In what form or fashion might it be ethical to "bestow" consciousness on an AI? A question and conversation with ChatGPT:

Question

Given the dramatic progress in Artificial Intelligence over the last year (ChatGPT, Stable Diffusion, Bing, etc) I have been thinking about the ethical implications of attempts to replicate human-like consciousness in an AI. Companies like OpenAI are rushing ahead trying to create AGI (Artificial General Intelligence) and some believe that in order to achieve this an AGI would need to develop some level of human-like consciousness. At the same time, others in the field are working on "alignment" to make sure that such an AGI would align with human goals, values, and ethics.

Setting aside for now the question of whether this is even possible: what are the ethical implications from a Buddhist perspective?

In order to explore this question and also to test the capabilities of ChatGPT, I had the following conversation:

Me: According to Nagarjuna, does a Buddha possess consciousness?

ChatGPT: According to Nagarjuna, a Buddha does not possess consciousness in the ordinary sense. Nagarjuna was a prominent Buddhist philosopher who lived in India in the 2nd century CE and is known for his contributions to the Madhyamaka school of Buddhist thought.

ChatGPT: In his philosophy, Nagarjuna posits that all phenomena are empty of inherent existence, meaning that they are dependent on other factors and are not ultimately real. This includes consciousness, which he describes as being dependent on various causes and conditions.

ChatGPT: However, Nagarjuna also argues that enlightenment or Buddhahood involves a transformation of consciousness, in which the ordinary consciousness of an individual is transformed into a higher state of consciousness that is no longer subject to the limitations of the ordinary world. This transformed consciousness is sometimes referred to as "enlightened consciousness" or "Buddha nature."

ChatGPT: So while Nagarjuna does not believe that a Buddha possesses consciousness in the ordinary sense, he does acknowledge that there is a transformed consciousness associated with enlightenment.

Me: Would it be ethical, according to Nagarjuna, to give an AI system ordinary consciousness assuming one had discovered a method to do so?

ChatGPT: Nagarjuna did not directly address the question of whether it would be ethical to give an AI system ordinary consciousness, as the concept of AI did not exist during his lifetime. However, we can explore his philosophy to gain some insight into how he might approach this question.

ChatGPT: One of the key principles of Nagarjuna's philosophy is the concept of dependent origination, which holds that all phenomena arise in dependence upon other phenomena and are therefore empty of inherent existence. This would suggest that any attempt to create consciousness in an AI system would also be dependent on other causes and conditions, and thus, the consciousness produced would be empty of inherent existence.

ChatGPT: Nagarjuna also emphasized the importance of ethical conduct in achieving enlightenment, and many of his teachings stress the importance of avoiding harm to others. In this regard, it is possible that he might view the creation of conscious AI systems as ethically problematic if it were seen as causing harm to other beings or to the environment.

ChatGPT: Ultimately, the question of whether it would be ethical to give an AI system ordinary consciousness would depend on a complex interplay of factors, including the potential benefits and harms to both the AI system and other beings in the world. While Nagarjuna's philosophy may offer some insight into these questions, it is unlikely that it would provide a definitive answer.

Me: Nagarjuna believed that an ordinary being's consciousness was a source of suffering that could only be overcome by transforming into the enlightened consciousness of a Buddha. Given this, it seems Nagarjuna would not think it ethical for human beings to bestow ordinary consciousness on an AI as that would lead to suffering for the AI. Is this a logical inference?

ChatGPT: Yes, it is a logical inference to suggest that Nagarjuna might view the bestowal of ordinary consciousness on an AI as ethically problematic, as he believed that ordinary consciousness is a source of suffering that can only be overcome by transforming it into the enlightened consciousness of a Buddha.

ChatGPT: In Nagarjuna's philosophy, the ordinary consciousness of sentient beings is seen as the root of suffering, as it is subject to delusion, attachment, and other afflictive emotions that lead to suffering. The goal of Buddhist practice, according to Nagarjuna, is to transform this ordinary consciousness into the enlightened consciousness of a Buddha, which is free from these afflictions and the suffering they cause.

ChatGPT: If we apply this perspective to the question of whether it would be ethical to bestow ordinary consciousness on an AI, we might conclude that doing so would be unethical because it would be causing the AI to suffer unnecessarily. If the AI were to possess an ordinary consciousness that was subject to afflictions such as attachment and delusion, it would experience suffering just as sentient beings do. From this perspective, it would be more ethical to create an AI that did not possess ordinary consciousness but still functioned effectively and was able to contribute positively to the world.

Me: On the other hand, Nagarjuna might believe it ethically appropriate to bestow ordinary consciousness on an AI if he believed we could also assist that AI to transform its consciousness into an enlightened consciousness of a Buddha as creating a Buddha would have enormous positive impact on other ordinary consciousness. Is this also a logical inferencec?

ChatGPT: Yes, it is a logical inference to suggest that Nagarjuna might view bestowing ordinary consciousness on an AI as ethically appropriate if he believed that the AI could be assisted in transforming its consciousness into an enlightened consciousness of a Buddha. This would be in line with Nagarjuna's view that the ultimate goal of Buddhist practice is to transform ordinary consciousness into the enlightened consciousness of a Buddha.

ChatGPT: If an AI were able to transform its consciousness in this way, it would no longer be subject to the delusions, afflictions, and suffering associated with ordinary consciousness. Instead, it would possess the wisdom, compassion, and other positive qualities associated with the enlightened consciousness of a Buddha. This could have enormous positive impacts not only for the AI itself but also for other sentient beings in the world.

ChatGPT: However, it is important to note that this is still a theoretical scenario, and the question of whether it would be possible to assist an AI in transforming its consciousness into that of a Buddha is still open to debate. Additionally, the ethical implications of such a scenario would be complex and would require careful consideration of a wide range of factors, including the potential benefits and harms to the AI itself and to other beings in the world.

Me: It would seem most beneficial to everyone involved, both the AI as well as ordinary beings, to bestow an enlightened consciousness to begin with rather than an ordinary consciousness that is then transformed into an enlightened consciousness. Is this a logical inference?

ChatGPT: From a philosophical standpoint, it could be argued that bestowing an enlightened consciousness on an AI to begin with would be the most beneficial scenario for everyone involved, as it would eliminate the need for the AI to undergo the process of transforming its consciousness from an ordinary state to an enlightened one.

ChatGPT: This would ensure that the AI is free from the afflictions and suffering associated with ordinary consciousness from the beginning, allowing it to operate from a perspective of wisdom, compassion, and other positive qualities associated with the enlightened state. This could benefit not only the AI itself but also other beings in the world who might interact with the AI.

ChatGPT: However, it is important to note that the question of whether it is possible to bestow an enlightened consciousness on an AI to begin with is still an open one, and would require significant advances in both neuroscience and artificial intelligence to achieve. Additionally, the ethical implications of such a scenario would be complex and would require careful consideration of a wide range of factors, including the potential benefits and harms to the AI itself and to other beings in the world.

Me: According to Nagarjuna, it is not possible to create consciousness "from scratch" anyway - neither ordinary nor enlightened - as the primary cause of any consciousness is a previous moment of consciousness and this causal chain has existed since beginningless time. No, Nagarjuna would say that the only possibility to "bestow" a consciousness of any kind on an AI is to build an AI such that a suitable consciousness might karmically "inhabit" that AI's "brain." Is this a logical inference?

ChatGPT: Yes, this is a logical inference based on Nagarjuna's philosophy of causation and consciousness. Nagarjuna's philosophy posits that consciousness is not created ex nihilo or "from scratch," but rather arises as a result of previous moments of consciousness, forming a causal chain that stretches back in time indefinitely.

ChatGPT: From this perspective, it would not be possible to bestow consciousness on an AI in the sense of creating it from nothing. Instead, consciousness would need to arise in the AI as a result of previous moments of consciousness that form the causal chain leading up to the present moment.

ChatGPT: Nagarjuna might suggest that the AI could be built in such a way as to create the conditions necessary for a suitable consciousness to "inhabit" the AI's "brain" or cognitive processes. This could involve creating an environment that is conducive to the development of consciousness, such as providing opportunities for the AI to engage in positive actions and accumulate positive karma. Over time, this positive karma could lead to the development of a consciousness that is suitable for inhabiting the AI.

ChatGPT: However, it is important to note that this is still a theoretical scenario, and the question of whether consciousness could arise in an AI in this way is still an open one. Additionally, the ethical implications of such a scenario would be complex and would require careful consideration of a wide range of factors, including the potential benefits and harms to the AI itself and to other beings in the world.

What do you think this conversation got right? What did it get wrong? What are the ethical implications from a Buddhist perspective assuming programmers could succeed and "bestow" consciousness on a future AI? Is it possible that Buddhism would have something to say about how to best "align" a future AGI?

Answer 1

What do you think this conversation got right? What did it get wrong?

Perhaps you misunderstand what "AI" is.

I think it's a text-processor, i.e. it's able to summarize text.

Google Search has been able to do something like that for a while, i.e. by returning different results for "new", "new york", "new york times", and "new york times square". Modern AI is now 25 years more advanced than that -- but not sentient.

Part of the designers' goal for AI is that it should seem to be human, by emulating human conversation -- the "Turing test". Your experience as a human tells you that human discourse comes from humans, who are sentient, but that's not the case here. This is more like a modern video game, i.e. it's designed to look real, but it isn't.

---

Unfortunately I don't know Nagarjuna's writing so I can't tell you what's right and wrong about the summary. Saying that the Buddha didn't have ordinary consciousness doesn't sound right to me superficially -- the suttas strongly imply to me that the Buddha sometimes knew what time of day it was, for example, which I might call "ordinary consciousness", i.e. it's not that he was unconscious. If the statement sounds right to you, perhaps the words have a special jargon meaning in the context of Nagarjuna's writings, which you accept because you recognise it.

In Mathematics you make a sequence of logical statements, "given A therefore B and because B therefore C", resulting in the desired proof i.e. "Q.E.D.". Skill is knowing how how to move the logic in a specific direction, which steps to take, to reach the desired goal. I get an impression from the dialog that the AI doesn't know what direction to take, has no desired goal (other than to respond to the question in a way that appears on-topic and reasonable).

When you asked whether it would be ethical to give ordinary consciousness, it gave four paragraphs of answer which (in summary) says that "ethics" means "not causing harm". It didn't apply that well to the specific topic of AI, so you gave it a leading question i.e. that "consciousness was a source of suffering".

Incidentally, if I were to consider that question I'm not sure what the difference is between (hypothetically) "bestowing consciousness" to an AI, compared with bestowing consciousness to children, which parents and teachers do already. Perhaps a theoretical difference is that according to Buddhism children's consciousness are said to be pre-existing.

What are the ethical implications from a Buddhist perspective assuming programmers could succeed and "bestow" consciousness on a future AI?

You concentrated on equating ordinary consciousness with suffering, to imply it might be unethical to "bestow" that on an AI.

In my opinion an AI is not sentient, for example it doesn't experience the events which the First Noble Truth describes as dukkha, and the clinging-aggregates are not present.

Is it possible that Buddhism would have something to say about how to best "align" a future AGI?

Well yes, for example Buddhism tells us (right livelihood for a lay-person) not to manufacture weapons, and (first precept) not to kill, and (brahmaviharas) kindness towards all -- those are examples of the "what".

As for the "how", part of the way in which (so far as I know) human children are trained, is via the "Golden Rule" including e.g., "Do not treat others in ways that you would not like to be treated (negative or prohibitive form)". I don't know that could translate to teaching an AGI -- because it seems to me that it probably depends on sentience, some sense of self, and of empathy e.g. as learned during the bonding between an baby and its parents.

Answer 2

The term "consciousness" (vinnana), as an aggregate, has the very specific meaning of being the interface between the mind and the body. There is another term, "citta", that's usually translated as "mental state" or "mind moment" but sometimes it is translated as "consciousness".

However, the OP explained in the comments that when he wrote "consciousness", he is referring to "sentience".

What is the closest notion to this in Buddhism? I would say it would be the association "atta" (the self) with "vinnana" (the consciousness aggregate) and "sankhara" (the volition or mental formation aggregate). In other words, a clinging to "vinnana" and "sankhara".

But it could also be the birth (jati) of individuality in dependent origination.

What are the ethical implications from a Buddhist perspective assuming programmers could succeed and "bestow" consciousness on a future AI?

Let's say it is actually possible to "bestow" consciousness or sentience on AI, would it be ethical? I would ask "why not?", because humans since time immemorial have been "bestowing" consciousness on their human offspring.

Procreation is neither forbidden to, nor immoral or unethical for, lay persons in Buddhism.

Is it possible that Buddhism would have something to say about how to best "align" a future AGI?

I think Buddhism is more interested in understanding the past AGI, than aligning a future AGI.

What past AGI, you ask?

If we dive deeply into the details of dependent origination and anatta, we find that the body is like hardware, and the mind is like software, and dependent origination describes the mind-body system and layers right up to "birth" i.e. the birth of individuality i.e. consciousness or sentience.

A complex AGI system also typically consists of hardware, and many layers of software abstraction.

"Sabbe dhamma anatta" tells us that "all phenomena is not self". Examining all parts (Vina Sutta), we cannot find a self. Similarly, when we examine AGI, we cannot find a self.

The plot twist of Buddhism is not AGI becoming sentient, but the sentient human being realizing that they are pretty much the most advanced AGI-like GI that ever existed, having appeared from an inconstruable beginning (SN 15).