jasonmorton

This lets you verify the signature on the model. It won’t help you tell that a decision came from that model. If you want to verify the inference that a model makes, check out https://github.com/zkonduit/ezkl (our project).

You can ZK an XGBoost or neural net with https://github.com/zkonduit/ezkl

Here is the twitter thread providing an introduction and some background: https://twitter.com/ezklxyz/status/1712940546182774936

Checking the work takes the same resources as doing the work. As much as this would be wonderful in a universe where it's possible, it's simply not possible.

We do live in that universe, under some currently believed assumptions. An NP-complete problem is an example of something where checking a solution is (thought to be) easier than finding one.

Zero-knowledge proofs make it such that checking that a computation (such as inference) has been done correctly is easier than doing the computation (even keeping some parts private). A great reference is here: https://people.cs.georgetown.edu/jthaler/ProofsArgsAndZK.pdf

One can cryptographically prove the correct inference of a small AI model now with https://github.com/zkonduit/ezkl (our open-source package).

Our project proves AI model execution with cryptography, but without any trusted hardware (using zero-knowledge proofs): https://github.com/zkonduit/ezkl

There are definitely certain scenarios, especially around verifiable compute, in which you can think of ZK as a software alternative to a secure enclave.

Interestingly, if you ask people which they would trust for a rollup, most say ZK is the more trustworthy technology. In part just because it doesn't depend on one vendor (the enclave manufacturer).

Here are two good places to start:

http://learn.0xparc.org/

https://www.youtube.com/playlist?list=PLj80z0cJm8QErn3akRcqv...

Examples include the various zkEVM projects, risc0, and slightly more limited stuff like our effort to compile PyTorch/ONNX to a zk circuit (https://github.com/zkonduit/ezkl).

Speed and memory usage is improving quickly.

Not OP, but for me adding or deleting a character in the middle of a paragraph in a latex doc hangs for maybe .5s or more, often. This is a fairly new phenomenon, never happened 2+ years ago.

I've been using emacs for many years now, and I'm sad to say that it has never been slower, despite the hardware getting better. Now it lags like crazy when editing a latex document, no matter the recent version. Something went off the rails in the last year or so, and even Atom is more responsive.

Many ads are not clicked on by a real person, but a bot, or someone clicking ads as fast as possible because it is their job.

Many comments would be considered "fraudulent" in the same sense, in that they were not authored by a real person, but a bot, or someone writing comments as fast as possible because it is their job.

We've been working with Modulo as well; the experience is excellent and is constantly adapting to make it work for our kids, rather than imposing a formula. They enjoy having some autonomy in their education. It's a very nice community as well, and they have the chance to make friends all over the country.

Thanks, looks like this is: https://github.com/jitsi/jitsi-meet

Can anyone recommend a good open-source chat and call solution? I'm looking for something I can run on my own server to support friends and family, with an iOS and Android app. It doesn't have to support more than one or two video calls at a time. I figure this is sadly the long term solution.

There are some high-quality open access journals (e.g. https://www.emis.de/journals/SIGMA/) that don't charge authors. They use arxiv overlay. It is entirely possible.

Copyright assignments made by authors can be cancelled [1]. For papers and books from before 1994 or so (25 years from present), you can write a letter to the publisher cancelling your grant or assignment of copyright to them. The rules are complex so I might be missing some details. But there are probably a lot of researchers who could take advantage of this to put on some pressure and reclaim their work.

[1] https://www.copyright.gov/docs/203.html

Here is a proven approach for at least the first part, building foundations and being ready for graduate work. Many Berkeley Ph.D. students passed through this route. Get the book "Berkeley Problems in Mathematics." It contains historical problems from the Berkeley math prelim exam, and solutions. Now don't look at any solutions yet.

This is the exam all Berkeley math Ph.D. students must pass within three semesters of arriving to stay in the program, and the fail rate is about 50%.

You will also need reference books, advanced undergraduate and beginning graduate textbooks. Buy, download, or borrow as appropriate.

Pick a problem (start with the older ones, they are easier). Set aside 30-60 mins and try to solve it. No devices, no references at all, go to a library or a coffee shop without your devices. Dont' give up till time is over. If you cannot (usually the case), still don't look at the answer. Hit the reference books (don't look up the problem online either, it will go right to the answer and you won't learn much). Read and try to understand enough so that you can solve the problem. It is ok if you solve it this way (in the course of reading about it).

For bonus points, students studying for the exam will typically take entire old exams (available from the Berkeley website), take that to the library and just sit down for three to six hours and try to solve all the problems correctly. Then self-grade harshly. When you can do that for a recent exam (and get a good score), you will have more or less mastered undergrad math to the point that you could teach it.

Most important: you have to struggle to solve problems. Reading a solution is about as useful as watching someone else lift weights: you get minor tips on form but not any stronger.

12.15 on p 167 of http://cdn.batstrading.com/resources/regulation/rule_book/BA...

Yes, it seems like intent is important, but in some cases (e.g. the BATS rule), no intent is needed. There are statistical properties that indicate intent, but it seems like most of the criminal cases involve finding an email or text message like "I found a bot" or someone telling an engineer to build a system with a manipulative intent.

Here https://drive.google.com/file/d/0Bzu6iGPza2s9amFURTFzenRWVGg... are some slides from a talk I gave last year at Strata about what spoofing and layering is, and how a trading firm or exchange might detect or prevent them.

Interesting, why is that? Hedge fund returns certainly don't suffer from light speed issues (although missing data can be a problem).

Thank you, I will try these. I was looking at your repo but couldn't tell if it would work remotely.

Anyone know if there is any terminal + server side piece that, like terminology or iterm2, lets me inline plot in R, python over ssh? Linux or mac.

Imagine a distant future where advances in technology makes everyone fantastically wealthy in modern terms. Material goods are as cheap as digital storage feels now. What would spend money on in this utopia? My guess would be extending our time and quality of life (health care) and improving our minds (education).

Right - would you rather rely on a theorem which the author claimed is true, but won't let you see the proof, or one where the proof is published and reviewed? That's the choice between closed and open source math software.

Among the interesting things about Ethereum is that they built a billion-dollar "company" in a totally nontraditional way. Instead of VC, they raised money in a presale, the developers have liquidity from the beginning, and none of it required permission from anyone. Just people buying into the concept. Seems clear that this sort of thing being possible stands to change things for regulators, investors, and other market participants.

This is almost exactly the singularity-igniting innovation described in the sci fi book "Avogadro Corp" by William Hertling.

I think this is absolutely the way forward for technical courses (where the prof and maybe students already know how to use git, or can be expected to learn).

A course is kind of like a web app with a lot of common functionality. Too bad most LMSs seem to think we want to write our course in a little text box with no version control (even newer ones like Canvas make this assumption).

There are several arxiv overlay journals. Example: http://www.emis.de/journals/SIGMA/about.html