suspect, but that is not an impediment to “understand” the answers.

I am so glad you said the bold, When I read some big words I often say, why? When small words work just as well.

The word ‘God’ alone should be able to be reasonably ‘understood’ fairly quickly to have all the fancy words someone can think up to mean 1 ‘being’ was, is, and will be, and that ‘being’ is the ultimate source of everything anyone can think up.

With regard to an examination of a person (the new example), I would argue understanding is not a concern, what is of concern is getting someone to speak.

In doing so, they will either dig their own grave, or not.

Understanding what happened is not necessary to accomplish the goal which is to move the case along.

In fact, I don’t know if understanding ever has to come to fruition.

I think if you asked prosecutors, investigators, and judges why someone kills, they might shrug in the same way you and I might say ‘how could someone do that horrible thing!’.

But I think you and I are building an understanding on some things.

So there is some experience to analyze!

Take care,

Mike

 

[PRmerger]

You ask the other party to tell you their perception of your meaning, in their own words. It is, of course, a Turing test, which is a great way to avoid any possibility of prejudice.

The Turing test is deficient, as shown by Searle in his famous Chinese Room Argument.

plato.stanford.edu/entries/chinese-room/