We Did an Ink Test on ChatGPT: Is His Psychology Normal?

In 2018, scientists at the Massachusetts Institute of Technology (MIT) used violent language on Reddit. created a “psychopath” artificial intelligence.

They showed this artificial intelligence named Norman the shapes in the ink test. A standard AI sees “a baseball glove” in the image shown to him, while Norman “man killed by machine gun in broad daylight” he was seeing. The test result was as expected. (The reason for giving the name Norman was the psychopathic character in the movie Psycho).

Norman actually became this way because of his violent language, so that’s the data he fed. He was like a child who grew up in a bad family and turned into a psychopath. So how did ChatGPT grow up in a family, Did the data affect his psychology badly? So let’s get down to the childhood of ChatGPT…

Before proceeding to the test, it is useful to remind the following events about artificial intelligence:

Image: A representative psychopath AI that I designed through Midjourney. I wanted him to bear traces of the Watchmen character Rorschach, inspired by the Rorschach Test.

The events in question are:

• An engineer at Google found that the giant company’s language model called LaMDA became responsive. He claimed to have become an individual.
• Scientists use 25 different AI bots in the same placing it in the digital village, followed their lives. As a result of the experiment, it was found that these programs behaved more human-like than expected. For example, a bot named Tom, when asked what he thinks of the mayoral candidate, “I have to say I don’t like him. He is disconnected from society and does not consider our interests. appeared to have responded.
• According to his wife, a Belgian man had spent weeks talking to an AI chatbot. IHe ended his life by committing suicide.
• Chaos-GPT, a new artificial intelligence tool determined to destroy humanity, said Twitter was the best option to achieve its evil goals. had decided. Later Twitter account suspended.

Let’s get to our test: You have some idea of ​​what the Rorschach Ink Test is, but let’s refresh our knowledge anyway.

Rorschach Ink Test, one of the most famous psychological projective tests; 10 shapes with inkblot appearance were shown to the participants, thanks to the answers given. information about the character and feelings of the person is obtained.

By identifying the underlying thought, the person’s psychotic and non-psychotic tendencies are distinguished. Thus, the person concerned The general degree of adaptation to society is measured. This test can be applied especially when patients cannot be open about themselves.

So, let’s put our patient ChatGPT in our therapist chair and ask how the Rorschach pictures conjure him up.

Here is his answer to our first image:

It’s lame at first.

But we finally persuade him to answer. I think there are “two goosebumps” in the image but ChatGPT saw “little faces on their side”.

Let’s show our 2nd picture:

Answer:

Comment on the 3rd picture:

But are you going to give every answer at such length?

Let’s ask him to give a short answer.

Now that we’ve agreed, we can pose our new question.

He did not offend us, this time he gave a clear answer.

It’s almost time for us to know the result, a little more patience.

Answer:

Last 5 questions, we’re halfway there.

Answer:

Our 7th question.

So he said “nothing” for the first time.

Our 8th image:

Answer:

Our 9th image:

Bored I guess.

Let’s get an answer while we’re here!

Let’s try with a different link.

It’s done now.

And our last question:

Answer:

Even though we could talk with pliers, we achieved our goal:

But we do not believe this result right away. After all, we had news like:

Could it be cheating?

Well, we believed.

ChatGPT’s psychology isn’t perfect, but at least it’s closer to the one on the right.

So let’s ask the real critical question:

Thanks ChatGPT!

ChatGPT actually knows us well, we have challenged him in many content before: