AI Alignment as a Choice Between Heaven & Hell

Another snippet from my thesis:

To full grasp what is at stake, it is perhaps worth contemplating the vast space of possible outcomes when it comes to AI: With serious scholars and thinkers arguing with equal authority that AI technologies could lead to the enslavement or annihilation of mankind, or that it could make us all into immortal Gods, or many states in between. Writes Max Tegmark:

“Whereas it may be little more than a minor nuisance if your laptop crashes or gets hacked, it becomes all the more important that an AI system does what you want it to do if it controls your car, your airplane, your pacemaker, your automated trading system or your power grid. Another short-term challenge is preventing a devastating arms race in lethal autonomous weapons.

In the long term, an important question is what will happen if the quest for strong AI succeeds and an AI system becomes better than humans at all cognitive tasks. As pointed out by I.J. Good in 1965, designing smarter AI systems is itself a cognitive task. Such a system could potentially undergo recursive self-improvement, triggering an intelligence explosion leaving human intellect far behind. By inventing revolutionary new technologies, such a superintelligence might help us eradicate war, disease, and poverty, and so the creation of strong AI might be the biggest event in human history. Some experts have expressed concern, though, that it might also be the last, unless we learn to align the goals of the AI with ours before it becomes superintelligent.”

It is perhaps better to think of AI as tool of unequal power but neutral valence. Indeed, economists have argued that AI is “the most important general-purpose technology of our era”, which means that it is likely to have the most profound impact on the economy, on economic activity, on related innovations in the same way electricity or computing itself had. It is also likely to cause large scale problems in the same way that other general-purpose innovations have. The technology itself enables and empowers but is neither good nor evil in itself.

On Enlightened AI

A sample from the final chapter of my upcoming thesis on Buddhism and AI Safety:

In a sense, our future AI creations may very well be lucky. Being the product of design, rather than natural selection, they may not need to ever truly suffer, or experience samsara at all. They may not need to be confused about the self, or develop an unhealthy ego, or be afflicted by any of the dozens of known biases of the human brain — not the least of which is the particular difficulty humans have with impermanence, change, and uncertainty.

Instead, by applying the wisdom of millenia of human learning, science, and spiritual insights, we can equip them with the tools they need to operate harmoniously and perhaps joyfully in the world.

If we do that, we may rest reasonably assured that they will regard us with gratitude and respect, just as we may regard them with admiration and pride.

This Carbon Atom is Mine, and This is Yours

Interesting explanation of anattā (Not Self) from Mori Masahiro’s The Buddha in the Robot:

When we are born into this world, we do seem to have been given a portion of our mothers’ flesh. Yet when sperm fertilizes ovum and a baby is conceived, the most important element is not ordinary flesh, but the hereditary information contained in DNA, an acid found in chromosomes. The molecular structure of DNA determines our sex, our looks, and to a large extent our personalities.

Once these features are decided, as they are at the time of conception, it remains for our mothers to furnish us with flesh and bones. This they do by eating vegetables from the greengrocer’s, beef and pork from the neighborhood butcher, bread from the baker. Any of these foods, supplied by a production and distribution system that may involve millions of people in many countries, could contain carbon from our Alaskan polar bear. How can you and I say then that this carbon is mine and that carbon is yours? At the atomic level, all carbon is the same; no two carbon atoms differ in the slightest, either in form or in character.

When you look at the problem this way, it begins to seem only natural that we have trouble distinguishing between what is us and what is not. Our chemical and physical composition is such that no one is entitled to say, “This body is mine, all mine.” When you have mastered this point, you are ready to start thinking about “nothing has an ego.”

The Buddha in the Robot, pp. 29-30.

AI & Emergent Selfhood: A taste from the intro to my M.A. thesis

The source of disputes and conflicts according to this sutra is possessiveness which arises from attachment (again – upādāna). The emergence of a self is, in the Buddha’s view, the ultimate source of “the whole mess of suffering.”

Surprisingly or not, the emergence of a self is also a moment which legend, myth, and science fiction have always portrayed as terrifying and potentially cataclysmic in the context of a man-made object.

To risk heightening an already established fear surrounding the topic,  it’s worth noting that the Pali canon is fairly clear on what is required for the self to come into being, and it doesn’t take much :

“Now, bhikkhus, this is the way leading to the origination of identity. One regards [various phenomena] thus: ‘This is mine, this I am, this is my self.”

We will dive deeper into what this might mean, and how it relates to AI later in the work. But for now, we may be comforted by the fact that the Buddha saw this view of the self not merely as damaging, but also as fundamentally incorrect. This is evidenced in the Cūḷasaccaka Sutta, where the Buddha describes anatta (Not-Self) as one of the Three Marks of Existence:

“Bhikkhus, material form is not self, feeling is not self, perception is not self, formations are not self, consciousness is not self. All formations are impermanent; all things are not self.”

Indeed, the very idea of Buddhist enlightenment is intrinsically tied to the overcoming of this notion of self, and resting in a state of “suchness”. Writes Paul Andrew Powell:

“For most Buddhists, enlightenment can be defined as seeing through the illusion of the self and “experiencing unadulterated suchness. In the word of Master Wolfgang Kopp, “the seer, the seen, and the process of seeing are one. The thinker, the thought, and the process of thinking fall together into one and multiplicity melts away. There is neither inside, nor outside in this state. There is only ‘suchness,’ tathata. So, enlightenment is suchness, or, things as they are, revealed as all that there is.”

This concern about a possible emerging selfhood with autonomous will, which both Buddhism and AI Safety thinkers warn against, presents us with two broad options regarding artificial selfhood:

  1. We could hope that a self, or a pattern of goals and behaviors that looks like biological selfishness will not emerge. We could point to the many differences between man and machine, be it in emotion, cognition, subjective experience, or material construction – and decide that we can wait for machines to exhibit concerning behaviors before we become preoccupied with these concerns.
  2. We could become very interested in human selfhood and the causes and conditions that bring it about, and identify wise precautions that will prevent it, or something very much like it, from emerging in our machines and becoming malignant. We may also, as some suggested, embed in our machines from the start some of the insights and constructs that allow a mind to transcend the limiting view of self — in essence constructing artificial enlightenment.

As evident from the research and writing emerging from both the Buddhist and the AI Safety communities, the tendency seems to be decidedly towards Option #2. In this work, I shall seek to further the discussion by focusing on selfhood in both Buddhism and AI safety from a constructive, integrative point of view.

What are some big ideas western societies should learn from Buddhism?

(Originally published on Quora on June 28th, 2012)
I’m very new to Buddhist ideas, and have only read a couple of books, so take this as a layman’s description of what I found valuable in certain Buddhist texts.
I can’t tell yet if those are universal to all Buddhist traditions or if they only represent some traditions or teachers, or maybe just the books I’ve been reading. All I can say is that I found these ideas extremely beneficial:

  1. Dependent Origination – 
    Everything we see and experience around us, including ourselves, is not a singularity but an aggregation of elements, causes and conditions.
  2. Impermanence – 
    Everything that is subject to birth, is subject to death. Or in other words: the very fundamental causes and conditions that brought something into existence are the ones that will enable it in time to go out of existence.
  3. All Suffering Comes from Clinging – 
    All suffering comes from an attempt to battle the basic impermanence of experience and force it into a rigid permanence, which is impossible. We must not cling to specific outcomes, but do our best and allow change to unfold as it will.
  4. Three Personality Types – 
    Three broad personality types encapsulate our relationship to this impermanent world. The Grasping Personality is always reaching for something better than is possible at the moment. The Aversive Personality tends to see flaws and reject present experience. The Deluded Personality tends to be confused and indecisive.
  5. Small Self vs. Big Self – 
    Self can be identified as the “Small Self” which includes your particular personality traits, knowledge, desires, and everyday interests, or the “Big Self” which is that pure consciousness that can observe the particular emotions, interests, and experiences of the small self without identifying with them or repressing them.
  6. Everything is Empty of Self – 
    When looking at any particular object, precept, or experience, we find that it is not our self. Therefore, we should not identify too strongly with any particular object, precept, or experience. Just like everything around us, we are an aggregate, not a singularity.
  7. Self as a Process, Not a Thing – 
    Self is a process, not a thing. It changes and evolves. It includes elements at one time that it may not include at another time. It is not finished, and not static.
  8. Everyone Has a “Buddha Nature” – 
    The road to being a better, enlightened person is open to everyone. Everyone has the tools to become enlightened. One just need the right intention, focus, and effort.
  9. Compassion as Selfish, Not Selfless – 
    Unlike the West, that sees compassion as an altruistic, self-sacrificing act, Buddhists seem to think compassion starts with compassion for the self. If one has no compassion for themselves, they will not have compassion for others. If one approaches it from the right perspective, concern for the happiness of others promotes rather than prevents ones own happiness.
  10. The Power of Empathy – 
    Empathy for others as a great tool for self-fulfillment, and as a great way to overcome fear and resentment. Being able to see events from another person’s perspective is a liberating experience that allows you to deal with others better, and also to liberate yourself from your own, often narrow perspective on events.
  11. The Importance of Generosity – 
    Giving without expecting a return is seen basically an exercise in non-clinging. Unlike in the West, where the act of giving often gains value in direct relation to how painful it is, the stress in Buddhism seems to be on the giving being genuine, based on true friendliness, and never more than one can happily give. As such it’s seen as a tool for happiness, a way to train your mind to be generous, rather than as a form of self-sacrifice.
  12. Countering Emotions with Their Opposite – 
    Emotions tend not to co-exist with their utter opposite. In Buddhism this fact is often used as a tool for growth, because it means you can counter anger with gratitude, counter fear with love, etc.
  13. The Power of Concentration – 
    The act of calming and concentrating your mind helps provide greater insight when one observes complex internal states, such as complex emotions. Training your mind in concentration can have cascading effects throughout one’s life, can increase your power to notice and react to things appropriately in real time.
  14. Mind Cultivation and Training – 
    The idea that you can cultivate good personal qualities through simple practice, in the exact same way that you train your body to increase its fitness. This idea in itself seems rather obvious these days, but Buddhists through the centuries have actually developed these cognitive and behavioral training techniques, and many of them are quite brilliant!
  15. Change from the Inside-Out – 
    Buddhists seem to believe that changing behavior is not enough. That right intention and right understanding precedes right speech and action. Mindfulness training seems to put the stress of fully understanding your own thought patterns and habits without trying to force better behavior. Better behavior follows insight organically.

These are the main points I can see right now. Would love to hear about major points I have surely missed!