HN Theater @HNTheaterMonth

Here are few thoughts on Aaranson's comments:

1) I have absolutely zero understanding of quantum computers (even after watching this video: https://www.youtube.com/watch?v=OWJCfOvochA), though I like to read up on quantum physics. However, it seems to me that I've lived through many "breakthroughs" in this technology. For example, here's an excerpt from a NYT article from 4/28/1998: https://www.nytimes.com/1998/04/28/science/quantum-computing...:

  The discovery has touched off a wave of excitement among physicists and computer scientists and is leading dozens of research centers worldwide to embark on similar experiments heralding the advent of an era of so-called quantum computers, specialized machines that may one day prove thousands or even millions of times faster than today's most powerful supercomputers.

It seems quantum computers that can kick ass is always 10-15 years away.

2) Here's a writeup from Pano Kanelos, President of the University of Austin of University of Austin, where he explains the motivation: https://bariweiss.substack.com/p/we-cant-wait-for-universiti.... Others can debate the anti-wokeness, but this part is pretty much fact:

  Over the last three decades, the cost of a degree from a four-year private college has nearly doubled; the cost of a degree from a public university has nearly tripled.

Colleges in the US are insanely expensive! If the new UoA starts fixing that I'm all of it, but I'll believe it when I see their tuition rates and first batch of graduates. This is the first question listed here (along with the high cost of healthcare in the US): https://patrickcollison.com/questions

3) The blog post he mentions is interesting. Of course, the same argument may be made for any job area for which demographic stats does not match the society at large. The argument is irrefutable because any point against it is labeled as offensive or biased, which makes it funny coming from a mathematician.

⬐ romwell

>The blog post he mentions is interesting

And he misses the point of the blog post by a mile.

⬐ ryan93

Wanna explain?

⬐ romwell

Well, my comment was flagged, so nobody would see my explanation.

I would recommend reading Piper's post and the follow-up to it, and then see if a "mass call to resign" was really the point.

One of the IBM scientists seems to have an old quantum computer as a piece of office decoration, so I guess there's some hope to preserve them: https://www.youtube.com/watch?v=OWJCfOvochA

I’ve begun to put together useful resources for learning more here: https://quantumjavascript.app/resources.html

I’d like to really flesh out that page; I’ve now come across so many folks that are amazing at explaining this stuff... It’s just a time crunch to juggle all this as you might imagine ;)

These two videos in particular helped me get started:

1. What is a physical quantum computer? https://www.youtube.com/watch?v=OWJCfOvochA

2. How might one compute using it? https://www.youtube.com/watch?v=F_Riqjdh2oM

"sparsity is a useful measure of interpretability, since humans can handle at most 7±2 cognitive entities at once "

If you are going to limit models to 9 features, or 9 combinations of features, or 9 rules, models are not going to work as well. This just seems like a very weak argument to introduce.

"A black box model is either a function that is too complicated for any human to comprehend, or a function that is proprietary" These seem like two very different things. A false equivalence is then made with many arguments against proprietary models being made as if they apply to complex models.

"It is a myth that there is necessarily a trade-off between accuracy and interpretability....However, this is often not true" It's a bit of a straw man to suggest that there is "necessarily" a trade-off, as there are surely cases where there is not. One could say that there is "often" a trade-off between accuracy and interpretability. I don't think there are many people out there with the naive view that one should never do feature engineering.

"If the explanation was completely faithful to what the original model computes, the explanation would equal the original model" This is just nonsensical to me. The idea is not to be completely faithful, but to raise things up to another level of abstraction. The series of videos from Wired comes to mind where a concept is explained at multiple levels. https://www.youtube.com/watch?v=OWJCfOvochA

"Black box models are often not compatible in situations where information outside the database needs to be combined with a risk assessment." This is absolutely untrue, depending on one's definition of often. The output of a model, whether it falls into this incorrect definition of a model or not, can be treated as a feature in another model. Ensemble learning exists.

COMPAS is the punching bag, but no one seems to know what it is. I haven't seen the evidence that its performance is equal to three if statements. It certainly doesn't have anything to say about machine learning in general, as it is set of expert designed rules. So, it is actually the kind of algorithm the author favors, except proprietary.

"typographical errors seem to be common in computing COMPAS.... This, unfortunately, is a drawback of using black box models" Unclear why typographical errors only affect black box models.

The BreezoMeter case is not clear evidence of anything broader either. It is unclear whether the one error noted out of millions of predictions is from bad source data. Stretching this to concern any sort of proprietary prediction, such as mortgage ratings, is a stretch that doesn't really tell us anything.

"Solving constrained problems is generally harder than solving unconstrained problems." This doesn't make sense to me at all. All evidence is that ML works better on constrained problems.

The idea that CORELS is somehow better, even if it comes up with a ruleset of millions of rules, doesn't make sense. The proposed workaround for this "the model would contain an additional term only if this additional term reduced the error by at least 1%" could result in the failure to create a model if no one term provided 1% on its own.

Scoring systems are useful, but it's like a single layer perceptron in the example provided. You need to consider combinations of factors to see their impact. A high X is bad, unless Y is also high and Z is is low.

The fundamental problem here is trying to limit the power of the algorithm to the power of the human mind. From the very beginning we have used computers to do things that are difficult or impossible for us to do. In some cases the answers were provably correct, but we have now reached a point where computer generated proofs are accepted. The "oracle" mode of computing will compute an answer for us, but we ultimately have to choose whether or not to accept it on the basis of the evidence, much like we do with the opinion of an expert. Simple techniques such as providing one's own test data set, and the kind of analysis done by Tetlock around Superforecasters, can go a long way to building that understanding of accuracy of predictions, such that we have a guideline for evaluating algorithms that are beyond our ability to understand.

⬐ buckminster

With an interpretable model typographical errors are obvious in the result. For example, if the system denies bail because you have four convictions, but you actually don't, then the problem is obvious. If the system denies bail with no interpretation then the typographical error goes unnoticed.

⬐ mattmcknight

I guess I don't see that part. If the typo is in the number of convictions, wouldn't an interpretable model also be subject to that typo? An interpretable model would only consider number of convictions as one of the factors. So if you look at a model like one of the scoring models shown and there are 20-30 factors under consideration, the impact would not be any more apparent than it would be from reviewing the input data. Like if it said a person has zero convictions and allowed bail, but they had four convictions, it wouldn't be obvious from the result that there was a typo somewhere.

⬐ sanchezdev

I appreciate this thoughtful and detailed reply. I was thinking all those things in my head while reading as well, but couldn't bring myself to invest the time to address them all. I got the impression that the author wasn't someone who has a lot of experience building real-world predictive models otherwise they'd appreciate the trade-offs that need to be made sometimes to get something that works well and can be debugged/interpreted without too much trouble. Of course this isn't to say we shouldn't be striving to develop more interpretable solutions, but I don't think this paper is very helpful to due to its lack of rigor and straw-man tactics.

I found this video to be a great explanation https://www.youtube.com/watch?v=OWJCfOvochA

Explains to 5 different levels of CS knowledge.

Try this video - https://www.youtube.com/watch?v=OWJCfOvochA

I found it easy to comprehend. It starts explaining to a child & works its way up to a doctorate level.

Here's another great job explaining quantum computing starting at a basic level & working up - https://www.youtube.com/watch?v=OWJCfOvochA

What constitutes reasonable?

Here's an example of someone explaining something to five different levels of audiences. https://www.youtube.com/watch?v=OWJCfOvochA

None of these approaches are prima facie "unreasonable." But if you get the wrong level it won't feel useful.

I don't think it's crazy to have open teaching discussions on HN, I think you lay out a fair hope that someone can break something down here. But I'm just not sure which level you're looking for.

Precision and simplicity are often competing constraints.

⬐ chvid

Okay.

Let me try getting my point across with an insult instead.

Some dude in a marginal university in Norway has managed to convince a few people, a couple of students and a writer for a glittery university magazine, that he is a genius.

His approach is to do something trivial but bury it in a completely obscure terminology. This technically sorta works on a toy example but not really on more challenging tests. But it is very convincing for people who are not trying to understand what is going on but rather are impressed by technical words and academic rank.

That is why it is getting negative feedback here and on other technical forums.

And that is why noone can answer a simple technical question.

⬐ coldtea

Here's the code, check it out:

https://github.com/cair/TsetlinMachine

>And that is why noone can answer a simple technical question.

Who is "noone"? Some people in a small thread on HN that first read about this today and most of which do not have anything to do with NN or Tsetlin's automata (e.g. startup founders, JS programmers, embedded programmers, and so on)?

Note also that you haven't made any "simple technical question" -- not a specific one that is. Just asked for it to be explained to you, and then rudely complained again when someone did that it wasn't "reasonably" explained.

You also seem impatient to wait for one (considering that you posted 4 hours ago, and most of US is sleeping still), it has to be pronto for you it seems...

Here's also a simple explanation from another member:

"The Tsetlin machine is based on Tsetlin’s automata, the so called finite automata with linear tactics. The are basically counters of rewards (up) and penalties (down) with various strategies of integration, proportionization and differentiation. They have minimum necessary for digital emulation of the perception functionality. They str not only due to Tsetlin but also yo Krinsky, Krylov, Varshavsky. Mikhail Tsetlin is the most prominent creator. It’s the work in the USSR in the 60s. So not new but very elegant!"

>His approach is to do something trivial but bury it in a completely obscure terminology. This technically sorta works on a toy example but not really on more challenging tests. But it is very convincing for people who are not trying to understand what is going on but rather are impressed by technical words and academic rank. That is why it is getting negative feedback here and on other technical forums.

Possibly. Also possible that all the negative feedback are from people who don't understand the math either, and/or are too invested in traditional NN, and is the usual resistance to new ideas.

Apparently you don't have the means to classify it to one or the other case, but you do want to be rude and make a judgement nonetheless.

⬐ isoprophlex

Dude. Take a break.

Your problem is your own inability or unwillingness to set aside a few moments to read the - in my opinion reasonably clear - definitions in the linked articles. Don't take it out on others.

You are not entitled to an easily grokkable explanation, tailor made for your exact level of knowledge.

⬐ majewsky

> Let me try getting my point across with an insult instead.

As a rule of thumb, when you write such a sentence, just stop and go outside for a while.

⬐ chvid

Sorry about my choice of wording; I was just annoyed with the bold claims and the apparent lack of definition of this specifc concept.