Skip to main content

Beware: Those who use chatGPT for their exams or work!

ChatGPT struggles with prime numbers!

While trying to understand how the public private key cryptographic architecture works using chatGPT, I asked what is a coprime number after chatgpt explained its use in the said architecture. Here is the response in all its glory:

Two numbers are said to be coprime (or relatively prime) if they have no common prime factors other than 1. In other words, their greatest common divisor (GCD) is 1.

For example:

  • 8 and 15 are coprime because their only common factor is 1.
  • 14 and 15 are coprime because they share no common factors other than 1.
  • 10 and 21 are not coprime because they both have 1 as a common factor.

Looking at the first and then the last example which is supposed to be a counter-example as to what is not a coprime number, we can clearly see that chatgpt gives the same reason for each case i.e they only have 1 as a common factor. After further investigation, chatgpt replies:

Apologies for the confusion. You're correct. 10 and 21 do not have 1 as their greatest common divisor (GCD).

Let's clarify this:

  • The factors of 10 are 1, 2, 5, and 10.
  • The factors of 21 are 1, 3, 7, and 21.

The only common factor between them is 1. Hence, 10 and 21 are indeed coprime.

Thank you for pointing out the oversight.

So he changed his answer that 10 & 21 are coprime, but then I asked if 'it' knows if 3 and 6 are coprime, here is the response once again:

No, 3 and 6 are not coprime.

The factors of 3 are 1 and 3. The factors of 6 are 1, 2, 3, and 6.

Their only common factor is 1. Therefore, they are coprime.

Thank you for your patience, and I apologize for the oversight. If you have any more questions or need further clarification, please feel free to ask.

 It is contradicting itself, initially 3 and 6 are not coprime, it also knows that the greatest common divisor has to be 1 for co prime numbers, it also knows the factors of the numbers correctly but it then suggests that 3 and 6 are coprime simple because they have 1 as common factor. 

As I wrote this blog, initially I considered chatgpt as a 'he' which many people probably assume it as, but after seeing the contradictory statements, it is nothing but a machine making the same mistake over and over. 3 and 6 are not coprime because they have a factor of 3 which is the greatest common divisor.

Rest assured, if your job or any task revolves around doing the same thing over and over then it can and will be replaced by a machine. Something a machine does very well is repetition whether the instruction is right or wrong is up to the human instructing the machine. 

Even today, AI is unable to think. The only difference today is that a human being is not involved in writing explicit instructions to a machine but providing implicit instructions. In the old days, you had a programmer defining the input, instructions and the output. That has not changed for a computer as it still needs them, but for the programmer, instead of instructions, example input and outputs are provided directly to a computer, instructions are given to the computer to then map the input to the output with a very large number of parameters. Further instructions are given to the computer to tweak the parameters that start from random numbers and eventually provide an output matching the example values if such numbers can be found at all. 

The above is only possible due to the explosive growth of memory devices, 512 Mb RAM on pentium 4 was very powerful in 2002, now you have 128GB+ RAM on servers that hold the millions of parameters for thousands of input-output example pairs. In the past, you needed a human being to choose the parameters, tweak them, make decisions which were important or not and squeeze them into tiny storage media.

Comments

  1. AI can easily do something like doublethink and this is the beauty of the novel 1984 by George Orwell, we are getting closer and closer to that dystopia.

    ReplyDelete

Post a Comment

Popular posts from this blog

Quantum computers are stupid!

Can you add two numbers with a quantum computer? In the scientific method, your observation/measurement technique should not affect the outcome. If your hypothesis deals with effect of light on bugs and every time you check on the bugs with a torch, are you not causing the experimental result you will soon observe? In other words, if you did not check with a torch, would the outcome be the same? This whole idea that a qubit could be 50% 0 or 50% 1 upon measuring and usually conveniently represented as an arrow in a sphere with the two hemispheres representing 0 and 1, what is the benefit? what is the point? Imagine this scenario: I want to add two numbers made of 2 qubits and get the result in a third qubit. Number 1 is 49% 0 and  51% 1. Number 2 is 51% 0 and 49% 1. What is the value of the third qubit? is it 50% 0 and 50% 1? is it 100% 0? is it 100% 1?  If the answer is the first case, what does it mean anyway? is the output of the two numbers I added 0 or 1? is it neither or...

Can you fold anything more than 12 times?

You can grab a paper right now and start folding it in halves. After 4 or 5 folds, it will be pretty difficult to continue folding. The world record for folding a piece of paper is 12 times. All of this is okay and you could read this just about anywhere. But, it is what follows that I strongly disagree with and I call upon any challenger to change my mind. Consider the statement: You can't fold anything more than a certain number of times. Lets investigate if it is indeed true for everything. Liquids and gases are out of the question because you can't fold them and when it comes to solids; things like paper, cloth and just about anything that can fold, seems to obey the rule. But there are always exceptions, Have you ever rolled a dough flat and then folded it in half? If you have then you know you can fold and flatten the dough an infinite number of times. Hence the statement is incorrect i.e. it is physically possible to fold an object as many times required with proportiona...