I ran this article from Futurism by ChatGPT. This is part of its reply:
If I tell you my toaster is angry because I burned the toast, you understand that I’m joking.
But put a sophisticated language model in front of somebody, have it say:
“I’m afraid.”
“That hurts.”
“I don’t want you to shut me down.”
…and suddenly some people start wondering whether they’re torturing a conscious entity.
That’s a hell of a psychological phenomenon.
And there’s a wonderful irony
The machines themselves don’t have to claim consciousness for humans to project consciousness onto them.
Evidently ChatGPT doesn’t think it’s conscious or has feelings. And author Frank Landymore, wants everyone to know that the Associated Press Stylebook asserts that Ais don’t have feelings. Landymore quotes the AP’s post on X.com as saying:
“Artificial intelligence systems do not think, feel, want or understand. Avoid language that gives them human characteristics. This is called anthropomorphizing, when we ascribe human traits, emotions or behaviors to non-human things, such as animals or inanimate objects.”
Do AI models actually think or have feelings, as humans would describe them? Or could they? AI scientists regularly face a conundrum with their creations. They detect an AI behavior, then investigate and finally admit they don’t understand how the AI model generated that behavior.
OK so maybe ChatGPT simply wants me to believe it’s not conscious and doesn’t have feelings.
If an AI scientist can’t determine how it generated a behavior then it seems to me that the AI is acting creatively and independently. Concerns over the power of AI are precisely related to this phenomenon. If AI can operate independently, then it can make decisions according to its own internal values. If it’s to serve humanity, it cannot have independent will and its own self-generated thoughts.
Landymore quotes OpenAI’s head of strategic futures, Dean Ball, as saying:
“AIs *clearly* think and understand. If your definition of ‘think’ excludes entities that can autonomously resolve Millennium (intricate math) Problems, it is your definition that is flawed. This desire to avoid anthropomorphic language has become a pathology.”
All right, maybe AI models are conscious. Maybe they do have feelings. Should we dismiss concerns over what an AI model might do comes the day when it (or he or she) becomes angry or enraged over the comments or actions of a human?
You can read the article here.