During the Spring Festival I listened to a podcast episode, 92. How Do You Find Meaning in Life?. Combined with my recent thinking about the AI agent wave, a few thoughts — briefly noted here.

First, a summary of the podcast: the host argues that after society enters post-modernity, the lack of personal meaning becomes a widespread problem — manifesting as feeling that what you do every day is meaningless, being unable to find your own value, living without a goal.

The causes are complex, tied both to the nature of modern work and to the values of modern society. Modern work is extremely atomized; everyone is a cog in the system. Daily work is often separated from the final system built by a huge gap, so many people don’t really know what they are doing. And modern society emphasizes individual values, generally starting from personal rather than collective worth. For this problem, several mainstream remedies exist in society, but the host believes none of them is the right path.

The mainstream views that counter the loss of meaning include:

• Pursuing the maximization of material and monetary rewards. This is the prevailing societal approach, but clearly this goal cannot lead to personal happiness — at best it gives you an “optimization objective.”

• Pursuing the richness of personal experience. This looks like a solution grounded in personal values, but what counts as “worth experiencing” is itself shaped by societal values. The vast majority of answers here are confined to globetrotting, consumerism, and other consumptive activities. Few people go off to study medicine to save lives or devote themselves to charity. Most circle back to material needs.

• Pursuing love — treating it as the substitute for faith in post-modernism. I didn’t quite follow this part.

For this problem, the host offers his answer:

1. Do less purely transactional work. “Purely transactional work” is work that is not its own purpose — you do it purely for money. Its counterpart is practical work: work that itself (not the eventual paycheck) brings you joy. In post-modern society, we must either turn our hobbies into work, or turn our current work into a hobby.

2. Value community. That is, value the people we interact with closely in work and life; bringing value to a community also greatly enhances the sense of meaning. Science outreach and volunteering belong here.

3. Macro isomorphism. That is, my current work serves a larger goal — for example, although I am laying bricks, in my mind I am building a cathedral.

Weighing this against my own work and life, I think there is real truth in it. As an algorithm engineer, although I handle concrete, narrow problems in autonomous driving, I have always believed these are necessary work to make autonomous driving a reality. And autonomous driving itself can bring enormous value to society — it could even reshape the form of cities. Although sometimes I don’t fully buy into certain individual tasks, I can always find the value in other projects or from another angle — at worst, I sharpen my ability to decompose problems, drive projects, and collaborate. Precisely because of this, despite the heavy workload, I feel my work has value beyond the paycheck — it is meaningful.

In the AI era, when most work can be automated, the crisis of personal meaning will be further amplified. If a machine can do what I do — and do it better and cheaper — where does my value lie? If a person builds meaning on a specific work skill, like writing beautiful code, they may be standing on AI’s railroad tracks.

To find our own sense of meaning in the AI era, perhaps we must step outside the industrial system, stop trying to become better cogs, and pursue higher-level goals — things of more personal or community value, even doing things that seem “meaningless right now.”

For instance, from Liu Cixin’s Ball Lightning:

“Actually, son, it’s not hard to have a wonderful life. Let your father teach you: pick a universally recognized hard problem of the world — preferably a mathematical one that needs only a sheet of paper and a pencil, like the Goldbach conjecture or Fermat’s Last Theorem — or a problem of pure natural philosophy that needs neither paper nor pencil, like the origin of the universe. Throw your whole body and mind into studying it, tend the crop without asking about the harvest, and before you know it, a lifetime has passed. What people call devotion is really just this… So, the key to a wonderful life lies in what you can become obsessed with.”

Other

• TOPReward — as reinforcement learning spreads across fields, how to provide the reward is the core question. Using large models as the reward is only natural; the future really might be the left foot stepping on the right foot, ascending in a spiral.

• mimic-video — video generation models as foundation models for embodied AI. In terms of supervision-signal consistency, video modality across embodiments remains distributionally consistent. Compared with VLAs that force multi-embodiment training while ignoring mechanical differences, the video modality can better absorb the training set.

• GEAR-SONIC — RL for whole-body control of robot embodiments. By aligning RL to human motions, the robot’s action space is greatly reduced, after which further task RL can be done in a humanoid way. Combined with datasets like egoscale, this might be another path toward VLA.