No mention of tool use. If the model cannot emit both text and audio at the same time, to enable tools, it’s not really useful at all for voice agents.
This take might have been true years ago, especially for cheaper makes, but modern EVs from china use high-quality components and designs that often surpass european automakers. The premium brands - Nio, Zeekr, Polestar, Lotus, etc - have design and R&D offices in Europe, and source parts from suppliers all over the world. Nio uses Nvidia Orin chips, Qualcomm SoCs, Brembo brakes, Bosch controllers, ZF suspension systems, Continental/Pirelli tires, and ClearMotion (in their flagship model); can't get any better than that.
The driving feel is definitely a thing, the chinese cars are very soft and 'boaty' which is not as desirable elsewhere. They are also on average much larger and heavier than their western counterparts, cities in China have road infrastructure built in the past 20-30 years with spacious lanes.
> have design and R&D offices in Europe, and source parts from suppliers all over the world. Nio uses Nvidia Orin chips, Qualcomm SoCs, Brembo brakes, Bosch controllers, ZF suspension systems, Continental/Pirelli tires, and ClearMotion (in their flagship model); can't get any better than that.
Exactly my point - if premium Chinese cars use Western components are built by Western robots, and designed in Western offices - that means the Chinese technological superiority does not exist.
As for their cars and roads being big - I don't see that as an upside, I have heard from many Chinese residents that the physical size of their cities means (most European cities are denser than Chinese ones) that navigating them becomes that much harder and painful.
Which lays to rest another myth about how China is light-years ahead of the West in urban planning , when in reality they have their own issues.
Yaya. Sorry. I didn't state my assumptions, am mixing issues.
To me, Kia/Hyundai is an example of a foreign manufacturer successively building EVs in North America for North America. Satisfying American standards, tastes, regulations, etc.
I expect the larger Chinese competitors to do the same.
Decades ago, European and then Japanese competitors did as well.
The library is actually ~1400 lines of code, and 51KB in size. Not very slim by JS framework standards, so focusing on that for the headline feels misleading.
It renders terribly on sizes below ~14px, even on a Retina display. No anti-aliasing setting can save it.
A modern typeface like IBM Plex Mono [1] or Office Code Pro [2] that offers more weights and can still render nicely at 10px size would be a better choice, maintaining the same aesthetic.
Yep, on my phone I opened this page and closed it in 20 seconds. Can't see shit, captain! Didn't figure out if it's the font, weight, or size but this is the worst looking website I've hit in a while.
During a war with national mobilization, that would make sense. Or in a country like China. This kind of coercion is not an expected part of democratic rule.
It has always been a part of democratic rule, in peacetime and war. All telco's share virtually all of their technology with the government. Governments in europe and elsewhere routinely requisition services from many of their large corporations. I think it's absurd to think llm's can meaningfully participate in realworld cmd+ctrl systems and the government already has access to ml-enhanced targeting capabilities. I really have no idea what dod normies think of ai, other than that it's infinitely smarter than them, but that's not saying much.
Not the same thing. The parent comment was talking about government “requisiting” services as in forceful compliance, takeovers, not collaboration or regulatory compliance.
If you're referring to telcos sharing their tech with government there are a few examples of Ericsson working with the Swedish military:
> Brigadier-General Mattias Hanson, CIO, Swedish Armed Forces, says: “Strengthening Sweden’s militarily and acting as part of a collective defense requires us to increase our defensive capabilities. We need to utilize the latest technology and all the innovative power of the Swedish private sector. Sweden has unique skills and capabilities in both telecoms and defense technology..." [0]
The question of whether or not the government should be able to use AI for targeting without the involvement of humans is a wartime question, since that is the only time the military should be killing people.
Under such a scenario, requisition applies, and so all of this talk is moot.
The fact that the military is killing people without a declaration of war is the problem, and that's where energy and effort should be directed.
Edit:
There's a yet larger question on whether any legal constraints on the military's use of technology even makes sense at all, since any safeguards will be quickly yielded if a real enemy presents itself. As a course of natural law, no society will willingly handicap its means of defense against an external threat.
It follows then that the only time these ethical concerns apply is when we are the aggressor, which we almost always are. It's the aggression that we should be limiting, not the technology.
You could view various non-proliferation agreements as a legislative constraint on military technology.
Same for chemical and biologicals. Those do prove your point that the law will be ignored if expedient. But it doesn't invalidate the notion of a society putting constraints on itself.
One alternative would be to call the government's bluff: if they truly are as indispensable as they claim then they can leverage that advantage into a deal.
But at a more general level, I'd say that unethical actions do not suddenly become ethical when one's business is at risk. If Anthropic considers that using their technology for X is unethical and then decide that their money and power is worth more than the lives of the foreigners that will be affected by doing X then good for them, but they shouldn't then make a grandstand about how hard they fought to ensure that only foreigners get their necks under the boots.
> so two docker builds can end up with the same hash but different output
The cache key includes the state of the filesystem so I don’t think that would ever be true.
Regardless, the purpose of the tool is to generate [layer] images to be reused, exactly to avoid the pitfalls of reproducible builds, isn’t it? In the context of the article, what makes builds reproducible is the shared cache.
Ah you're right, the hash wouldn't be the same but a Dockerfile could produce different outputs on different machines whereas nix will produce identical output on different machines.
Producing different outputs isn't dockerfile's fault.
Dockerfile doesn't enforce reproducibility but reproducibility can be achieved with it.
Nix isn't some magical thing that makes things reproducible either.
nix is simply pinning build inputs and relying on caches.
nixpkgs is entirely git based so you end up pinning the entire package tree.
If you are building a binary on different arches, it will not be the same. I have many container builds that I can run while disabling the cache and get the same hash/bytes in the end, i.e. reproducible across machines, which also requires whatever you build inside be byte reproducible (like Go)
reply