But there are other more important (imho) values.

First, it is an excellent opportunity for personal worth. Socially: change of social group, freedom to explore, independence without much responsibility. Intellectually: mastering an advanced topic, in a program less directed than previous education, geared toward people with some demonstrated skill/affinity, meaning more depth. Also, no teacher or parent asking for grades, meaning more responsibility over own learning and organization.

It's a time for exploration. As it's said in classical AI, too much exploitation and not enough exploration means suboptimal solutions.

But also, higher education provides a resource for society. People with niche knowledge, expertise in a variety of topics, means a society is more resilient to disruption, and also more likely to innovate.

University is also a place for intellectuality and individuality, values very important to sustain a democracy, since democracy's main enemy is the inside enemy, in the form of ignorance and gullibilty.

Of course, to me all this means higher education is mostly a public service and therefore should be mostly taxpayer funded, but again, I'm from Europe :)

Content from TFA:

> Ambassadors on Friday will consider an “invitation of the President of the European Parliament [to] proceed with the Council’s first reading position” on the proposal to allow tech companies to choose to scan for CSAM

> The European Commission proposed the temporary CSAM bill as a stopgap measure to allow companies to scan while legislators agreed on a more permanent solution.

> Tech firms continued to scan, despite the legal limbo.

> If capitals choose to adopt their position, the law does not automatically pass: The Parliament would need to either accept it or re-enter negotiations. “There is no certainty that [the Parliament] would adopt the legislative act in second reading in line with the Council’s first reading position,” Cyprus wrote in the note.

> Gregorová rejected the suggestion that lawmakers would budge. “The Parliament mandate is clear: A majority voted it down, meaning that we reject the extension.”

so I'm always surprised when some linguists decry LLMs, and cling to old linguistics paradigms instead of reclaiming the important role of language as (a) vehicle of intelligence.

Also, just as you can drop many IPA symbols and just get the basic set needed to represent a particular language, I guess you could use "simplified" SW ignoring the fine differences.

Now, it may not be obvious that there is a necessity for a writing system for a minority language embedded in a larger community (spoken language), but there are many uses: preservation, digital use, teaching, linguistic study...

Also, it's not a zero sum game at all. First, even if the long stretches of good and bad governments equalized over the long term, it's much worse to live all your life under a bad government than have one half the time. Also,thinking as an engineer, think about the stability implications of a wildly swinging signal vs a smoothly swinging one, even if both are centered.

In any case, coming from a country which had a long dictatorship until not so long ago, it's fairly clear it's not a zero sum game.

There has been unusually heavy rain in Spain for an unusual length of time now, and reservoirs are well beyond average capacity.[1]

I wonder if this anomalous situation has anything to do with the blackout. Maybe the inability to continue pumping water led to the shutdown of the solar, maybe for economic reasons as is suggested in some other comment?

[1] https://www.embalses.net/

Unfortunately there's no getting away from that. While the scarcity of data indeed is an issue, and your idea is nice (congratulations!) the actual problem is the scarcity of useful data. Since sign language doesn't correspond to the oral language, there are many problems with alignment and just what to translate to. Glosses (oral language words used as representation for signs) are not enough at all, since they don't capture the morphology and grammar of the language, which among other things heavily relies on space and movement. Video + audio/audio captions is nearly useless.

Good luck with your efforts, this is a fascinating area where we get to combine the best of CS, AI, linguistics... but it's hard! As I said, let me know if you want some literature, by PM/email if you want, and I'll get back to you later.

The article is excessively negative in tone, and very dramatic and aggressive. I have found many people adjacent to academia, drop outs, or even some inside, very disenchanted and angry at how it works. And it's true, the sets of incentives, structures and political organisation in academia don't relate at all to academic excellence, and are something we have to "suffer". I wish we could come up with a better set of incentives, but it's very hard to do in a mostly vocational and passion-based activity. So what people have come up is structure the incentives along the chores (eg teaching) and easily measurable results (eg publications). And whenever you come up with an incentive structure, some people will game it. And the current state of publication stress (publish or perish) is extreme and counterproductive. But please note that these measurement requirements and incentives are imposed from outside academia. Of course, I'm not saying leave us to our devices, academia is nepotistic and political enough. But the system sure could use some overhaul. Suggestions welcome.

On the other hand, this "fraud" is incentive fraud, but not "truth" fraud. The way science Truth works is by accumulation of imperfect, even erroneous results, leading to an ever more refined understanding of the world. Scientists don't just blindly trust others, even if they cite each other (nowadays, citations are a political and incentive-gaming tool more than actual references). So these massive scale frauds don't bother us so much because they don't make understanding necessarily go backwards. Of course the payer feels it's a waste of money, but in academia we see money as support for research, which is mostly failed anyways because you only make discoveries by failing and failing again.

And progress in knowledge is nowadays still going on, even in the medical fields. And academia still works, much as healthcare and compulsory education, becausemany people feel a calling to do these professions properly, even if it doesn't seem so from outside. So let's be optimistic, even while trying to come up with improvements to the current model.

PS: So sorry for the wall of text

not tldr but provocative thought derived: why insist that "it's" as a possessive is wrong, when it follows an existing pattern (john's, everyone's) and speech doesn't make the distinction?

I think the concept of the basis of functions, a series of functions which you can combine to approximate any other function is something that engineers should be able to understand. Then you can see the time to frequency as a (very useful) application.

BTW, the visualization in the post as to how the transform works is awesome, and can also work with the function basis explanation instead of the frequency one. In fact, it might make even more sense! And the posterior mention of the cosine transform wouldn't need to be hand wavy about real and imaginary parts.

In any case, I've seen so many engineers insist in the time to frequency explanation that it must somehow be easier to understand for people. I just lament that the beauty of function spaces is lost in these explanations, as well as the underlying understanding of why Fourier transform is not only useful but deep.

Anyway I'm excited and looking forward to the code and models to be released, hopefully I can use them for my research! I think it's easy to overlook how revolutionary the transformer "way" of doing things has been, and the fact that so many different tasks can be reformulated in a "language" way I believe hints at something deeper about how the universe, our minds and language work.

1) they use a single, shared embedding space for the two languages, forcing the model to learn "semantics" independently (or rather, interdependently) of language 2) using back-translation for training. I'm not sure that I got this right, but this seems to be round-trip translation? So the model can self-assess its performance by checking the spanish->english->spanish difference.

Sounds very promising and interesting! However, it seems they only tested on spanish and english. I wonder if the similarity of the languages at the lexical level made these results possible.