rjurney

So feed it to kids!

So long as its safe. Almost always, these things don't pan out... or at least the first molecular iteration. That's the primary reason medicine is so expensive.

There are many ways of approaching quantitative trading and many people do employ time series analysis, especially for high frequency trading.

They aren’t so hot, but recent efforts at transfer learning were promising.

And NVIDIA is playing in that game too. Why would they not play in higher level services as well? They already publish the source to their entire software stack. A comparison to Android is completely useless. Google is a multi-sided platform that does lots of things for free for some people (web users, Android users) so it can charge other people for their data (ad buyers). That isn't the chip business whatsoever. The original comment only makes sense if you know nothing about their respective business models.

Android is a complement to Google's business, which is when open source works. What would be the complement worth $1 Trillion to NVIDIA to build a truly open platform? There isn't one. That was his point.

Claude is much better. Overwhelmingly better. It not only implements deep learning models for me, it has great suggestions on evolving them to actually work.

That are better for someone else.

Cost cutting is usually driven by pricing pressure, and cheaper products are better.

I believe this is in dispute among sources.

But I JUST switched from GPT4o to Claude! :( Kidding, but it isn't clear how to use this thing, as others have pointed out.

Herodotus defined the singularity as occurring during the reign of Cyrus and concluding during the reign of Artaxerxes.

If a product isn't better for someone, people don't buy it. Marketing counts for something, but quality is still the most important factor in the success of a product. It just may not be the quality axis that's important to you, so it can seem arbitrary.

Not having a layer like this has left AMD completely out of the AI game that has made NVDA the world's most valuable company.

If it's efficient, this is very good for competition.

I love articles that completely dispel the idea of singularity - that point out that the feeling of our age that things are changing too quickly has been around for a very long time.

But I thought ULID will work in a UUID field in Postgres? Will it in other DBs?

Lots of people use both VSCode and native Jupyter notebooks that would love this app as it develops.

Don't listen to the grinches that stole Swift. Electron is fine. We have lots of RAM. Just make it good. Data scientists won't care.

I've started using python-ulid as ULIDs seem superior to UUIDs. https://pypi.org/project/python-ulid/

You should try it sometime. You might fall in love with it.

God, I love Processing. Most beautiful programs I’ve ever written were in Processing. I had this plan to recover from atypical depression by visualizing all of basic linear algebra using Processing but… I couldn’t program a computer for some time. Doh!

This is especially true for a data scientist, where most code is throwaway. If you make it all spectacular, you aren't getting anything done. Data scientists' code should be "eventually good," that is to say it gets refactored as it approaches a production environment. I talk about this in my last book, Agile Data Science 2.0 (Amazon 4.1 stars 7 years after publishing).

https://www.amazon.com/Agile-Data-Science-2-0-Applications/d...

I will say that after 20 years of working as a software engineer, data engineer, data scientist and ML engineer, I can write pretty clean Python all the time but this isn't common.

It is time to burn traditional academic publishing to the ground and rebuild it from the ground up, or else let's rebel against the government, give Alexandra Elbakyan citizenship and put her in charge of the Department of Energy to make all computer science research open domain for all the world to see. Requesting $48 for to read a paper funded by your taxes before you even know it is useful to your work will result in deportation or a firing squad, whichever a panel of independent reviewers decides.

A less spirited method is to let traditional publishing slowly die by adding review features to highly viable alternatives like arXiv, PubMed and others which will then replace it. I hate academic publishers, they are vultures that serve no purpose. I hope they all go out of business or somehow pay big :)

There is a chain. There is legal recourse. And there are considerations in government IT that you would not believe and they are incredibly difficult to deal with on minimal resources. It has to be harder than the private sector and this application isn't any different than buggy mainframe software run by major banks. It sits and gets crufty.

Reminds me of Russia. There's your release date... and then there's your court date. You aren't released until the court date after your release date, and without a bribe that could be years. You might never get out.

I haven't been there in a while, but this is how it was when I lived in Moscow.

I wonder if anyone here gives a shit about this?

I think its a no brainer and possible doom scenario. Is it possible? There are common operations like... the L1L2 norm that are used across machine learning. Is there a way to craft an operation like a regularizer that is both useful and profitable? Or could up you use a block as an initialization parameter and somehow make use of it? A prediction task itself might work - solve this provably hard problem and you own the cash. I don’t know but it seems like what is needed.

So is EDI used by banks with other currencies. I am old enough to have used it :)

What is needed is an AI digital currency where training deep networks for representation learning has a side effect that mints coins without costing a lot of extra resources. Then mining coins is useful and we get more better AI to enslave humanity.