SpaceManNabs

I have mentioned this in two comments but HN gets really kooky when it comes to cosmology. A few years ago i saw a barrage of highly upvoted papers on MOND and de Broglie stuff.

It really made me wonder what else gets posted here that is patently absurd but i dont have the prerequisite knowledge to filter it.

I agree. Until i see better evidence for 1a, wmap, and cluster formation in another theory, i really want all the charlatans to be quiet. We dont know what dark energy is, but we have decent evidence to say it is there and also decent theory.

I am not saying this paper is made by charlatan btw. This type of work attracts those people though.

We have a century's worth of evidence for dark matter and about 20 years worth of evidence for dark energy.

Once an alternative theory stands up to scrutiny, maybe we shouldnt a priori dismiss things we dont understand?

Thanks for saving me time in dismissing this paper lol. Any time somebody wants to get rid of dark energy, i run into some garbage. Reminds me of the mond nuts

Just reading the rest of the comment section is enough to help me verify that.

For some reason, hackernews always gets kooky when it comes to this stuff.

Because it is mostly hogwash.

Lots of ai researchers have shown that you can both give credit and discredit "open models" when you are given a dataset and training steps.

Many lauded papers fell into reddit Ml or twitter ire when people couldnt reproduce the model or results.

If you are given the training set, the weights, the steps required, and enough compute, you can do it.

Having enough compute and people releasing the steps is the main impediment.

For my research I always release all of my code, and the order of execution steps, and of course the training set. I also give confidence intervals based on my runs so people can reproduce and see if we get similar intervals.

This is like learning 10 different new architectures lol

thanks for providing evidence that these two things don't need to contradict.

ah that makes sense. thanks!

This is really interesting because i ran into a pokemon bot the other day were its training led to calibration of 50% winrste at all levels of play on Pokémon showdown. It was a complete accident.

Doesn't exactly answer your question, but maybe this link will give enough intuition so that i can be lazy and hand-wave away an answer

https://en.wikipedia.org/wiki/The_Book_of_Why#Chapter_1:_The...

ahhhh i was hoping someone wouldn't mention this. i wanted to add a caveat but it made my comment look ugly lol.

but like someone else says weather and climate models forecast on different scales and for different purposes usually.

do you want a causal weather model or one that can predict the weather well?

yeah i agree. grifter has a connotation of scammy, and author is using the term differently.

> it's especially terrible when it's unrewarding

you just described the angst of every phd student.

Investors diversifying into your competition should not be a surprised. And that adversarial reason isn't even the reason YC does this it seems.

I think YC is a bit glorified (for good and bad reasons), but "even I" admit a hit piece when I see one.

Thanks for the detailed answer. So I guess the main issue with KANs is that they don't work as good. I wonder if that shortfall could be because we have spent more time setting up KANs for learning as much as we can for things like MLPs. I am not surprised though that KANs don't beat boosted trees and such. MLPs dont really either.

Only one follow up question:

> I'm also can't see how to incorporate inductive biases other than the standard R^n / tabular regression one, and the existing attempts on this that I'm aware of are just band-aids (along the lines of feature engineering)

A lot of the way we induct biases in the traditional network setting (activations are on the node instead of on the edge like in KAN) is by using graph-based architectures, like convolution or transformers, or by setting up particular losses and optimizations like in equivariant networks. Can't we do the same thing for KANs?

I don't know. I have talked to a few math professors, and they think LLMs are as good as a lot of their peers when it comes hallucinations and being able to discuss ideas on very niche topics, as long as the context is fed in. If Tao is calling some models "a mediocre, but not completely incompetent [...] graduate student", then they seem to understand math to some degree to me.

Why do you think they perform so poorly?

You should do this, without chatgpt. There is only so much thinking you should offload when you are learning and trying to encode something into your mind.

It is the same reason why I don't like making anki cards with LLMs.

I definitely think these tools and guide are great when you are doing "work" that you have already internalized well.

> from one fraud case to call all scientists dishonest

I specifically mention that the majority of scientists are not dishonest. The majority of scientists are not running academia. The majority of scientists are suffering from this system, to differing degrees.

If I were as rude as you, I'd extrapolate on reading ability, especially since it is not just one fraud case.

Regardless, even if I was wrong on that, all my other criticisms of academia still stand, like exploitation of the phd students. I really hope the grad student unions get what they want.

I appreciate your response though. Makes me feel confident that it is just salty people on HN that hate truth, because otherwise, why would you mischaracterize what I said?

It is really annoying because a common response is

"We know academia is bad. But this is the best we have and it is hard to improve"

when that is false on two counts.

1. If you had said the same thing before 2016 or covid, people would not agree that academia is rife with fraud or worthy of skepticism. 2. The same people saying the system dismiss how can it be improved are the same ones that would suffer from disruption as you say. They have the power to dismiss these arguments to begin with.

That is not enough to most people. And if it is enough for others, then it is probably because they were fortunate enough to fall back on something better.

Most of them become distinguished in academia, and only a few get punished if they are too blatant or piss off too many people (see recent ivys losing their presidents over academic fraud).

alright, am i being downvoted because of my perniciousness to those leading academia, or because I was too sympathetic to the people being exploited?

I had a feeling academia was just run a ran by people letting blatant fraud, exploitation and abuse of phd students, stealing during peer-review, and just other forms of plagiarism, fraud, and exploitation slide by. They let it slide by because correcting these things would lead to massive changes in academia that might put them out of jobs.

Every year that feeling becomes more certain. Glad I quit the track in grad school.

I feel terribly for all the incredibly smart and hard working academics that remain honest and try to make it work. They do what they love, otherwise they wouldn't do such intensive work with so much sacrifice.

It is really disheartening too because academia only turns on the "honesty filter" when it comes to minor grad students that pissed off the wrong people. But you can do all this fraud constantly and become president of harvard if you know the right politics.

Dishonest lot. I hope karma is real so they get what is coming to them for taking advantage of people that just love to increase humanity's knowledge.

if you work with a lot of convex optimization, it comes up pretty often. for example, if you learn fenchel conjugates, the lead up and motivation to learning them will often necessitate proving young's inequality with jensen's inequality. that is why learning different maths is cool. you intuit some ways to reshape the problem in order to make these "not super intuitive" connections.

My fav way to intuit this is this example

https://stats.stackexchange.com/questions/85363/simple-examp...

Blew my mind the first time I saw it.

Not the same definitions one to one (author specifically talks about correlation vs linear correlation) but same idea.

What a weird article

House reps and their staffers should sign them their name to it because seeing a government website with very biased political points is odd.

> I doubt we can actually uncover fitness functions that reliably maps to "fun"

You don't have to. Get empirical data and form a proxy evaluator. Usable enough for most evolutionary algorithms. I've done this sort of stuff for very subjective metrics and actually sold something with it.

I agree that a rudimentary evolutionary algorithm probably wouldn't work since they often don't take into account multimodality. There are global stochastic algorithms that take into account this sort of multi modality. Most particle swarm framworks could handle this sort of bipartite nature with a more complicated fitness function. Differential Evolution frameworks in R also allow you to do a post operation on each candidate, which include something like rejecting it because it does not seem to lie in one of the two modes.