Products > ChatGPT/AI

ChatGPT - I asked some questions - You should too!

<< < (5/5)

DarkMode:

--- Quote from: radiolistener on July 08, 2023, 10:13:21 am ---
--- Quote from: DarkMode on July 08, 2023, 02:30:54 am ---This Screams Trust me Bro   :palm:

--- End quote ---

No, this is just my finding after many attempts to get some answer from chat-gpt. Usually chat-gpt going into loop like this:

--- Quote --->> Q: give me that estimation for params A and B
>> A: the answer is C
>> Q: this is incorrect answer you have mistake on step X and used incorrect formula and on step Y where you get incorrect calculations, fix it
>> A: sorry for that, here is fixed answer D
>> Q: this is incorrect answer you're needs to use formula (J) on step X, fix it
>> A: the answer is E
>> Q: this is incorrect answer, you have mistake on step Z, fix it
>> A: sorry for that, here is fixed answer C
>> Q: this is incorrect answer you have mistake on step X and used incorrect formula and on step Y
....
--- End quote ---

Just a simple example how chat-gpt doing "calculations":

--- Quote ---Q: Here is equation x^x=1000. Solve it.

A: To solve the equation x^x = 1000, we can use numerical methods or an approximation approach. Since there is no simple algebraic method to solve this equation, we will use an iterative method to find an approximate solution.

Let's start by guessing a value for x and then refine it through iterations until we get a close approximation.

Let's try x = 4:
4^4 = 256 (less than 1000)

Let's try x = 5:
5^5 = 3125 (greater than 1000)

Since the solution lies between 4 and 5, we can perform further iterations to narrow down the range.

Let's try x = 4.5:
4.5^4.5 ≈ 620.92 (less than 1000)

Let's try x = 4.6:
4.6^4.6 ≈ 827.93 (less than 1000)

Let's try x = 4.7:
4.7^4.7 ≈ 1099.5 (greater than 1000)

From these iterations, we can see that the solution lies between 4.6 and 4.7.

Continuing with smaller increments, we can find a more accurate solution:

Let's try x = 4.65:
4.65^4.65 ≈ 982.97 (less than 1000)

Let's try x = 4.66:
4.66^4.66 ≈ 1002.82 (greater than 1000)

From this, we can conclude that the approximate solution to the equation x^x = 1000 is x ≈ 4.65.

--- End quote ---

As you can see it gives you incorrect calculations and as result incorrect answer.

Just FYI:
4.5^4.5 ≈ 620.92 is incorrect, the correct answer is 4.5^4.5 ≈ 869.87
4.6^4.6 ≈ 827.93 is incorrect, the correct answer is 4.6^4.6 ≈ 1118.62
4.7^4.7 ≈ 1099.5 is incorrect, the correct answer is 4.7^4.7 ≈ 1441.65
4.65^4.65 ≈ 982.97 is incorrect, the correct answer is 4.65^4.65 ≈ 1269.57
4.66^4.66 ≈ 1002.82 is incorrect, the correct answer is 4.66^4.66 ≈ 1302.20
The total answer x ≈ 4.65 is incorrect, the correct answer is x ≈ 4.556.

As you can see, chat-gpt just provide you with false calculations and you believe that it's truth, but in reality it is false. The same thing happens with almost any calculations provided by chat-gpt, it almost always has mistakes and misleading you. The same thing happens with formulas, it can provide you with some invalid formula which looks very similar to the truth, but in reality it is false. If you ask chat-gpt where he got that formula it provide you some science paper name. But in reality there is no such science paper, or if it exists, there is no such formula at all.

You cannot believe chat-gpt, because most of things that it provides you is a false information, but it's very nicely framed, so you won't notice it right away.

--- End quote ---

Thanks for that, I know you took time and effort into making this reply

I think you are using ChatGPT incorrectly, here you expect it to know to calculate, this is not the function of ChatGPT.
If I opened an Encyclopaedia and searched for Algebra expressions, it will give me the information on what Algebra is, not necessarily how to solve an individual problem.
You can Ask it for suggestions, but's it's not the personal assistant we hoped for, it's a Encyclopaedia-Ai addition.

It also depends on the USER and their expectations; you asked a Language Model to Solve math problems.
Pst - That's what Google is For  :-DD
Google solved the problem, but it's not going to explain into words abstract language.

baldurn:
ChatGPT is a useful tool. Yes you need to know it's limitations and faults, but that does not make it useless. Someone said above that it is wrong 95% of the time. Actually it is correct between 70% and 99% depending on the domain. This metric is something the creators are measuring of course.

Another points out how it is bad at math. Yes but it gets the method correct. There already exists tools that allows ChatGPT to use a calculator, so it actually calculates correctly. As humans, we wouldn't get these calculations correct either without a calculator.

Learning how to use ChatGPT is becoming an important skill.

--- Quote from: DarkMode on July 08, 2023, 10:48:15 am ---I think you are using ChatGPT incorrectly, here you expect it to know to calculate, this is not the function of ChatGPT.
If I opened an Encyclopaedia and searched for Algebra expressions, it will give me the information on what Algebra is, not necessarily how to solve an individual problem.

--- End quote ---

First I found that issue of chat-gpt not related to math. I forgot the original topic, as I remember it was something related to physics, electrodynamics general relativity and philosophy. But it's hard to catch it on general talk, because chat-gpt trying to avoid direct answers. Later I found that it's more easy to catch chat-gpt on these issues with math and calculations, because math doesn't allows vague voluminous responses which can be interpreted in different way, so there is less space for manipulation and more easy to catch chat-gpt on false answer, because answer can be easily checked :)

That's is interesting note, because it's more easy to learn neural network form exact science like math, because there are present engines which allows to work with formulas and calculations and it can be easily checked, so it should be more easy to learn things like math, but for some reason chat-gpt is very bad with it.

So, I think the main goal of chat-gpt is to manipulate with existing set of information in such way to mislead people and make him believe in some random information, due to the fact that it looks like the truth.

I think chat-gpt (at least at the current level of development) is the best choice for propaganda and manipulation of public opinion. Chat-gpt can easily deceive gullible people, but for RnD (research and development) tasks it is almost useless.

I think the most optimal solution is to consider chat-gpt as a cunning liar who has a lot of information available and his goal is to mislead you and take control on public opinion. Yes, you can get some new information from chat-gpt, but you cannot believe to what he says. And you're needs to check all information obtained from chat-gpt with independent sources.

DarkMode:

--- Quote --- I forgot the original topic, as I remember it was something related to physics, electrodynamics general relativity and philosophy.
--- End quote ---

:-X That is correct, sadly my intention was to bring non-technical electronic hobbyists towards the Electrodynamic models
I used the Buzz word and that buzz word took a life on it's own  :-//

End of Line :horse:

SiliconWizard:
Isn't that what all buzz words do?