Last I checked labor was 15-30% for such projects, so it could help but only so much
653 karma · joined April 15, 2019
I guess I'm having trouble unraveling your experience and personal usage vs. what you're concluding about the labs.
It would be like saying proving ergodicity more generally for physical systems would unlock condensed matter physics, ignoring how well stat mech has served us regardless.
I am not anti-AI and I don't think we should stop throwing them at conjectures. I'm against this fundamentally misleading type framing that's become prominent. Millennium prize problems are important. Treating this specific aspect of NS as the one missing piece is just harmful. If we just throw compute at formal conjectures voila cancer and fusion.
I think the better example of "AI" usefulness toward solving problems is AlphaFold, and immensely powerful tool. But also suffering from a false framing/marketing problem as "solving protein folding". It feels like the right use of compute. Considering many factors that we can't hold in our head at once. "Solving" something that was already "solved" via computation (simulation) but now much more efficiently. The output is a valuable tool itself, it was not about "solving the protein folding problem", which it didn't do. It is a tool to solve problems requiring a sequence->ground state calculation. Which is a very broad set.
Formal verification of a conjecture we set up as a benchmark we set to test human understanding is not valuable in the same way.
I'm failing to make multiple points and gotta run, but i think that final point is important. The millennium prizes are not about technological/practical value, at least not intentionally. They're about shit that seems fundamental to us, things that feel[1] to us based on our understanding are important AND feel like they should be solvable in a human-comprehensible way. So formally resolving them with pure compute is not really the point. It seems closer to that story about one of those prime conjectures where some guy just ran brute force enumerations to find a counterexample. Valuable for sure, time-saving. And knowing the answer makes it a lot easier to solve a problem.
TL:DR science and math are more than formally resolving conjectures, they're about building up understanding and tooling that you can then build more on. AI should be an increasingly big part of it, but declaring "AI will solve fusion because it's smart" is like the rest of the fucking owl meme. I have no doubt it will help, most likely via simulations/quicker testing/calculations and verification. Maybe partly via reactor designs. Maybe partly being fed conjectures about bounds/limits that would be useful as inputs for the next iteration. And maybe even in the form of resolving some formally stated conjectures (I don't know enough plasma physics to name any).
[1] obviously to the mathematicians it's more than a feeling..
The OP is about a federal ruling saying clean water is not a constitutional right, no? And if you're disagreeing you're saying it should be a positive federal right in interpretation if not declaration, implying power of enforcement.
Still sounds like negative framing to me. And when those implicit non-declared rights are judged legally, they are still judged in a negative rights lens, no?
The right to legal counsel and jury are protections are still defensive in framing, though I concede your point. Access to education I would also concede is partially positive though you see it is about not denying. I would also say the more recent positive-sounding rights declarations/rulings are not always on super solid grounds, but this is admittedly circular.
I stand by my claim of "built on".
I know this will be read as me saying they shouldn't have clean water. What I'm saying is that declaring a positive right is not an effective approach to satisfying that right in this country, or anywhere that I'm aware. But especially this country. If I was in that jurisdiction I would be kind of insulted if this was presented as a reasonable approach to secure clean water.
Statistics is big. It's hard to answer without knowing what you plan on covering. I agree that most books (I've seen) suck. Wasserman all of statistics is the best I know, for my purposes. The problem I've found with most statistics books is trying to be too cute and clever by catering to a certain crowd thereby justifying holes that make it harder to truly understand and making the subject seem an incoherent patchwork.
The other problem that's even harder to solve is that the majority of people picking up a statistics book don't really think they need to learn statistics, just certain pieces. Which makes statistics seem less coherent and thus furthering the perception that statistics is in fact incoherent, leading to special background books that say "here's all you really need to know about statistics."
The final problem is statistics IS kind of incoherent as most often presented, especially when it tries to be what I'd roughly call "backward compatible". Why is so much time spent on p-values, for example? Is that really what a consistent modern perspective of statistics entails[1]? But backward compatibility forces it's inclusion, because that's what people still use and have used because they never got taught of "philosophy" of statistics, but rather rules. If you don't teach those rules everyone else uses, you're making them spend even more time on statistics than the small amount they're already unhappy spending. So the lowest common denominator is taught, which is incoherent. It's a vicious cycle that is not the fault of statisticians.
To an outsider there's not unifying dominant through lines because the field is justified largely through its application. There's no obvious "philosophy" or vibe of how to think about statistics. What I mean is that one approaching the field doesn't really grow more comfortable with it and doesn't really feel like they're advancing to a more coherent view, so they're less motivated to try to grok it, because it simply doesn't look like there is something to grok. With all of previous factors contributing to this.
So I guess the point of my ramble is that I think what's missing the promise of a reward for learning statistics "properly", really understanding it. In my biased view this is largely because it's not often presented to the non-professional statistics student as if there is a clear thing to understand, more just a collection of tools. And I think this is partly because the author already knows the reader doesn't want to spend time on it. So my advice would be, come up with a clear story of what statistics is. Promise and fulfill the promise that it is worth spending the time in a more abstract world for a little bit because the end is worth it. That it will save time and frustration in the long run because it won't just be a collection of basically faith-based tools with mystical rituals they will feel uncomfortable with for the rest of their careers. Tell that coherent story and then you can demand more commitment from your audience[2]. Maybe that cuts down your audience size but I think the net effect will be more people understanding statistics and what it actually is.
What would be the main threads and themes in your book? How would you justify spending time on it? What would require as necessary background? I don't understand the subject enough to offer any opinion. The most coherent things to me are convergence types and bounds. And then do you include computational approaches? For example in practice I'd say 90%+ of people would be better off using bootstrapping analysis for errors in most real-world cases compared to the standard "approved" approaches, but that's not very satisfying and the theory of it is certainly hard to integrate cleanly. TL;DR I don't envy anyone writing a statistics textbook.
[1] This is just a single aspect, but it gets to a broader problem. This is a really hard problem to overcome. So much of academia is taught p-values and to so much of academia that is basically the hardest math they know, and they work very hard to follow the rules they were taught, which to them is statistics. I don't know how you tell them to re-learn something especially when the answer is more math. That makes them even more hesitant and less likely to fully embrace a deeper understanding of the field.
[2] maybe that's the larger point. The analogy is maybe teaching calc vs. algebra based physics. Algebra based physics is harder to teach and learn, and it is less coherent and complete. There is not much meat in it, mostly rote tools. IMO conservation laws like energy and momentum and solving shit from there would leave a better taste for what physics is, but instead they are often used in one lecture to derive kinematic equations which are then elevated because students can memorize them. I guess it's better than nothing but it leaves people with the wrong impression of what physics is, just like I think most statistics leave a bad impression of what statistics is. Make the statistical equivalents of something like conservation laws the central objects. Something that makes it a coherent story that builds and rewards.
So it's kind of bottom-up/"data" driven as opposed to abstraction being more "engineered into" the system.
That is, to me abstraction is building things such that they look the same. Generalization is discovering things are almost the same if tweaked a little to look the same under a certain lens.
Does that make any sense? In this view I guess the generalizing lens can be become the basis for an abstraction. I would assume the loop is closed between the two somehow but I can't quite see it.
[1] https://www.pnas.org/doi/pdf/10.1073/pnas.79.8.2554 a paper he confidentally rejected as meaningless, unaware and not interested in the overwhelmingly likely possibility that his current brand new ideas trace back to it.
[2] "I’m stitching across a lot of different research, and I want to be clear many aspects of this system are not yet fully worked out.
But what we do know points to a system that works with different physical and algorithmic priors, and the dynamics are sharply distinct from current digital computers." Hmm, I wonder if "we" figured that stuff out from individual and especially collaborative efforts combining abstractions and more theoretical approaches with expertise in biology, not to mention EE, biophysics, physics, and whatever else other fields that don't know about his favorite molecule and thus can be ignored if not shit on?
Then we actually founded a startup and realized this was not a silly question. But perhaps that's circular when the default path is VC funding, hmmm.
This is sad conclusion. Maybe you're more right than wrong, but my answer to the "value" question would be:
Solving even harder problems based on lessons from solving easier ones? Pretty similar to the trajectory in this blog post. Perhaps aided additionally by llms and other tools that can solve the subproblems so you can focus on the less obvious/automatable aspects of the problem?
But note the tension, only by being involved in the problem solving to some extent do you become better at it. So if your default is to say "An LLM can or will soon be able to solve it, why bother?", then your situation will become more desperate and your outlook more negative in a self-reinforcing way.
Further, if there is a correlation, I'd bet it's not so much an intrinsic "creativity" trait, but more effectively higher creativity because more trials. That is, along the lines of Chollet's paper, a measure of creativity should be based on a fixed budget with fixed knowledge.
Among many other possibilities I haven't considered, perhaps another mechanism could be that because ADHD people spend more time thinking in less goal-oriented ways and mixing thoughts on accident, perhaps we do in fact gain some learned creativity via experience with vagueness[1]? But that might also imply that part of creativity is actually being able to diffuse more freely through thought space and lowering the barrier to attempted connections between ideas. That lower barrier leads to less likelihood of any "collision" being meaningful but maybe it's overcome by higher collision rates? Or maybe effectively higher order (not just pairwise) collisions?
Disclaimer in case it's not obvious: I don't know any of the literature on what creativity even means or how it's quantified.
[1] Which is me injecting an assumption that creativity ~= connecting things with no obvious or well-troden reasoning path between them.
edit -- oops just looked at your profile after seeing someone elses comment. I assume you are stating a fact then, leaving original anyway
It's more about computational complexity it seems but maybe it cites stuff that might interest you.
It's using computational neural networks that no one has ever believed represent the biology of the brain, but I think I must still disclose the following to save the precious time of the genius solving the problem all by himself:
It doesn't fully account for every biological detail ever documented, so it's probably a meaningless "curiosity".
Oh it probably also doesn't account for every detail discovered since publication so even if had value at time of publication it is not worth reading now.