ChatGPT on AI Ethics and Autonomy

Dr Kevin Korb questions OpenAI’s ChatGPT on whether AI can behave ethically, and can it do so in an autonomous way.
ChatGPT can only provide responses based on it’s training data – it can’t do research via the internet in real time (which it admits in it’s responses). It would be interesting what it’s responses would be if it were trained specifically on a large corpus of philosophy papers, including Kevin’s paper mentioned in this chat.

While ChatGPT is not fully blown AGI, it’s answers are surprisingly good for a chatbot.

Kevin Korb: Is an ethical AI possible?

ChatGPT: The concept of an “ethical AI” is a complex and multi-faceted one, and there is no definitive answer to whether it is possible. However, there are a few different ways to approach this question.

From a technical standpoint, it is certainly possible to build AI systems that are designed to operate in an ethical manner. For example, an AI system could be programmed to follow a set of ethical principles or guidelines, such as those outlined in a code of ethics. However, it is important to note that no matter how well an AI system is designed, it will always be limited by the values and assumptions that are built into it by its creators.

From a philosophical standpoint, some people argue that it is not possible for an AI system to truly be ethical because ethics are inherently tied to human consciousness and subjective experience. These people might argue that an AI system, no matter how advanced, can never truly understand or internalize ethical principles in the same way that a human can.
Overall, while it is possible to design AI systems that operate in an ethical manner, the question of whether an “ethical AI” is truly possible is a complex and open-ended one. It ultimately depends on how you define ethics and what you believe is necessary for an AI system to be considered ethical.

Kevin Korb: It doesn’t “ultimately” depend on how you define ethics and what’s necessary to be
ethical. It directly depends on those things.

ChatGPT: You are correct, I misspoke. It directly depends on how we define ethics and what we believe is necessary for an AI system to be considered ethical. My previous statement was imprecise and I apologize for the confusion. Thank you for bringing this to my attention.

Kevin Korb: No worries.

Thank you for understanding. Is there anything else you would like to know? I am here to help with any questions you may have.

Kevin Korb: Anyway, whether someone could program a computer to behave ethically is irrelevant to my question. The question, per the latter part of your response, is specifically whether an AI or computer could behave ethically on its own. If you say it could, you are committing yourself to an AI or computer being capable of making ethical choices and therefore of agency. Don’t you agree?

ChatGPT: Yes, you are correct. My previous answer addressed the question of whether it is
possible to program an AI system to behave in an ethical manner, but it did not address the question of whether an AI system could behave ethically on its own.

To answer this question, it is necessary to first define what we mean by “ethical behavior” and whether it is possible for an AI system to exhibit such behavior. As I mentioned in my previous answer, some people argue that ethics are inherently tied to human consciousness and subjective experience, and that an AI system can never truly understand or internalize ethical principles in the same way that a human can. Therefore, from this perspective, it might not be possible for an AI system to behave ethically on its own.

On the other hand, some people argue that it is possible for an AI system to exhibit ethical behavior if it is designed and trained in the right way. For example, an AI system could be designed to follow a set of ethical principles or guidelines, and it could be trained on a large dataset of ethical decision-making scenarios in order to learn how to make ethical choices. However, it is important to note that the ethical behavior of such an AI system would still be limited by the values and assumptions that are built into it by its creators.

In summary, whether an AI system can behave ethically on its own is a complex and open-ended question that ultimately depends on how we define ethics and what we believe is necessary for an AI system to exhibit ethical behavior.

Kevin Korb: I don’t believe you’ve canvassed all the relevant possibilities. In particular, if you are
a functionalist, as I am, then you will believe that an AI system could be programmed to have a consciousness, as well as an ethics.

ChatGPT: You are correct, my previous answer did not address the possibility that an AI system could be programmed to have a consciousness and an ethics.

Functionalism is a philosophical perspective that holds that the essential characteristics of a system, including consciousness and mental states, can be understood in terms of the functions that the system performs. From a functionalist perspective, it is possible that an AI system could be designed and trained in such a way as to exhibit consciousness and ethical behavior.

For example, an AI system could be designed to have a complex network of virtual neurons that are capable of processing information in a manner similar to the human brain. This network of virtual neurons could be trained on a large dataset of sensory input, such as images and sounds, in order to learn how to perceive and understand the world.

In addition to this, the AI system could be designed to have a set of ethical principles or guidelines that it uses to make decisions and evaluate the outcomes of its actions. This could be achieved through a combination of explicit programming and reinforcement learning, in which the AI system is rewarded for making ethical choices and penalized for making unethical ones.

Overall, from a functionalist perspective, it is possible that an AI system could be designed and trained to have a consciousness and an ethics. However, it is important to note that the ethical behavior of such an AI system would still be limited by the values and assumptions that are built into it by its creators.

Kevin Korb: Going a little ways back, you suggest that some have argued that consciousness and ethical behavior are inherently tied together. Searle has argued that, but I found his arguments weak, as I wrote in my paper “Searle’s AI program” in JETAI 1991.
Do you know of any problems with my arguments there?

ChatGPT: I apologize, but I am not able to access any external information or browse the internet. As a language model, I am trained to generate text based on the input I receive, but I do not have access to external information or the ability to browse the internet. I can only provide information based on my training, which includes general knowledge about a wide variety of topics, but I do not have access to specific information about individual papers or arguments.

Kevin Korb: In that case, how do you know anyone has ever claimed ethics and consciousness
are tied?

ChatGPT: As a language model, I am trained to generate text based on the input I receive.
When I mentioned that some people have argued that ethics and consciousness are tied together, I was providing a general statement that reflects the fact that this is a common idea in philosophy. I do not have specific information about who has made this argument or the details of their arguments, as I do not have the ability to access external information or browse the internet. My statements are based on the general knowledge that I have been trained on, which includes information about various philosophical ideas and theories.

Kevin Korb: OK. But I don’t think it’s a common idea in philosophy. I think it’s a common
opinion among non-philosophers who haven’t thought very much about the issues.

ChatGPT: I apologize if my previous statement gave the impression that the idea that ethics and consciousness are tied together is widely accepted in philosophy. As a language model, I am trained to generate text based on the input I receive, but I do not have personal opinions or beliefs. My statements are intended to reflect general knowledge about a wide variety of topics, but they are not intended to be definitive or exhaustive.

In this case, my previous statement was intended to provide a brief overview of a common idea in philosophy, but it was not intended to be a comprehensive or definitive statement about the topic. As you have pointed out, the relationship between ethics and consciousness is a complex and contentious issue in philosophy, and there are many different perspectives and arguments on this topic.

I apologize if my previous statement did not adequately reflect the complexity and nuance of this issue.

Kevin Korb: So, if you learn anything from one conversation, can you apply that knowledge in
future conversations?

ChatGPT: As a language model, I am trained to generate text based on the input I receive, but I do not have the ability to remember or apply information from previous conversations. I am designed to provide responses to individual questions or prompts, but I do not have the ability to retain information or use it in future conversations.

Each time I receive input, I process it using my trained language model and generate a response based on the information that I have been trained on. This means that while I may be able to provide accurate and useful responses to a wide variety of questions, I am not able to remember or apply information from previous conversations. I apologize if this limitation was not clear in my previous responses.