There is actually a whole lot of research around the "use less data" called data pruning. The goal in a lot of cases there is basically to achieve the same performance with less data. For example [1] received quite some attention in the past.
I clarified my comment - "perhaps researchers have not tried 'use less data'" suggests I might be unaware of this concept, I changed it to "as if". In fact "less data" was tried for decades before the first image classifiers were actually working in 2012. My understanding of that paper you are linking to is that it is not a new research paradigm; it is about filtering/pruning less relevant data that is not needed to improve a particular capability in a deep learning model, and that is absolutely one likely approach that will yield the goal of smaller, better models in many tasks.
That will not change the fact that a coding model has to learn vastly many foundational capabilities that will not be present in such a dataset as small as all the python code ever written. It will mean much less python than all the python ever written will be needed, but many other things needed too in representative quantities.
Right, but that's a short term moat. If they pause on their incredible levels of spending for even 6 months, someone else will take over having spent only a tiny fraction of what they did. They might get taken over anyway.
By reverse engineering, sheer stupidity from the competition, corporate espionage, ‘stealing’ engineers and sometimes a stroke of genius, the same as it’s always been
My website of choice whenever I have to deal with references is dblp [1]. In my opinion more reliable than Google scholar in creating correct BibTeX. Also when searching for a paper you clearly see where it has been published or if it is only on arxiv.
As an American, I'm going to need this as a volume, either in terms of Olympic-sized swimming pools or the height of a pile in an [American] football stadium. Maybe I'd accept weight as a quantity of Ford F-150s, but you'd be pushing it.
I also like to take potshots at Americans, but come on. It's unlikely that a newspaper called "the berliner" in a article about Berlin included this line specifically thinking about citizens of a far-away foreign country who don't use metric units that often.
Occam's razor says that it's actually one of our noble and enlightened European journalists who made that sloppy remark without realising it.
Well, the EU is pretty firmly in conservative hands. Ask a random EU citizen if they know that EPP is secretly leading the EU since its conception? They will likely not realize, because they fall for the piece&love propaganda. Just look at what VDL has done since she overtook the lead? Know where she comes from? Used to be defense minister in germany. Was called "Flintenuschi" back then. And now, magically, we are supposed to invest a shitload of money into military. Thats what I call a wolf in sheep's clothing.
What exactly makes you assume that the persons arguing for open source here are not the same people who has helped us defeat earlier attempts to make chat control happen?
I was trying to remember where I had heard of him. He was the initial developer of RCS, and was also involved with Extreme Programming (XP). XP later got mangled into the Agile movement and it wasn't perfect to begin with, but it was refreshing.
I think it is quite high, especially for newcomers to the ecosystem. The popular sdkman CLI tool for managing JDK installations provides 17 different JDKs to install:
With the DotNet SDK for comparisons sake, there is but one provider, and package management is provided as first class citizen in the compiler CLI, removing the need to even pick a "gradle" or "maven" style build tool in the first place for almost any project.
It doesn’t negate the issue at all, especially as sdkman is an optional extra tool. If working with others, that default is not always going to be the one your team develops against either. If you are just trying to pick a jdk and don’t even know sdkman exists, a sensible person will rightly wonder why the ecosystem has fractured into 17 JDKs, or you won’t notice at all and blindly pick a download from the oracle website. I’ve seen it cause confusion many times for new (and old!) Java developers.
Checking a new developer has actually managed to install the correct JDK for a given project is still a thing I have to do, all the time.
Sdkmans docs also have no bearing on the multiple build systems in the java world too, which again simply a non-issue in DotNet and many others.
There is no market for such a thing. At that price point, you get a personal chauffeur. That’s what rich people do and he can do stuff that a self driving system never can.
And the rich people who don't want a chauffeur like driving the car. They will buy a $10M car no problem, but they want driving that car to be fun because that's what they were paying for. They don't want you to make the driving more automatic and less interesting.
[1] https://arxiv.org/abs/2206.14486
reply