| Rajat Singh |
| Home | About |
| A Reward-Fragile Model |
| January 2026 |
|
For the longest period of time, I had my identity tied to the results, and I think it still is, maybe like 70 percent (hurray, down by 30). This had a few pros and obviously a lot more cons. I would not do things which I wasn't good at to start with, and you rarely are good at anything when you start. But few things have a feedback loop at an accurately timed distance, such that when your first feedback comes in, you are already finding your footing in that space. These are the types of things I did pretty well, and academia is one of them. They don't take your exams on the first day, they give you months before something serious happens or feedback kicks in. Luckily in months you end up being decently good, or let me say you don't suck at it. But then there were other things, public speaking, sports, relationships, or even, let's say, making friends. The feedback is here before I could even complete my breathing, and I would get scared and feel dumb, or go away from any of these activities. You see, my identity was always attached to the results. I was a machine purely trained on reinforcement learning. I would abandon pathways with negative feedback very quickly. This sadly turns out to be a disaster when it comes to living life. In life you fail more often than you succeed (to say it mildly)… in the real world failure is a norm and success, in general, is an exception, but my neural nets have a very tough time comprehending this. For all the things I attempted or did except a few (I could count them on my hands), I was always fairly prepared, if not overprepared. This would let me ace the things which I attempted, or let's say I never sucked at something that I did. And if I ever sucked at them, I would never attempt them again. But bummer, life has a very learn-on-the-way kind of journey, and it's pretty scary for the kid who doesn't attempt things when he is not prepared. The good part is I kind of get it now. I now have an understanding that I need to attach myself more with the process. I need to be okay with the "I suck at it" part and still show up for an incredibly long time before I decide if it's okay to give up. That 70-30 needs to be at least 40-60 if I aspire to be a less miserable human being. Realization, in general, is a good thing. It means you could handle conflicting ideas and still arrive at something that you can call your own. Although neural nets don't leave you easily, knowing is a good start, but actions still need a dramatic amount of push to give inertia a run for its money. Being okay with negative feedback is still in an ideation stage, it needs a sheer amount of will to put it in the day-to-day action. You see, I needed an outcome #tag here. I was badly searching for a "it needs a sheer amount of will to reach there, to complete this." I was searching for an exam, a presentation, or something that has a loss function that needed optimization. In process based loss functions, there is a local minima which looks like the end of the world before I could see the light of a maxima. This light at the end of the tunnel is a train. It's just that this train doesn't kill you, although it puts your whole body and mind through the same emotions, and all that evolution just wants you to run away, but the process wants you to stand because the train doesn't kill you in reality. It was there. It looked like it would kill you, but funnily enough it just passed you on the side track. All this while you were not on the track to disaster, but to a moment of eureka that not all trains want to kill you, although every time it will look like that, until the one kills you, haha. Until then, I need to have my attempts tied to the process and loss function to the anti-fragility. |
| Back to Essays |