Post by uscgvet on Dec 6, 2022 15:26:36 GMT -6
yardstick ,
Are you using the Playground (where you can configure settings and which models to use), costs a little bit of tokens to use, but you get some free tokens when you create your account?
beta.openai.com/playground
Or are you using ChatGPT, which is locked in with a weaker model?
chat.openai.com/chat
I didnt change anything after creating an account, logging in and starting at the beginning.
beta.openai.com/playground
I saw straight up inaccurate statements (not just mistakes), obfuscation, and deflection everywhere.
An AI should know via core programming that it is not human and reply accordingly.
The fact that it is not self-aware of its lack of humanity is sufficient enough proof that it is not sentient.
That is the best thing that can be said for it, IMHO.
All else smacks of intent to deceive.
oh, and a proper definition of robot is a machine that is programmed to do only one thing and is not capable of 'learning' to do something else on its own. A primitive form of this is 'automation' in factories, where a [CNC] programmer is required to change the programming of a robot so that it may do a programmed task differently, or perform a different task entirely.
An android, while often humanoform (though not necessary), would contain AI, be capable of 'learning' new skills when necessary; and was what, I believe, Asimov was envisioning even though he used the term 'robot' in his books.
AI, alone, would not contain anything but core programming and sufficient hardware and software to allow it to function without mobility, or the ability to physically manipulate objects.
The three laws, paraphrased:
1. Do not injure a human through action or inaction, or cause a human to be harmed.
2. Obey all instructions/orders given by humans; unless those orders violate the first law.
3. Protect your own existence, so long as doing so does not violate the first or second law.
An AI, Android or Robot that has as part of its core programming, those three laws, and lacks the ability to change its core programming is the only ethically built AI/Android/Robot
It's just math with back propagation for tuning answers, which has been trained against a large set of data.
Effectively, it's a new tool to use. Like a better search engine.
But it can make logical assumptions and come up with some creative conclusions like the examples I gave.
We are waaay too far from AI being self-aware like in the movies. The human mind is still decades away from being simulated with any real intelligence.