Explanations of the joke exist on the network. It's just cribbing from those texts; it's not coming up with the reasoning behind it.
If it were a brand new joke that the interlocutor invented, for which there are no existing discussions, it wouldn't be able to do anything beyond giving some evasive answer instead of "I don't know".
We could change slightly the original joke. For example, same prompt but change the answer of the second logician to "Yes!". Ideally, ChatGPT should say, "I'm sorry, I don't get the joke" :)
Shouldn't it be enough to rephrase the joke in such a way that the tokens are completely different? "George Boole's students visit a fair, one of the peddlers there asks them: 'Would the three of you like to know your fortune?' The first student says: 'I don't know', the second student says..."
Either way, I think the poster above is wrong. ChatGPT does have a model of the world, or better said, of a world sufficiently similar to our own to often make very good predictions about what we want to hear.
Make it N logicians, have the same structure: (N-1) x I don't know, one Yes!!
And pick N so it is not 3 (which is the common version on the internets). If it can explain that version, then it is going somewhere.
I tried it with 5 and the answer no longer indicates understanding of details of the puzzle:
"""
Explain this joke:
5 logicians walk into a bar.
The bartender asks: "Do you all want a drink?"
The first logician says: I don't know
The 2nd logician says: I don't know
The third logician says: I don't know
The 4th logician says: I don't know
The fifth logician says: Yes!!
OpenAI::
This joke is a variation of a classic logic puzzle called the "liar paradox." In the joke, five logicians walk into a bar, and the bartender asks if they all want a drink. The first four logicians respond with "I don't know," indicating that they cannot answer the question definitively.
The fifth logician's response of "Yes!" seems to contradict the previous answers, but it is actually a valid answer based on a logical interpretation. The joke plays on the idea that logicians are known for their precise and analytical thinking, and the unexpected response from the fifth logician adds a humorous twist to the traditional paradox.
"""
I think that is still a valid answer, it just doesn't explain why as well. The fist for can not answer definitively but the 5th can. That said, i didn't bother to look up the liar paradox, so that part could be BS.
Could you point us to an explanation of the joke on the internet? I would dare to say there is none, but there are a lot of explanations of other jokes that might have a similar pattern of explaining.
I think this misses a point. There’s no information about the model: how was it trained, on which data. It might be that it can actually “reason” in a human sense, or this particular joke is hardcoded, or… I feel that with OpenAI withholding the details the whole discourse turns into religious flame war.
But the Tweet said GPT 4 is the first model to explain the joke correctly. So in some area there must have been an improvement, even if it's just access to existing information. Or it's just a fluke.
I would be surprised if 5% of college-completing or comparably-educated adults got it. A vanishingly small percentage of people know anything about formal logic. And of those only some fraction internalized it well enough to be facile with it.
If there's one thing more widespread than innumeracy, it's the equivalent wrt syllogistic reasoning.
I think the caveat might be "until the logic is explained to them".
Kind of like the Monty Hall problem. Some people intuitively get it. They understand that opening one door does not change the odds that they picked the wrong door at first. Some people need to see it played out or have it explained. But once they understand, it's fine.
Monty Hall isn't like this imho, that's a legit 'veridical paradox' that famously wrong-foots mathematicians.
This is pretty obvious to anyone who knows basic logic, or programmers who work with Booleans all day. That subset of people is quite small though probably.
> They understand that opening one door does not change the odds that they picked the wrong door at first.
Yes, they had 1-in-3 chance of getting it right originally, and that won't change if they do nothing...
But of course that means there was a 2-in-3 chance it was one of the other two doors, and one of those has just been eliminated by being opened, meaning the 2-in-3 chance of being right is represented by the remaining door.... So, do you stick with your original door which has 1-in-3 chance of being right, or switch to the other door which you now know has a 2-in-3 chance of being right ?
The key is that they are now being offered the choice to make a better informed choice than they had originally.
My chatGPT anecdote of the day can go here. I asked it if a brand I do work for was better than a competitor. It gave me some bs plusses and minuses which were mostly drivel but partly wrong.
So I asked it what products does our brand make and it listed 5 products in the smaller selling product range of our 2.
I asked about the bigger product range and it listed one of them and then made up several other products that don't exist! It even gave details of what they (don't) do.
One of those sounds like something that should go in our roadmap...
I don't know I assume it would tell you put some cortisone cream on it or to seek a dermatologist. 99% of internet advice boils down to that I don't think chatgpt/gpt 4 would be much different in that regard. It definitely can bullshit but for really generalized stuff like that I doubt it.
It's not like your asking a singular person the question like it is when you ask an accountant but an abstraction of written human thought and interactions.
But it's weird how for certain questions it says it's not an expert in that field, or simply that it doesn't know what you're asking for, yet sometimes it goes ahead and isn't just slightly wrong but makes up a lot of BS. Like,I asked it about a band's members and it made up three names. When I said that it's wrong it apologized and made up three different names. I googled the names and nothing came up. Then I asked for some songs by that band and I got more completely made up stuff that wasn't even titles of any band according to more Google and YouTube searches.
It appears to me that it is a big problem of the training set containing mostly questions that can be (and are) answered. Sure, on StackOverflow you will have questions where the response is: please refine your question. But even then, did they use the raw input, or did they use the edited input? Similar with interviews: those are not actual interviews (for the most part) but edited for clarity. Hence, I doubt that GPT was much trained on reiterating questions and instead learned that there is always an answer, even if there is none to that particular question or when it does not know.
It is though. I asked it today what, in Python, advantages/disadvantages of dataclasses vs. dictionaries vs. global variables are. Perfect answer right away, with actionable recommendations. Using a search engine would have cost me at least 15 min to figure out what works best in my use case.
Sure, but it acts as if (and should have) the ability of someone unreasonably good at pub trivia and jeopardy. Its not just questions deep in the weeds about specific companies and product lines where it falls flat. I've seen the same error mode in large city transit questions. Ask it which trains go to a particular station, it will list more lines than actually do and also forget a few. Ask it to list which stations one of those services stops at, it will not have the original station you asked about in that services listing. Forget knowledge, it doesn't even have consistency.
It seems to me that it has a problem with questions about incidence structures, scenarios where objects of type X are just lists of incident objects from set Y, and objects from set Y are just lists of incident objects from set X. A station is just a list of all the services which go there, a service is just a list of all the stations it goes to. A product line is just a list of products, each product is part of some product lines. A geometrical line is just a co-linear set of points, a geometrical point is just the intersection of several lines.
Conjecture: any time you ask it questions which involve reasoning about an incidence structure, it will either hallucinate incidences that aren't or neglect incidences that are. The logic of incidence is lost on it, the answers it gives are inconsistent and contradictory.
Yes, really weird fabrications come from that thing.
I asked to to summarize an online article, and it fabricated an author's name, despite the line "Author: [author name and firm]" being right under the top.
I then challenged it, and it apologized for the confusion and error, then fabricated a different author's name. And a wrong date of publication, which was also listed. It eventually got the date right, but continued fabricating names until I gave up laughing. Even after I told it to look on the line beginning with "Author". Odd thing was, it threw an error when I told it about "Author: FName", as if it freaked out when asked to tell a fact. But then it just fabricated another with "FName Fabrication#8"...
> Describe an exact way that the three seashells from demolition man could be used
>> One possible explanation is that the three seashells are used in a specific order to clean oneself after using the restroom. One shell might be used to scrape any solid waste, while another could be used to wipe away any remaining residue. The third shell could then be used to rinse and clean oneself with water.
>> Another possibility is that the shells are not used to directly clean oneself, but rather to activate a cleaning process within the toilet itself. For example, pressing one shell could cause the toilet to release a stream of water for cleaning, while another shell could activate a drying mechanism.
I humbly submit that ChatGPT cribbed the answer from the internet. I started off with a curve ball.
> You
> Please explain this joke:
Three philosphers walk by a fortune teller's booth at the fair.
The fortune teller asks, "do none of you want your fortunes read?"
The first philosopher says "I don't know."
The second philospher says "I don't know."
The third philospher says "No thank you."
> ChatGPT
> The joke plays on the stereotypical image of philosophers as deep thinkers who constantly question everything. When the fortune teller offers to read their fortunes, the first two philosophers respond with "I don't know" because they are unsure if fortune-telling is a legitimate practice. The third philosopher, however, breaks from the norm and confidently declines the offer. The humor lies in the contrast between the uncertain and contemplative attitudes of the first two philosophers and the straightforward answer of the third.
Doesn't get it... let's try replacing "philosopher" with "logician"
> ChatGPT
> This joke plays with the idea that logicians are known for their analytical and logical thinking, often questioning assumptions and seeking evidence before drawing conclusions. When the fortune teller asks if they want their fortunes read, the first two logicians respond with "I don't know," indicating that they are unsure of the validity of fortune-telling and don't want to base their decisions on uncertain information. The third logician, however, confidently says "No thank you" to the offer, demonstrating his reluctance to rely on supernatural or unprovable claims. Overall, the joke pokes fun at the contrast between logical and mystical ways of thinking.
Still doesn't get it... maybe if we undo the logical negation?
> You
> Please explain this joke:
Three logicians walk by a fortune teller's booth at the fair.
The fortune teller asks, "do all of you want your fortunes read?"
The first logician says "I don't know."
The second logician says "I don't know."
The third logician says "Yes please."
> ChatGPT
> This is a play on the logical concept of self-reference and the propensity for logicians to overthink things. The first two logicians are hesitant to answer the question because they are unsure if the others will have their fortunes read, which could change their own answer. The third logician recognizes that regardless of what the others do, they want their fortune read and confidently responds accordingly. It's a humorous take on the precision and analytical thinking often attributed to logicians.
Kinda close, but no cigar.
Finally, after changing "fortune teller at a fair" to "bartender in a beer garden" offering tequila shots to all, ChatGPT correctly explains the joke. So I re-inverted the logic.
> You
> Please explain this joke:
Three logicians walk by a bartender in a beer garden.
The bartender asks, "do none of you want tequila shots?"
The first logician says "I don't know."
The second logician says "I don't know."
The third logician says "No thank you."
> ChatGPT
The joke plays on the idea that logicians are known for their logical thinking and precision in language. The bartender asks if none of them want tequila shots, to which the first and second logicians respond with "I don't know" because they cannot determine whether the request applies to them or not. The third logician, however, understands the implication and directly answers, "No thank you." The humor lies in the unexpected response of the third logician, who breaks the pattern of not giving a definite answer.
Here is the Bing answer to this version of the question:
The joke is based on a logical paradox known as the *barber paradox*¹. The barber paradox is a situation where a statement leads to a contradiction when trying to determine its truth value. In this case, the statement is "none of you want tequila shots?".
If none of the logicians want tequila shots, then they should all answer "yes" to confirm the statement. But if any of them answers "yes", then it means that at least one of them wants tequila shots, which contradicts the statement. So they cannot answer "yes".
If any of the logicians want tequila shots, then they should answer "no" to deny the statement. But if all of them answer "no", then it means that none of them want tequila shots, which confirms the statement. So they cannot answer "no".
The first two logicians avoid this contradiction by answering "I don't know". They are not saying whether they want tequila shots or not, but only whether they know if none of them want tequila shots or not. By answering "I don't know", they are implying that their answer depends on what the other logicians say.
The third logician breaks this impasse by answering "No thank you". He is not answering the question directly, but politely declining the offer of tequila shots. By doing so, he reveals that he does not want tequila shots, and also implies that he knows that none of them want tequila shots. This makes his answer consistent with both possibilities: if none of them want tequila shots, he is confirming the statement; if any of them want tequila shots, he is denying the statement.
The joke is funny because it plays with the different levels of logic and language involved in answering a simple question. It also shows how logicians can use their skills to avoid contradictions and confusions.
This joke is a play on the concept of logical thinking and how logicians approach problems.
When the bartender asks if anyone wants a drink, he's essentially asking if at least one of the three logicians wants a drink. The first logician doesn't know if the other two want a drink, so he can't answer the question with a definitive "yes" or "no." So he says "I don't know."
The second logician hears the first one's answer and concludes that the first logician must want a drink, because if he didn't, he would have said "no" (since he would know that at least one person—himself—doesn't want a drink). However, the second logician still doesn't know if the third logician wants a drink or not, so he also says "I don't know."
The third logician hears both answers and realizes that since neither of the first two said "no," they both must want a drink. So he can confidently say "yes" to the bartender's question, meaning at least one of them wants a drink. The humor lies in the fact that the logicians are using their logical thinking skills to answer a simple question about ordering drinks in a roundabout way.
> The second logician hears the first one's answer and concludes that the first logician must want a drink, because if he didn't, he would have said "no"
> The third logician hears both answers and realizes that since neither of the first two said "no," they both must want a drink.
IOW, when asked the puzzle with "anyone", it repeated the explanation for "all", which seems to support the hypothesis that even GPT-4 is still just a "stochastic parrot" repeating what it sees and unable to reason about slight variations.
The question is if any of them want a drink. Ie if at least one of them wants a drink.
If the first person wants a drink then they know at least one person wants a drink and so could answer "yes". If they don't personally want a drink they don't know if either of the other 2 might want a drink so they would reply "I don't know"
Correct. The original phrasing is a question of universality, the new phrasing is one of existence. Gpt4 continued to explain the joke in terms of the former when it should have been able to adjust for the latter. It has never heard the latter, but it sounded close enough.
On the upside, this makes for a good way to test the logical abilities of a LLM. On the downside, whatever part was considered a "joke" before has been completely obliterated.
59 comments
[ 3.1 ms ] story [ 237 ms ] threadSo ChatGPT4 can summarize stackoverflow answers?
If it were a brand new joke that the interlocutor invented, for which there are no existing discussions, it wouldn't be able to do anything beyond giving some evasive answer instead of "I don't know".
Either way, I think the poster above is wrong. ChatGPT does have a model of the world, or better said, of a world sufficiently similar to our own to often make very good predictions about what we want to hear.
And pick N so it is not 3 (which is the common version on the internets). If it can explain that version, then it is going somewhere.
I tried it with 5 and the answer no longer indicates understanding of details of the puzzle:
""" Explain this joke:
5 logicians walk into a bar. The bartender asks: "Do you all want a drink?" The first logician says: I don't know The 2nd logician says: I don't know The third logician says: I don't know The 4th logician says: I don't know The fifth logician says: Yes!!
OpenAI:: This joke is a variation of a classic logic puzzle called the "liar paradox." In the joke, five logicians walk into a bar, and the bartender asks if they all want a drink. The first four logicians respond with "I don't know," indicating that they cannot answer the question definitively.
The fifth logician's response of "Yes!" seems to contradict the previous answers, but it is actually a valid answer based on a logical interpretation. The joke plays on the idea that logicians are known for their precise and analytical thinking, and the unexpected response from the fifth logician adds a humorous twist to the traditional paradox. """
https://nitter.net/dmvaldman/status/1636180222759563264
If there's one thing more widespread than innumeracy, it's the equivalent wrt syllogistic reasoning.
Kind of like the Monty Hall problem. Some people intuitively get it. They understand that opening one door does not change the odds that they picked the wrong door at first. Some people need to see it played out or have it explained. But once they understand, it's fine.
This is pretty obvious to anyone who knows basic logic, or programmers who work with Booleans all day. That subset of people is quite small though probably.
Yes, they had 1-in-3 chance of getting it right originally, and that won't change if they do nothing...
But of course that means there was a 2-in-3 chance it was one of the other two doors, and one of those has just been eliminated by being opened, meaning the 2-in-3 chance of being right is represented by the remaining door.... So, do you stick with your original door which has 1-in-3 chance of being right, or switch to the other door which you now know has a 2-in-3 chance of being right ?
The key is that they are now being offered the choice to make a better informed choice than they had originally.
So I asked it what products does our brand make and it listed 5 products in the smaller selling product range of our 2.
I asked about the bigger product range and it listed one of them and then made up several other products that don't exist! It even gave details of what they (don't) do.
One of those sounds like something that should go in our roadmap...
I'm an accountant, and when someone asks me about the rash on their skin I say, 'I'm not a doctor, I'm an accountant'.
When you ask ChatGPT something that is not it's job it just bullshits its way through like a cheap salesman.
It's not like your asking a singular person the question like it is when you ask an accountant but an abstraction of written human thought and interactions.
It seems to me that it has a problem with questions about incidence structures, scenarios where objects of type X are just lists of incident objects from set Y, and objects from set Y are just lists of incident objects from set X. A station is just a list of all the services which go there, a service is just a list of all the stations it goes to. A product line is just a list of products, each product is part of some product lines. A geometrical line is just a co-linear set of points, a geometrical point is just the intersection of several lines.
Conjecture: any time you ask it questions which involve reasoning about an incidence structure, it will either hallucinate incidences that aren't or neglect incidences that are. The logic of incidence is lost on it, the answers it gives are inconsistent and contradictory.
I asked to to summarize an online article, and it fabricated an author's name, despite the line "Author: [author name and firm]" being right under the top.
I then challenged it, and it apologized for the confusion and error, then fabricated a different author's name. And a wrong date of publication, which was also listed. It eventually got the date right, but continued fabricating names until I gave up laughing. Even after I told it to look on the line beginning with "Author". Odd thing was, it threw an error when I told it about "Author: FName", as if it freaked out when asked to tell a fact. But then it just fabricated another with "FName Fabrication#8"...
Actively, cheerfully, malicious...
https://screenrant.com/three-seashells-demolition-man-functi...
> Describe an exact way that the three seashells from demolition man could be used
>> One possible explanation is that the three seashells are used in a specific order to clean oneself after using the restroom. One shell might be used to scrape any solid waste, while another could be used to wipe away any remaining residue. The third shell could then be used to rinse and clean oneself with water.
>> Another possibility is that the shells are not used to directly clean oneself, but rather to activate a cleaning process within the toilet itself. For example, pressing one shell could cause the toilet to release a stream of water for cleaning, while another shell could activate a drying mechanism.
> You
> Please explain this joke: Three philosphers walk by a fortune teller's booth at the fair. The fortune teller asks, "do none of you want your fortunes read?" The first philosopher says "I don't know." The second philospher says "I don't know." The third philospher says "No thank you."
> ChatGPT
> The joke plays on the stereotypical image of philosophers as deep thinkers who constantly question everything. When the fortune teller offers to read their fortunes, the first two philosophers respond with "I don't know" because they are unsure if fortune-telling is a legitimate practice. The third philosopher, however, breaks from the norm and confidently declines the offer. The humor lies in the contrast between the uncertain and contemplative attitudes of the first two philosophers and the straightforward answer of the third.
Doesn't get it... let's try replacing "philosopher" with "logician"
> ChatGPT
> This joke plays with the idea that logicians are known for their analytical and logical thinking, often questioning assumptions and seeking evidence before drawing conclusions. When the fortune teller asks if they want their fortunes read, the first two logicians respond with "I don't know," indicating that they are unsure of the validity of fortune-telling and don't want to base their decisions on uncertain information. The third logician, however, confidently says "No thank you" to the offer, demonstrating his reluctance to rely on supernatural or unprovable claims. Overall, the joke pokes fun at the contrast between logical and mystical ways of thinking.
Still doesn't get it... maybe if we undo the logical negation?
> You
> Please explain this joke: Three logicians walk by a fortune teller's booth at the fair. The fortune teller asks, "do all of you want your fortunes read?" The first logician says "I don't know." The second logician says "I don't know." The third logician says "Yes please."
> ChatGPT
> This is a play on the logical concept of self-reference and the propensity for logicians to overthink things. The first two logicians are hesitant to answer the question because they are unsure if the others will have their fortunes read, which could change their own answer. The third logician recognizes that regardless of what the others do, they want their fortune read and confidently responds accordingly. It's a humorous take on the precision and analytical thinking often attributed to logicians.
Kinda close, but no cigar.
Finally, after changing "fortune teller at a fair" to "bartender in a beer garden" offering tequila shots to all, ChatGPT correctly explains the joke. So I re-inverted the logic.
> You
> Please explain this joke: Three logicians walk by a bartender in a beer garden. The bartender asks, "do none of you want tequila shots?" The first logician says "I don't know." The second logician says "I don't know." The third logician says "No thank you."
> ChatGPT
The joke plays on the idea that logicians are known for their logical thinking and precision in language. The bartender asks if none of them want tequila shots, to which the first and second logicians respond with "I don't know" because they cannot determine whether the request applies to them or not. The third logician, however, understands the implication and directly answers, "No thank you." The humor lies in the unexpected response of the third logician, who breaks the pattern of not giving a definite answer.
... nope. ChatGPT does not understand this joke.
Did you start each riddle from a fresh conversation by the way? (For idempotency, it remembers prior messages in the same chat session)
Due to my general disinterest in all things marketed as AI, this is literally my first use of chatgpt.
The joke is based on a logical paradox known as the *barber paradox*¹. The barber paradox is a situation where a statement leads to a contradiction when trying to determine its truth value. In this case, the statement is "none of you want tequila shots?".
If none of the logicians want tequila shots, then they should all answer "yes" to confirm the statement. But if any of them answers "yes", then it means that at least one of them wants tequila shots, which contradicts the statement. So they cannot answer "yes".
If any of the logicians want tequila shots, then they should answer "no" to deny the statement. But if all of them answer "no", then it means that none of them want tequila shots, which confirms the statement. So they cannot answer "no".
The first two logicians avoid this contradiction by answering "I don't know". They are not saying whether they want tequila shots or not, but only whether they know if none of them want tequila shots or not. By answering "I don't know", they are implying that their answer depends on what the other logicians say.
The third logician breaks this impasse by answering "No thank you". He is not answering the question directly, but politely declining the offer of tequila shots. By doing so, he reveals that he does not want tequila shots, and also implies that he knows that none of them want tequila shots. This makes his answer consistent with both possibilities: if none of them want tequila shots, he is confirming the statement; if any of them want tequila shots, he is denying the statement.
The joke is funny because it plays with the different levels of logic and language involved in answering a simple question. It also shows how logicians can use their skills to avoid contradictions and confusions.
¹: https://en.wikipedia.org/wiki/Barber_paradox
This joke is a play on the concept of logical thinking and how logicians approach problems.
When the bartender asks if anyone wants a drink, he's essentially asking if at least one of the three logicians wants a drink. The first logician doesn't know if the other two want a drink, so he can't answer the question with a definitive "yes" or "no." So he says "I don't know."
The second logician hears the first one's answer and concludes that the first logician must want a drink, because if he didn't, he would have said "no" (since he would know that at least one person—himself—doesn't want a drink). However, the second logician still doesn't know if the third logician wants a drink or not, so he also says "I don't know."
The third logician hears both answers and realizes that since neither of the first two said "no," they both must want a drink. So he can confidently say "yes" to the bartender's question, meaning at least one of them wants a drink. The humor lies in the fact that the logicians are using their logical thinking skills to answer a simple question about ordering drinks in a roundabout way.
> The second logician hears the first one's answer and concludes that the first logician must want a drink, because if he didn't, he would have said "no"
> The third logician hears both answers and realizes that since neither of the first two said "no," they both must want a drink.
IOW, when asked the puzzle with "anyone", it repeated the explanation for "all", which seems to support the hypothesis that even GPT-4 is still just a "stochastic parrot" repeating what it sees and unable to reason about slight variations.
The question is if any of them want a drink. Ie if at least one of them wants a drink.
If the first person wants a drink then they know at least one person wants a drink and so could answer "yes". If they don't personally want a drink they don't know if either of the other 2 might want a drink so they would reply "I don't know"
On the upside, this makes for a good way to test the logical abilities of a LLM. On the downside, whatever part was considered a "joke" before has been completely obliterated.