I have a hard time believing this, everybody needs money to live and they are his employer.
Then again, I’m a bit of a cynic .
What I tried to articulate before is more that in my opinion we shouldn’t associate ourselves with problematic companies.
Being at the whim of a single company (or source of income) is in general an unhealthy thing regardless of who they are.
For me it isn’t about Anduril specifically, they are just an easy example.
I wish we could find a guideline for all sponsors and large donors.
Which I think the human rights could lead as a foundation.
Maybe as a different example I think Google shouldn’t sponsor us because they violated the right for privacy multiple times for thousands of people and probably multiple other things as well.
In addition the project wouldn’t be FOSS anymore which I think a lot of people would have a problem with.
Re: relicensing — Nix is LGPL, Nixpkgs is MIT. A slow self-EEE is feasible, if enough people, and among them enough people competent in all the critical core parts, all agree it is a good idea. However, «we prefer free software or opensource software» was an uncontroversial part of the values, so I would expect that building that kind of an agreement will not be easy.
Yep, totally your right to doubt. I originally wrote “which I trust people can decide for themselves whether or not to believe, and whether they want stronger guardrails that don’t require belief to be upheld”. Ultimately, I decided that if I have to state trust in self-determination then I’ve already failed to show such trust, although now I’m wondering if I should have been more explicit.
Once again, I think we are in agreement in words. I certainly don’t think anyone is going to sign their name to something that says that NixOS should associate with problematic companies.
However, I feel sure that quite a few people have different interpretations of what “not to associate with” and “problematic” mean. Some people clearly believe that tomberek serving on the SC after his hiring is association; some people clearly do not. Some people think the project has already created distance by dropping a conference sponsorship and with the SC banning specifically Anduril from posting job listings here, while others clearly think that is not enough or are uncomfortable with the idea that these were even an option. Quite a lot of people find Anduril problematic; just as clearly there are people who are arguing otherwise.
I don’t think it’s unfair to recognize that there is feeling of tension between the policy decisions actually made regarding Anduril and people very ardently arguing that they perceive Nix to be in crisis due to potential Anduril takeover. (Before, yet again, people read something into this something that I haven’t written - I am not arguing either way on the policy decisions.)
My point, as it was in the original reply to your message, is simply that I don’t think anyone is seriously arguing that Nix should be catering to Anduril. When it isn’t functionally a strawman, I think it boils down to a difference in opinion in what acts would constitute Anduril or a different company or the MIC controlling the direction of Nix.
I’m quite open to the OP here; I like that evidence was linked and consolidated, and I think that the people who will be swayed will be. We can always try to persuade. I think we all want to be “good”, whatever that means to each individual. I think it’s more persuasive when we denounce acts instead of people, while also recognizing that eventually enough acts do define a person. When certain things are irreconcilable, I think what remains is simply community voting.
The way you phrase this tells me you’re picking up on the arguments that are taking shape. I can promise you that while those peddling hype etc will make wild claims and extrapolations, there is a serious core generating the foundation arguments.
A whole range of tough engineering problems like synthetic biochem are bottlenecked on insufficiently strong heuristics. We cannot automate the exploration of high-dimensional spaces and suggest where to run experiments. The periodic table is big. The structure of chemical bonds adds a lot of complexity on top. More labs and more people running experiments can’t really touch these problems due to the big O.
What LLMs demonstrated was that we’re already practically capable of generating novel, sometimes consistent formalisms, which are readily detectable through automated theorem provers. Work to integrate the two approaches, natural and formal, is underway in lots of places on Earth. The result will undercut the current compute capex cycle and consequent power demand because formal systems can be directly computed rather than inferred from multiplying giant matrices. It also will greatly assist in these heuristic exploration problems, providing automation useful to both theoretical and experimental research.
So what do we do with that? Biochemistry can pull off reactions that are hard using more traditional metal catalysts. Ripping carbon off of oxygen is one. Whereas algae do this to survive and turn lots of other wheels, a synthetic system can do it because we want the carbon. We might want the carbon to generate fibers like cotton without the land use. We might want concrete, steel, and plastic analogs that are closed loop on carbon, cheaper, higher performance, and less problematic for the environment.
This is all a very niche set of views because very few people know what formal systems are or how organic chemistry works, so it’s not going to percolate into pop science and politics without some revolution in public education. Since we all should know a bit of big O and formal vs natural, only a bit more digging will reveal the real work going on and that it’s a matter of doing the work rather than rolling the dice. Eric Schmidt may have to ask retail investors to “trust us” but grad students and faculty in universities around the world have independently deduced these conclusions in selection of their research.
AFAIK, it doesn’ matter what the original license is, it only matters whether the copyright holders agree to a new license, at least for existing code contributed under the old license. I’m generalizing, lawyers will need to advise for each country obviously. But I know this is the case in the US at least.
I don’t know what a self-EEE is, but you could change the license for new code contributions. Then you’d need to dual license and track which lines of code in each file were contributed under the old license, and then once they all disappear, the codebase could be considered fully transitioned to the new license. That event would likely not happen for many years.
But you could make it difficult for the people you want to prevent from using nixpkgs by proactively adding a bit of code to each file in the repo that was licensensed under the more restrictive license, essentially poisoning it. Some code in each file would be Newlicensed so they’d then practically be required to fork from a fully MIT snapshot of the repo, at least if they cared about legality.
But if you’re not going to actively police niixpkgs usage, all of this would be for naught. It might work in the US to prevent Anduril from using Nix, but it would be useless for similar companies in, say, China. I suspect most people would be against it just due to its long term PITA factor. And doing this before the board owns the trademark for the name Nix in countries they want to market in, would be pretty short sighted, because someone could just call any fork “Nix” anyway. This is IMO a much more pressing issue.
if i understand correctly, you argue computation can be part of the solution, rather than the problem. this would make supply catch up to demand, rendering innovation the way to tackle climate change. this is the classic liberal perspective on this.
such demand itself seems a function of supply tho, making this chasing shadows. the likes of big slop consider it in their incentive to throw as much compute at things as possible, emissions be damned. given new limits, they’ll be pushing those new limits, with regulators unable to keep up (orworse).
capitalist market incentive create principal-agent conflict on externalities.
to solve the externalities, the incentives thus need realigning to public interest. such accountability implies democratizing emitters, taking away their capitalist market incentive.
Sorry, what are you talking about? None of this makes any sense, mathematically. What formalism is LLMs coming up with? What does “detection by automated theorem provers” mean? What does “due to the big O” mean? Do you want to say that the time complexity of current algorithms is too large to practically compute? Then say that, O(1) is a “big O”, and it’s really not hard to compute things from there.
In an earlier post you link to a bunch of math wiki pages when arguing a point, and most don’t relate to the points at all beyond a superficial connection. Are you just posting word slop here to derail the topic while sounding smart?
Decarbonizing society undergoes has two main legs:
Decarbonize transportation
Decarbonize electricity
Assuming we keep modern material society, this is true — it applies to both “industrial capitalism”, “industrial communism”, and other hypothetical ways of organizing industrialized societies.
(Primitivism is a way out, but I know of no serious plan for how to compel the entire world to return to hit Ctrl-Z on modernity, other than just waiting for nuclear war. And I am quite sure that waiting for nuclear war is slower than other solutions, meaning that by the time Primitivism might succeed, there will be much more damage to the biosphere, and more much bad feedback loops that will take a long time to “burn out” causing much more environmental damage on an ongoing basis post-apocalypse too.)
The actual technocratic details of decarbonizing transportation and electricity are very interesting. Electricity is the easier one. You can keep a conventional grid and build Nuclear, but that is slow and expensive and hard to scale up (even for China, though I wish them luck overcoming that!). Or you can do far cheaper and more incremental Wind and Solar, but then you need to thoroughly reimagine the grid and build a bunch of storage too — a “hidden cost” compared to the very cheap capex of installing wind and solar itself.
There is a bunch of interesting literature about the design of electricity markets too, and how wind and solar run rough-shed over current designs. This is very interesting stuff — so much pop literature assumes that either markets are natural or that markets are evil. The middle ground that markets neither upstream or downstream from other “institutions” and can can be “dungeon mastered” — semi-planned, in a way that both strongly effects outcomes yet doesn’t leave market actors feeling they have no autonomy — is not discussed enough.
I am very interested in grid upgrade problems because it it reminds of Nixpkgs. They most people assume that “writing the software is the hard/expensive part” and coordinating “how the software is put together” is the easy/cheap part, reminds me how most people assume we just need solar panels and electric cars, and grid/storage will take car of itself. In both case, it is an error in too much atomistic and not enough holistic thinking.
(If any Nixpkgs person was or will be getting into grid management / electricity public policy, I would love to hear whether you also think this analogy rings true!)
‘Contributing back’ of course can be a slippery slope. Having any connection whatsoever with any such company can be detrimental in many unforeseeable ways and long term too. Dan Koditschek’s letter to Ghost Robotics is a good case in point. For one, Koditschek felt that his “formerly proud connection” to that company turned out to impact his reputation. So much so that he had to ask for the removal of any reference to him or the University of Pennsylvania from the company’s website, and make the letter public. As to whether his reputation has been restored afterwards, I very much doubt so. It may not help to know that Prof. Koditschek’s has been supported by the DoD for decades of course, has he himself admits. That his “sponsors and potential military users have always stressed their intention to use the technology to save lives” is disingenuous to say the least. Bottom line, he did profit, for decades, from his association with DoD programs, the same DoD that has killed and maimed around the world for decades.
As to the open source question, not being easy doesn’t mean unfeasible. I can see a serious discussion on the merits and demerits of open source to be inherently beneficial, and that can be done productively by discussing how to transition and implement a new license as @chrism@bme have demonstrated above.
On the same subject of open source and military applications, it is worth pondering on how unrestricted use bears “the potential to substantially improve software outcomes for the DoD (…) and at a lower total cost than the proprietary and closed systems the Department currently uses.” This much is said in an article published by the Center for a New American Security almost one decade ago. Now, shouldn’t we all make a good effort to make sure that our collective effort is put to good use? Otherwise, I wouldn’t be surprised to hear sometime in the future about how “wonderful intentions and achievements” were put to bad use. How disingenuous would that be?
For the record, I think relicensing this way would be a mistake to even think about before gaining the trademark. Not having a trademark for the brand is a gaping vulnerability right now, and when you start to make powerful enemies, it will be exploited.
Note that my position is, implementation of de-facto relicensing is easy, convincing (the currently active, it is not a reachability question) key people to actually go along with it is unlikely, so, discussing the finer points of technical transition is not moving forward on the hard part.
At least in the US, you need to file for trademark early and you need to use and defend your trademark, Nix has done neither of those, so I’m not confident they’d be grated the trademark in the first place, it is way too late.
.