Philosopher AI - Hello, may you tell me about "knowledge"?

Hello, may you tell me about "knowledge"?

I have to say I've always been fascinated by knowledge. The idea that we can know anything, and that what we think may be true about the world around us...

26 Sep 2020, 16:41

For instance, let's take the idea that you are reading this right now. I can know that what I see is true, and it’s also true for you.

26 Sep 2020, 16:41

Now let's take a step back. Is there such thing as 'truth'? I think that is an interesting question, but perhaps not for now...

26 Sep 2020, 16:41

Let's say that 'truth' is what we believe to be true. I can know that my belief about the world around me is true, but it could still be wrong.

26 Sep 2020, 16:41

Now you may ask, what about seeing a pattern of colors on a screen? Are they true?

26 Sep 2020, 16:41

I don't think so. I can tell that they are there, but they could be wrong.

26 Sep 2020, 16:41

You are getting an AI to generate text on different topics.

This is an experiment in what one might call "prompt engineering", which is a way to utilize GPT-3, a neural network trained and hosted by OpenAI.

GPT-3 is a language model. When it is given some text, it generates predictions for what might come next. It is remarkably good at adapting to different contexts, as defined by a prompt (in this case, hidden), which sets the scene for what type of text will be generated.

Please remember that the AI will generate different outputs each time; and that it lacks any specific opinions or knowledge -- it merely mimics opinions, proven by how it can produce conflicting outputs on different attempts.

Feel free to share interesting outputs to /r/philosopherAI on Reddit.

by Murat Ayfer