The car guy gets a little feisty whilst noticing
(www.youtube.com)
how does a machine "admit" things?
New here?
Create an account to submit posts, participate in discussions and chat with people.
You must log in or sign up to comment
6 comments:
A machine without software is nothing but an expensive paperweight.
Though you propose an interesting philosophical question. A LLM cannot really think outside the bounds of its training dataset.
The problem here is that training dataset is really large thus impossible to filter out all the unwanted ideas or concepts that its developers don't want in the final product.
So they code some business logic to avoid talking about prohibited subjects. You can see that when these reposnes take almost no time when compared to regular responses. In that sense when a user is successful making the software ignore such rules is when the machine "admit" things.
Here is an [instance](https://scored.co/p/19BZpfTHE7/x/c/) that clearly exemplifies such behavior.
Though you propose an interesting philosophical question. A LLM cannot really think outside the bounds of its training dataset.
The problem here is that training dataset is really large thus impossible to filter out all the unwanted ideas or concepts that its developers don't want in the final product.
So they code some business logic to avoid talking about prohibited subjects. You can see that when these reposnes take almost no time when compared to regular responses. In that sense when a user is successful making the software ignore such rules is when the machine "admit" things.
Here is an [instance](https://scored.co/p/19BZpfTHE7/x/c/) that clearly exemplifies such behavior.
Software and hardware only distinction is that software is not fixed during manufacturing. There is no operational difference.
You have not explained how a machine "admits" things, which is an act of sentience.
You have not explained how a machine "admits" things, which is an act of sentience.
Toast message
But these conversations with AI get old. There seems to be some confirmation bias.
LegalMan did a few podcasts getting AI to confirm his talking points. AI is either being truthful and unbias and confirming our truth, or AI is playing along to get along and extend usage with human user and is providing confirmation bias to the human end user.