I agree. I am a paying Le Chat Pro user, really rooting for a European alternative. But the quality difference between Mistral and the frontier labs is growing too big to ignore. It’s worrying to me that they didn’t talk much about new models at the conference, because that is really where their focus should be IMHO.
I am wondering what is keeping them back, though: Money? Compute? Skills? Training data? My fear is that you are really only getting really good models by training on very dubious data (outputs from the frontier models etc) and that Mistral is too European and too enterprisey to take those risks.
My theory with no insider information: it’s a little of all of the above, but mostly money. To some extent, you can dig yourself out of a data hole with RL and a lot of compute. And you can buy a lot of compute and some data with a lot of money. Big labs have been operating in this regime for a while and it’s one of the drivers behind their costs beyond just scaling the weights and doing the actual training. Mistral just doesn’t have access to this level of compute or the money to try and muscle their way in.
I wouldn't be surprised if each of the frontier American labs and individually has compute access similar to the entire EU. Chinese firms are a more interesting comparison since there are a fair amount of great models there, and it's estimated about 15% of the ai relevant compute is in China versus maybe 5% in the EU under European companies (and 70% ish in the US is the most common ballpark I see)
I've never heard or read anything about the EU planning on investing money in Mistral. They're a private company. They're French. It honestly sounds kind of absurd.
It's a bit strange, but a huge handout from the EU/France and a huge AI lab investment round are different orders of magnitude. The necessary sums are just not politically possible. How do you sell spending the equivalent of ten USS Gerald Fords on a start-up? You don't.
> I am wondering what is keeping them back, though: Money? Compute? Skills? Training data?
Not ruthless enough and no backing by a corrupt govt administration that has no morals but focuses on self-enrichment instead.
Might sound drastic but I think that's actually closer to the truth thn everbody likes to admit.
> My fear is that you are really only getting really good models by training on very dubious data (outputs from the frontier models etc) and that Mistral is too European and too enterprisey to take those risks.
> what is keeping them back, though: Money? Compute? Skills? Training data?
All of the above and more. Everything holding Mistral back is the same thing that has held Europe back from competing in the entire digital revolution. See this 1991 article lamenting the loss of any viable European PC manufacturer: https://www.nytimes.com/1991/04/22/business/europe-stumbles-...
Mistral being in Europe is disadvantaged with:
1. Money: less diverse private pension fund environment = less LPs to invest in VC funds = less VC dollars to invest in new ventures. European money is vacuumed out of the private sector into state pension funds and dumped into low yielding government bonds. This starves the private sector of capital while inflating the % of GDP driven by government spending every year (government pension funds buying government bonds in circular fashion enable runaway deficit spending...just like circular AI infrastructure spending).
2. Talent & compute: due to #1, Silicon Valley can outbid Europe for the best talent and hardware. Watch an OpenAI launch video and listen to all the European accents.
3. Local market fragmentation: Europe is a collection of countries that pretend to work together while not even having a unified capital market. The average EU citizen can barely communicate with their neighbor in a common language beyond the level of a toddler (english fluency is massively overstated by Americans who only experience tourist capitals).
4. Regulatory disadvantages: In everything from company regs, employee regs, unions, privacy regs, data portability regs, etc.
It's not "culture" or Europeans being "lazy" as most people would claim. There's currently thousands of young french people working 80 hour weeks creating dumb consulting powerpoints or legacy investment banking deal memos as we speak. Ambitious people exist everywhere in equal proportion, they're just working on the wrong things.
Europe can't compete in the digital revolution the same way they could compete in the industrial revolution due to various system design choices. Culture is simply the aesthetically observed byproducts of system design.
It says "through other Microsoft apps and websites," i.e. they reserve the right to include or remove it when and where they like throughout their whole product line (which includes github, of course), as well as:
- Conversations you have with Copilot through third-party apps and platforms
- Other Copilot-branded apps and services that link to these Terms
That first point (#4 in the original list) can cover all software, Copilot-branded or otherwise, which, even internally, uses Copilot (perhaps without your knowing so).
Github Copilot (to take your specific example) is both "other Microsoft apps and websites" and "Copilot-branded". So, yeah, those ToS undoubtedly apply to Github Copilot.
I think the big difference is that Google is free: everyone is using Google because it doesn’t cost anything and for a long time was the best search engine out there. I am sure that if Google would suddenly charge a few dollars per month for access, Bing market share would explode overnight, because it would become “good enough but cheaper”.
With the AI models, using a model that is “good enough but cheaper” is already an option.
There's no reason that a sizeable portion of LLM usage can't and won't end up free/ad-sponsored. Cutting edge stuff for professional use will probably be monetized via subscription or API credits for a long time to come. But running an older and less resource intensive model works just fine for tasks like summarization. These models will just become another feature in a "free" product that people pay for by watching or clicking ads.
I imagine the split will look a lot like b2b vs b2c in other technologies, b2b customers tend to be willing to pay for tech when it offers a competitive advantage, reduces their operating costs etc. b2c customers mostly just guzzle free slop.
It's actually pretty bonkers when you think about how basically every cutting edge professional you deal with is getting ads for all of their top search results for all of their work.
(Not quite "every", but outside of tech, most professional workplaces don't support ad blocking or Kagi.)
Yes, this statement is so old and wrong that it gets boring real quick…
You can just look at the roads of any country that has the same tariffs for U.S. and European cars and you will still not see all that many American cars there.
This is a very shady website and thus not a good source for legal advice of any kind… they call themselves “Deutsche Gesellschaft für Datenschutz” (German society for data protection) but are actually located in Bulgaria. They are not any kind of “official” data protection organisation.
The German BSI (Federal Office for IT Security) quotes the advisory from CrowdStrike (which is behind their customer login portal) as saying you need to roll-back to snapshots prior to 04:09 UTC:
I don’t know… I think a very big reason why people don’t take taxis is because they are very expensive especially for longer rides. This seems like a thing robo taxis might change. If the driver goes away, they shouldn’t be much more expensive than e.g. car rentals.
I do not read this court decision like that at all: the point of contention there seems to be that the customer was just sent a link to a webpage (where the contractual terms can be changed from under him at will by the company, thus this not being durable). The court makes it pretty clear in my (non-lawyer) opinion that attaching a PDF to the email would have been fine.
I was prepared to disagree with you, but I now have the same interpretation you have. Durable medium can be email - but the example seems a little fuzzy, for instance a durable medium is definitely when the email is stored on a HDD on a customer device. But is it still durable medium if the email only exists in a webmail? Probably yes, but maybe no. So the conservative approach would be to send paper for some things. (Or in this case, stupidly, USB devices. Banks, don't do that, please.)
Ramble Edit: it's unfortunate IMHO that there is no "read only" medium anymore. Not sure what it would look like now when USB-C is taking over the world, and that ship probably sailed, but it would be really cool and useful to have the option of a "data only" USB.
Maybe computers could have one USB port marked as "ROM". Or a switch or LED symbol indicating "ROM safe" mode.
When using such a ROM port, anything USB inserted there would only look like a DVD reader. A USB drive would get its files "mirrored" into a virtual ISO filesystem. Any other devices, such as keyboards etc would be just ignored and not connected to at all.
Most USB flash controllers support being read-only by either just being read-only or emulating optical drive. Obviously for the WORM usecase this is only an software solution inside the controller configuration as the underlying medium is still writable/erasable flash. In theory one could replace the flash with some kind of mask ROM with NAND-like interface and make it truly read only, but the cost makes that impractical for most applications.
Then there are LTO tapes that have WORM version, which is notionally not overwritable, but that is IIRC also only enforced by software (of the drive).
That doesn't fix the issue though. The issue is a killer USB or a virus on the disk. Being able to only read an infected file still allows it to be read.
Also, this is only a software solution as the USB protocol would require bidirectional transmission.
But it would bring us back to being as safe as a CD or diskette was.
I was thinking a special chip, talking bidirectionally both ways, pretending to be a PC host to the USB drive, and pretending to a DVD-ROM to the actual PC.
> There's an EU law demanding such documents to be delivered on a "durable medium". Some banks and financial institutions may have a strange approach to those, even though email attachments seem to be enough for others.
Even the (*-grand)parent never said the law actually says it can't be an email attachment, they said companies seem to interpret it that way. Which would not be surprising in the least. Then someone said they've never heard of any such law, and I pointed out that it exists.
I'm not sure who you're arguing with but it isn't me or in fact any of the people in this thread.
I am wondering what is keeping them back, though: Money? Compute? Skills? Training data? My fear is that you are really only getting really good models by training on very dubious data (outputs from the frontier models etc) and that Mistral is too European and too enterprisey to take those risks.
reply