2025-04-30 Angie: Expected role for you during my evaluation exam is bridge between Bayesian statistics and Bayesian cognition, because one of the committee members, Moshe benakiba, he does a lot of work on nested logic model, and his recent research is on online structural learning, which I think is kind of relevant with for us his work. So I my favorite chapter from for us his thesis is this part. And you also mentioned, yeah, that it's not always causal, but it's a program. But this thought is not very acceptable to causal inference oriented economists, but Scott Stern is one of my coming and he's open minded about it, and I've been trying to push back on causal inference, but we should do more proactive testing instead of passive testing. Does that make sense so far? Vikash: Maybe, yeah. I mean but, I mean, but you have a project that Scott is interested in already that would be a causal inference project. No, no, no, Angie: yeah. So they during the Bayesian reading group, we covered structural model instead of reduce for model. So he's interested in adding some more resource allocation and utility function, yeah, that sounds right, yeah. But it's not to the level of like, really privacy program and kind of my vision is like, when the Do you know, the birthday problem, yeah. So I observed that the reaction to the Facebook profit or modularizable time series differs, but in some level same, because between the Gen school and STEM school. This was, this is Ryan Bernstein's PhD thesis on we can have different modules, and based on that, we can build a network, and we can design some specification tools. I was collaborating with Ryan on this, and when I saw this, I thought, oh, maybe business model can be expressed like this, depending on what theory that you're using, you can somehow imagine, have a log likelihood fit of what you're observing, and you can have some region, Vikash: yeah, no. I mean, we've talked over a long period of time about versions of this idea which I agree are possible in theory. And the challenge is, where is there a very small, rigorous starting point? Yes, Angie: I made here. So I made an optimization framework which includes uncertainty and preference weight between four stakeholders, investor and the customer and operations, resource, partner and and based on what vector you're choosing and based on what your initial state is, your optimal path the first so for instance, like if you have different initial state versus better place, started from different capital level and different market understanding. And for instance, if your bull is from everything is uncertain in terms of operation, customer and the capital, then, depending on where you're starting from, your optimal first and also the optimal path can be when your preferences of the uncertainty is different. I pitched this to Charlie today, and he was quite acceptable with this good, great, yes. So I'm establishing trust. The only thing is, I don't have a lab. I'm the only student. Charlie, yeah, so, Vikash: yeah, I know it lonely, but I'm glad you have the Bayesian entrepreneurship group, yeah. And look forward to seeing you at the exam, Angie: yes, and having Josh Tenenbaum coming over and exchanging thoughts with Scott. Really enlightened Scott. So he thinks that now we should move a little away from social science, but to more computational. Vikash: I'd love to chat with Scott sometime, but I'll meet him, I guess, at the committee meeting. So 2025-04-24 without seeing the draft, advised to be conservative + address concrete problems + open to strong directional feedback, if mit phd is my goal i shared vikash charlie and angie's vision of implementing 10 scaling tools (which vikash called sci-fi 2024 summer) which i revisited (given his suprise of how much prob.prog can do) by sharing satelite was made thanks to author clark sci fi novelist (exchanged letters with mit scienteists) i modeled his updated evaluation and will compare it with his reaction on early may, before the chat after sharing my draft.