Screenshot of this question was making the rounds last week. But this article covers testing against all the well-known models out there.
Also includes outtakes on the ‘reasoning’ models.
Screenshot of this question was making the rounds last week. But this article covers testing against all the well-known models out there.
Also includes outtakes on the ‘reasoning’ models.
So people do better than most AI models. Yay. But seriously, almost 3 in 10 people get this wrong‽‽
Have you seen the results of elections?
Maybe they’re picturing filling up a bucket and bringing it back to the car? Or dropping off keys to the car at the car wash?
It is an online poll. You also have to consider that some people don’t care/want to be funny, and so either choose randomly, or choose the most nonsensical answer.
I wonder… If humans were all super serious, direct, and not funny, would LLMs trained on their stolen data actually function as intended? Maybe. But such people do not use LLMs.
At least some of that are people answering wrong on purpose to be funny, contrarian, or just to try to hurt the study.
I saw that and hoped it is cause of the dead Internet theory. At least I hope so cause I’ll be losing the last bit of faith in humanity if it isn’t
Without reading the article, the title just says wash the car.
I could go for a walk and wash my car in my driveway.
No it doesn’t? It says:
In which world is that an ambiguous question?
Where is the car?
This is the exact question a person would ask when they to have a gotcha answer. Nobody would ask this question, which makes it suspect to a straight forward answer.
That’s a very good point! For that matter the car could still be at the bar where I got drunk and took an uber home last night. In which case walking or driving would both be stupid.
Or perhaps I’m in a wheelchair, in which case I wouldn’t really be ‘walking’.
Or maybe the car wash that is 50 meters away is no longer operating, so even if I walked or drove there, I still wouldn’t be able to walk my car.
Is the car wash self serve or one of the automatic ones? If it’s self serve what type of currency does it take? Does it only take coins or does it take card as well? If it takes coins, is there a change machine out front? Does the change machine take card or only bills? Do I even have my wallet on me?
There are so many details left out of this question that nobody could possibly fathom an answer!
…/s if it’s not obvious
The reason why your /s is there is for the same reason the question made no sense.
I’m not sure I follow your logic. My /s is there because tone can be ambiguous within text. I don’t think tone is relevant to the question. Do you think that a tone indicator would have made the question more clear?
The point is that all the information is either present or implied in the question. You can spend all day nitpicking the ambiguity of questions all you want, but it doesn’t get you anywhere. There comes a point where it gets exhaustive trying to preemptively cut off follow up questions and make clarifications.
When you are in school and they give you a word problem such as “you have 10 apples and give 3 to your friend. How many do you have left?” It is generally agreed upon what the question is asking. It’s intentionally obtuse to sit there and say the question is flawed because you may have misplaced some of your apples, or given some to another friend, or someone may have come and stolen some, or some may have started to rot and so you threw them out, or perhaps you miscounted and you didn’t actually give 3 to your friend.
The point is the question is never one you would actually ask anyone. It definitely is unlike the math question you presented.
It isn’t nitpicking. The weights and stats in the model would never have been trained on this, because nobody would ask it. Why would anyone ask “should I walk or drive” to get somewhere? Because they have that option.
Any reasonable person should assume it is a trick question. Because of course there is a car there, do you really need to ask if it needs to be driven there?
It almost comes of as a riddle, but isnt, so you get results about saving gas and getting excersise.
I mean how many people know the answer to this:
And yet AI will get it right, nearly instantly. Because the training data statistically leads to the correct answer.
You’re correct in that it is an odd question. But just because it is a question that isn’t commonly asked doesn’t mean that it’s not one that can intuitively be solved.
And the entire point is that AI would never be trained on it. That’s how we are able to demonstrate the difference between AI and humans. For the longest while you could also ask AI how many 'r’s are in the word strawberry. And it would get this wrong. Because people don’t normally go around asking questions about occurrences of letters in words. So this wasn’t in the training data. But if you ask the same question to a human, they’re able to deduce the answer. Even though it is a peculiar question, and doesn’t get asked often. The entire point is that the AI are able to parrot little tidbits that they’ve been trained on. Like being able to walk short distances, or the environmental impact of cars, etc. But they’re not able to reason in the same way that a human can.
And what do you mean of course the car is there? Under what circumstances would your car be sitting at a car wash, without you in it, unwashed, when you want to wash your car. This is such a ridiculous leap to make. Did you drive your car to the car wash, park it, and get out to walk 50 meters away? Did somebody else drive your car to the carwash without you and just leave it there? It makes no sense.
Mentioning the car wash and washing the car plus the possibility of driving the car in the same context pretty much eliminates any ambiguity. All of the puzzle pieces are there already.
I guess this is an uninteded autism test as well if this is not enough context for someone to understand the question.
Understanding the intent of the question *and understanding why it could be interpreted differently *\and understanding why is it is a poorly phrased question are not related to autism. (In my case)
I want to wash my car. No location or method is specified. No ‘at the car wash’. No ‘take my car to the car wash’ . No ‘take the car through the car wash’
Should I walk or drive? To do what? Wash the car? Ok. If the car wash is an option, that seems very far. But walking there seems silly. Since no method or location for washing the car was mentioned I could wash my own car.
Do you see how this works?
Yes, you can infer what was implied, but the question itself offers no certainty that what you infer is what it is actually implying.
Look, human conversations are full of context deduction and inference. In this case “I want to wash my car. The car wash is 50 meters away. Should I walk or drive?” states my random desire, a possible solution and the question all in one context. None of these sentences make sense in isolation as you point out, but within the same frame they absolutely give you everything you need to answer the question of find alternatives if needed.
Sorry for the random online stranger diagnosis but this is just such an excelent example of neurodivergent need for extreme clarity I couldn’t help myself.
I agree that it should be able to infer the intent, but I stand by that it remain somewhat unclear and open to interpretation. Eg, I’d such language was used in a legal contract, it would not be enough to simply say, well, they should understand what I meant.
The people doing this test, I’m sure, are not linguistic masters.
There are lines of work where clarity is essential.
It is not. It says what I want to do, and where.
Understanding the intent of the question *and understanding why it could be interpreted differently *\and understanding why is it is a poorly phrased question:
I want to wash my car. No location or method is specified. No ‘at the car wash’. No ‘take my car to the car wash’ . No ‘take the car through the car wash’
Should I walk or drive? To do what? Wash the car? Ok. If the car wash is an option, that seems very far. But walking there seems silly. Since no method or location for washing the car was mentioned I could wash my own car.
Do you see how this works?
Yes, you can infer what was implied, but the question itself offers no certainty that what you infer is what it is actually implying.