2+2=5

Recently, I came across a program called “CalcGPT”
The intent of the program is to highlight the absurdities of trusting AI for everything, even something as simple as math equations. As the name suggests, it is a calculator running off a GPT model.

However.

This AI only hallucinates. For example:

It can even handle text inputs, leading to results such as:

The developer, Calvin Liang, designed it in such a way that it would not return an accurate output.
A professor I know said, (paraphrased) “I know AI will never outpace humans, since computers are only as smart as the people who program them.”
I think he’s correct. But it’s not so much people who do not know things that are the danger.
It’s those who are manipulating AI to deceive.
That’s the real danger I think.

If we rely on AI more than the articles it claims, then it is possible that the AI would say that someone said something that he never said. Which is probably the least of our worries.

To sort of return to the initial topic, math and calculations, what if an AI returns a calculation for a shuttle launch, but the developer had an agenda against the shuttle company? The consequences of that could be world-changing.

Basically, check your work. Check AI’s work. Read the sources. AI can be helpful, but it is not infallible.