I don’t see it as so different. Knuth was playing with bleeding edge tools to see how they could make beautiful documents, I’m sure peers told him plenty of times trying to run vector graphics through a PDP-10 wasn’t “carrying the tradition of…”
Knuth today is busy writing “Shock! Shock! Claude Opus is great at math” papers, not poo-pooing other with strange appeals to “tradition”
Qwen3.8 27B beats Kolibri 79.9 vs 70.8 in German in Kolibri's harness on Kolibri's benchmark.
Also, once the Cohere takeover is complete will they still be able to use this "sovereign" claim despite being 90% owned and 100% operated out of Toronto?
Disclaimer: I am part of the team that trained Kolibri, opinions are mine.
Qwen models have solid performance on benchmarks and we're transparent about this in the report. We're just happy to share a European alternative in the small model space, where I don't think we can afford to be fully dependent on China. We also put a lot of effort into German language quality things that don't show up in evals at all (style, Grammar, German reasoning, etc).
Sovereignty is imho mostly about choice and control over your data. Cohere is no different on that front and personally I'm quite excited about what we will build together post-merger.
Souvereignty is about ownership and about knowing the training data. That is especially important given a business model aiming at the government as a key customer. With Schwarz, Cohere, SAP, NVIDIA and the like as backers society can’t have souvereignty in any meaningful way. Of course it would be a plus to keep US agencies out of the data streams. But this technology will be used to support and make decisions that impact citizens. That being said, incredible achievement.
Well, it's still independent from the USA and China.
The main problems with big corp AI are due to control of access in the first place and control of what they output.
When you make your industry reliant on such choke points, you render yourself the opposite of "sovereign" for sure. Having multiple independent suppliers at least mediates that.
Yeah I think a model has to be actually good to claim sovereignty, as in competitive enough that people want to use it. Chinese and US LLMs are the only ones in these categories right now. Mistral and Cohere have the same problem, yes they are made in different countries but they are not competitive. They (France and Canada in this case) would be better off just downloading Chinese LLMs, even if they get cut off they still have the weights.
I’m most familiar with Canada, where sovereign is usually just an excuse to overpay someone connected for an inferior product with no strategic value.
We have to start somewhere. And I am not sure that you understand what the word sovereign means since it’s entirely disconnected from how “good” something is.
Mistral's medium is absolutely good enough to help with a ton of coding tasks. I've personally used it and found it very helpful. If hypothetically the US cut off the rest of the world from their models Mistral would be a very useful model to continue to use.
Also, it's not like it's fixed in time, they are continually building more powerful models as well. Simply because you're in second place doesn't mean you quit the journey. Though, I'm sure the US has a huge vested interest of convincing people to "just quit and submit."
I was also quite surprised to see that. Considering the effort Aleph Alpha put into to their new model, it seems like the Qwen team needs access to vast amounts of german data o.0
I'm really glad for these efforts for open models from within Europe.
This is visible in Figure 1, but the point they're making there seems to be all about the Pareto frontier: they say Kolibri is something like 3x faster at token generation. That can be a perfectly fine trade-off.
(It's MoE vs dense, I'm sure. But Qwen's own MoE models don't score as well as their dense models.)
This is called a 'plotter', there's a whole community of axi-draw enthusiasts making cool art, for 200 bones you can probably have one on your doorstep tomorrow
It’s not obvious to me that a “nobody wants this” megaproject by a decades old multinational forced to succeed by government mandate wasting over a decade debating national workshare is a great example of a “startup in aerospace”.
There’s tons of startups making cool fused harnesses and phased arrays and trying to get every permutation of diesel and electric and hydrogen engines certified. Drones and rc aircraft spawn many “built in a dorm room” prototypes for people who don’t value their eyebrows
> It’s not obvious to me that a “nobody wants this” megaproject by a decades old multinational forced to succeed by government mandate wasting over a decade debating national workshare is a great example of a “startup in aerospace”.
It looks like I wrote that in a confusing way! My intention was to ask about what features of the engineering discipline was motivating OP's comparison. Maybe a better series of comparisons would have been...
- the 747 – program conception in '63, delivery in '69;
– the A350 - first sales in '05, delivery in '13;
which I think captures the slow pace, and large dependency chain, of any significant program beyond a revision.
We already have 25kV for rail electrification which move 6MW all over the place, surely at boat speeds making some strange double-pantograph system for shipping could be strung up alongside most of the canals, maybe electric tugs could be used as well as normal shipping?
> surely at boat speeds making some strange double-pantograph system for shipping could be strung up alongside most of the canals, maybe electric tugs could be used as well as normal shipping?
I like it, very similar to how mules used to pull barges thru canals except with electrical cables and a rail pantograph running on distribution lines next to the canal instead of mules and ropes.
703 nm laser is almost “loo”, and 417 nm is almost “ool”. Searching those wavelengths gives me #FF0000 and #6400FF. Loo is all red and ool is all violet.
The correct solution is just giving people the option to charge people for phoning them.
Say minimum wage is 12euro/hour, and declining a call takes about a minute of your attention, so this is worth at least 20cent.
Instead of just "Answer" and "Hang-Up", there should be a gov't enforced "charge caller 20cent" option.
This would have no impact on 'real' calls, nobody would charge their Mother and if someone is tacky enough to charge their GP or mechanic they could just be billed for it, and it would hardly be worth someone's trouble to try to steal a phone to call themselves to collect 20c, but it would totally ruin the economics of these spam calls.
There a story, perhaps apochraphal, from 15 years ago of a man setting up a premium rate line and getting paid whenever telemarketers called him. I believe the phone service said anyone doing that would be fined because you need to inform potential callers it's a premium rate number
impressively convoluted. Spam calls are coming from all over the world using voip and prepaid sims. There aren't any accounts to "bill". Not to mention a whole new arena for fraud in spoofing regular numbers for micropayments
To end up on your phone, VOIP have to terminate somewhere in local phone network. These networks need to stop accepting calls from shady places, or every call like this that they cannot charge back to original (or intermediary network) becomes a loss to them. And telecom companies do not like losses...
Not really an HDL, as currently it has no concept of actually connecting the nodes. Each edge is an arbitrary number of conductors with no constraints on routing or timing. Curious to see where this goes though
> Although logic connectivity is not layout-dependent in the current prototype, we envision a future where physical proximity directly influences or determines logical connections. In this ultimate view, the circuit behaves like a digital organism whose physical form and logical connections evolve together under local structural constraints.
Once any HDL is translated to a bunch of primitives (gates, LUTs, flip-flops, etc), that's the job of an automated place-and-route, which already exist. It's independent of the starting language.
Knuth today is busy writing “Shock! Shock! Claude Opus is great at math” papers, not poo-pooing other with strange appeals to “tradition”