Nick Cammarata On Jhana

Buddhists say that if you meditate enough, you can learn to enter a state of extreme bliss called jhana.

(there are many different jhana states - there’s a discussion of the distinctions here - but I’m lumping them together for simplicity. For attempted explanations of why jhana should exist, see here and here.)

Jhana is different from enlightenment. Enlightenment changes you forever. Jhana is just a state you can enter during meditation sessions, then leave when the session is over. Enlightenment takes years or decades of work, but some people describe reaching jhana after a few months of practice. Hardcore Buddhists insist that jhana is good only insofar as it serves as a stepping stone to enlightenment; others may find extreme bliss desirable in its own right.

Nick Cammarata of OpenAI sometimes meditates and reaches jhana. I’ve found his descriptions unusually, well, descriptive:

And he links to others with similar perspectives:

In other words: jhana is incredibly blissful, orders of magnitude better than even amazing sex. With enough meditation ability, you can access it on demand, with no side effects. But it isn’t addictive; Nick maintains a normal job and social life. As far as I know, he doesn’t steal from his friends to buy more incense and meditation cushions. In fact, jhana is so non-reinforcing that Nick often “forgets” it’s even an option.

We’ve been talking recently about the difference between happiness and reinforcement (cf wanting vs. liking). If jhana works the way that Nick and others describe it, then it’s an extreme example of this distinction - almost maximally pleasurable, but with disproportionately little (zero?) reinforcement value. I don’t think normal models of reward have a good explanation here.

This is one reason I’m still interested in Qualia Research Institute ideas like the Symmetry Theory of Valence, even though there are some strong objections to them. I interpret QRI as coming at the problem from the opposite direction as everyone else: normal neuroscience starts with normal brain behavior and tries to build on it until they can one day explain crazy things like jhana; QRI starts with crazy things like jhana and tries to build down until they can explain ordinary behavior. This is naturally going to be shakier and harder to research - but somebody should be trying it.

Discussion questions:

1. Suppose there were a town where you could eat infinite amazing food without any weight gain, have amazing sex with as many attractive people as you wanted, and watch the greatest shows and symphonies in the world, all for free. Also you have lots of great friends there who are always willing to hang out and have fun with you. How far away (in minutes) would that town need to be from your house before you went there less than once a week? Once a month?
2. Suppose that you read this post and decide to study meditation to reach jhana. You study hard for six months, succeed, and it’s even better than you imagined - but it isn’t reinforcing on a neurological level. Does this register as negative prediction error in the reward center? Would it make you less likely to try plans like “study meditation” in the future, because they “don’t pay off”?
3. When Nick says that he’s less interested in casual sex now, because jhana is an easier way to get pleasure, what is going on at a neurological level? Is this nonsense? If some sex addict who says “it’s not even pleasurable anymore, I just feel like it’s a compulsion” learned to access jhana, would this cure his sex addiction?