edran

This is a great update! I hope the authors continue publishing new versions of their plots as the community builds up towards facility gain. It's hard to keep track of all the experiments going on around the world, and normalizing all the results into the same plot space (even wrt. just triple product / Lawson criteria) is actually tricky for various reasons and takes dedicated time.

Somewhat relevant, folks here might also be interested in a whitepaper we recently put up on arXiv that describes what we are doing at Pacific Fusion: https://arxiv.org/abs/2504.10680

Section 1 in particular gives some extra high-level context that might be useful to have while reading Sam and Scott's update, and the rest of the paper should also be a good introduction to the various subsystems that make up a high-yield fusion demonstration system (albeit focused on pulser-driven inertial fusion).

Cool. To be totally honest, I'd feel a lot better if we could just extend robots.txt especially given that we finally have a standard for it with IETF RFC 9309: https://datatracker.ietf.org/doc/html/rfc9309

Also might be worth looking at llms.txt (https://github.com/AnswerDotAI/llms-txt), though it's aimed to be more of an agent-friendly sitemap.xml alternative.

To me it seems fairly unethical for a [tech] company to raise prices without warning customers, especially given the large number of people impacted.

I wonder how the company will respond to the backlash.

I imagine you must be talking primarily about single services, but the question is primarily focused on the sort of necessary infrastructure that is required to build some kind of intranet for developers.

It's not a nightmare, it's a responsibility towards the people who use your product, and the society that allows you to gain wealth from such usage.

If you want to simplify the system, lobby for multiple nations to adopt a good standard, or move on.

The micromanagement task in StarCraft Broodwar (with TorchCraft) that has been used by a few people already in the past year to do some RL research has been tackled basically only with #units > 1, because getting a balanced 1v1 is possible only with certain combination of units, and it's indeed pointless otherwise.

Kiting is maybe a notable exception, but it is also restricted to a relatively small group of unit matchups.

See for instance the following papers:

- https://arxiv.org/abs/1609.02993

- https://arxiv.org/abs/1702.08887

- https://arxiv.org/abs/1703.10069

- https://arxiv.org/abs/1705.08926

Full disclosure: I'm one of such researchers, and I have authored some of the papers on the topic.

@wfunction: yes, TorchCraft includes a serializer that compresses the useful game state into a relatively small struct. That is then further compressed with other tricks and zstd.

More like 20, considering the years in which he passively controlled part of the political narrative.

>It is a very dangerous idea to seek a fairier distribution of wealth by taking it from the rich. If you doubt it go ask Venezuelans or Cubans if it worked out for them.

FYI most countries already do this in the form of somewhat progressive taxation. Some however do this more than the US (say, the great majority of Europe), and they do pretty much fine.

I also don't see how this is "dangerous". For who, the general population? The socio-economical system? The environment?

A lot of research done in Microsoft Research is blue skies research that has nothing to do with MSFT's goals. Hell, MSR for the past decade has set the example for open research in industry-controlled labs, so I would have been surprised if MSFT hadn't already started sponsoring theoretical research in STEM subjects.

(This doesn't justify a lot of the crap that MSFT and other tech corps have done to our industry over the years, but we should give credit where it's due)

We can reasonably assume it's nothing more complicated than what you can do using a combination of machine learning libraries and OpenCV (however if they have instead some new technique, I hope to find a paper from them in a few months :) ).

EDIT: Adding more details.

If you are looking for similar ideas, you should read papers in the area of object-class segmentation / classification[0][1][2], and generic supervised learning.

[0] https://arxiv.org/abs/1510.03727

[1] https://www.microsoft.com/en-us/research/publication/object-...

[2] https://www.ais.uni-bonn.de/papers/DAGM_NC2_2011_Schulz.pdf