The interview started with the most mundane question "Who are you?", and the very first sentence of Wales is either a lie or misleading. The journalists asks for clarification (thats a journalists job, btw), and in his second sentence of the interview Wales insults the journalist. I'm pretty sure who is the jerk here.
It also was Wales who bought up the topic, not the journalist. If he considers it a stupid topic he does not want to talk about, why is it the very first thing he talks about?
Sanger was originally hired to edit Nupedia, a web encyclopedia project with a strict peer review process, and only worked for Wales for about a year. Wikipedia was started as a side project (with Sanger contributing to the concept and some early organizing), but Wikipedia quickly became much more successful while Nupedia basically never got off the ground. My impression is that Sanger wanted to impose his own vision on Wikipedia, but couldn't because the community of volunteer editors disagreed, and when Wales stopped paying him as a full time Nupedia editor (Wales's company was tight on cash at that time), he stopped any involvement. This was long before most of the actual work of Wikipedia happened, and that should have been the end of the story.
But ever since, Sanger has been trash talking Wikipedia as a project and community ("broken beyond repair") and trying to undermine it. A few years later he started a competing project (which was predictably a total failure). For two decades he has been promoting himself as "cofounder of Wikipedia". Interviewer after interviewer asks the same lazy questions about the subject, without ever adding any new insight. (You can see that Sanger's ghost is chasing Wikipedia even into this discussion.)
It's beating a dead horse, and entirely off the topic of what the interview was supposed to be about. Answering the question clearly and accurately takes a lot of time and finesse, which is wasted on the interviewer and most of the audience. Wales clearly screwed up in that interview, but it's not hard to see where he's coming from, psychologically.
What an interview!
I had never seen this clip before, it's really something. Facts and context are important for sure, but as someone who isn't clued in on the Sanger drama, Wales could not possibly have made himself look worse. And in under a minute!
As you said, the interviewer is in the right, carrying out the job of interviewing, by pushing Wales as he did. To call him a "jerk" is silly, I think.
>(This is probably because Google Maps can be used for walking/biking too)
Please don't do that. The map is simply not good enough and does not have enough context (road quality, terrain, trail difficulty) for anything but very causal activity. Even then I highly recommend to use a proper map, electronic or paper.
It has a lot more map data accessible and you can even overlay National Park Service maps, land ownership, accurate cell service grids, mountain biking trails, weather conditions and things like that.
Disclaimer: Just because you see a route on a map, digital or paper, does not mean it is passable today. Or it may be passable but at an extremely arduous pace.
We used the walking directions for dual sport motorcycles once. It was pretty nice. We did have a few places where it became sketchy. Those and maybe more places would be sketchy for walking too. Not that google maps could do much about it. Terrain is a living thing. These were mostly huge cracks in the earth due to rain water.
Trail? Terrain? I use it for walking for 10-20mins around a (mostly flat) city and I expect that’s what 90% of people use it for, the comment didn’t mention hiking
It depends what you are doing but for hill walking in Italy I found the footpathapp.com app good. There are no decent paper maps in the area I go and Google maps are also rubbish for local paths but the app kind of draws in paths based on satellite images I think and you can draw on it to mark the ones you've been on.
Dubai has an entire active operation. It looks like it does work, but how well is debated. Seems to have enough of an impact (correlation or causation) that they haven't shut it down yet.
> However, there are only 24 permanent residents and five active farms on Hamnøya. Therefore, there is regular transport of tankers, concentrate feed and livestock trucks.
> Hamnøya is an island in Vevelstad Municipality in Nordland county, Norway. The 16.6-square-kilometre (6.4 sq mi) island lies about 500 to 700 metres (0.3 to 0.4 mi) off shore from the mainland of the municipality, separated by the Vevelstadsundet strait. The island is only accessible by boat and in 2021 it had 35 permanent residents living on the island.
I'm not sure if it's cheaper to upgrade both posts, but a bridge doesn't look so silly.
Arabic, even. An outlier, as it is AFAIK the only arabic dialect that is not written with the arabic alphabet. Also it's far removed from other arabic dialects.
Maltese isn’t an Arabic dialect. Yes, the grammar and phonology and core function words derive from Arabic, but more than half of the vocabulary comes from Italian/Sicilian-North African Arabic may borrow a few words from Italian here and there (just like English does), but not >50% of their vocabulary.
I think the family tree model of linguistic history is not very useful for English. Saying English is Germanic to the exclusion of everything else is not very useful.
The family tree model seems to assume that every language has only 1 direct ancestor. It seems to have been inspired by phylogenetic trees in biology. In phylogenetics, single-parent trees work fine because distantly related species can't breed with one other. By contrast, different languages borrow features from one another all the time. It could perhaps be useful for some languages, but not for English. I reckon.
Can’t read the Hebrew alphabet, but transliterated to Latin: “a shprakh iz a dyalekt mit an armey un flot” - I find it fascinating that despite knowing close to zero Yiddish, it makes complete sense.. well, I know a handful of German words (which covers “mit”)… and “flot” contextually makes sense as “navy”, especially if one knows English “flotsam and jetsam” (not navy but at least nautical)
It's not at all far removed from the North African dialects of arabic which is the dialect that it's derived from. Tunisians and Algerians can understand Maltese quite well.
> Tunisians and Algerians can understand Maltese quite well.
Not in my experience. Not at all actually. My experience with Arabic speakers is that they think they're understanding when you speak Maltese, because it sounds kind of familiar, but in actual fact they're not understanding much at all.
Which is not surprising after a thousand years of divergence.
Oh, stop it! What are you really trying to say? 'The same language' is usually just a desguised nationalistic claim. Ask yourself: what is the advantage of a language over a dialect or vice versa? Why are you fighting for it (or against it)?
Linguistically, it does not matter -- there is no objective definition of the difference between a language, a dialect, or whatever -lect.
>'The same language' is usually just a desguised nationalistic claim
It's the opposite: "it's a different language" is usually just a nationalistic desire for differentiation of what are essentially dialects/variants of a language.
>Linguistically, it does not matter -- there is no objective definition of the difference between a language, a dialect, or whatever -lect.
That's more because academic linguistics, as developed in the latter half of the 20th century, had to pay lip service into several ideologies, rather than there not actually being good practical ways to discern e.g. arabic as a single basic language with different variants.
> > 'The same language' is usually just a desguised nationalistic claim
> It's the opposite: "it's a different language" is usually just a nationalistic desire for differentiation of what are essentially dialects/variants of a language.
It's both. The idea that Ukrainian is an uneducated farmer's dialect of Russian is a common talking point in the "Greater Russia / Russkiy Mir" narrative. Conversely, asserting the status of the Ukrainian language is a big part of Ukrainian identity in the face of an imperial invasion.
> That's more because academic linguistics, as developed in the latter half of the 20th century, had to pay lip service into several ideologies, rather than there not actually being good practical ways to discern e.g. arabic as a single basic language with different variants.
As someone who once studied General Linguistics, I don't understand this remark. I've learned that calling something a language is a political act and often of great significance to the speakers, but is almost never well-defined from a purely linguistic perspective. That's a fact. Although you can sometimes find typological criteria to further argue that a variety is a language on its own, for example there are good grammatical reasons for not counting Swiss German as a variety of German, you will also find examples the other way around where two varieties have large lexical and grammatical differences and still count as the same language.
The strongest criteria for what counts as a language are based on language origins (as opposed to typology), and these do not generally suffice or make meaningful distinctions to varieties (~dialects). Mutual comprehensibility can be very low for speakers of the same language, which is why most research focuses on varieties or on speaker groups that are of particular sociolinguistic interest.
I don't get why you talk about "academic linguistics" as if there was a non-academic one and why you think linguistics "had to pay lip service into several ideologies." What are you talking about?
It's simple: linguistics is a politicized discipline, and there's a prevailing ideogically motivated tendency to put every language and dialect on equal footing.
>As someone who once studied General Linguistics, I don't understand this remark. I've learned that calling something a language is a political act and often of great significance to the speakers, but is almost never well-defined from a purely linguistic perspective. That's a fact.
Yes, this ideologically motivated idea after enough repetitionbecame "a fact" of the field, as if describing some objective physical law, and even non-political students will be taught and stick to the same (and anybody with a dissenting opinion will be getting an earful if not committing career suicide).
This wasn't always the case, it's more so with liberalism prevailing, especially in the latter half of the 20th century.
What you claim didn't happen, and can easily disproven with data. Your interpretation of a reasoning of a policy (that didn't happen) is bad faith.
You are wrong about both electricity [1], gas[2] and total energy [3].
Europe was very dependent on energy imports in the past and current policy is the by far most successful attempt in a long at changing it. It will help us for decades to come.
I don't understand your post. Europe absolutely spent many years phasing out nuclear energy while rationalizing that increased gas imports from Russia was good because trade will make us friends [0]. The data supports this (though obviously does not capture the political discourse around Russian gas reliance). I am in agreement that the current, post-Ukraine invasion policy is good.
I guess you’re talking about Germany. Shutting down nuclear reactors and huge imports of Russian gas are two entirely separate things.
Gas imports like explained in the Wikipedia article you linked started in the 70s. There certainly was too much of a reliance on Russia, not enough investment in alternatives like LNG terminals and warnings of partners in the east were ignored with the north stream pipelines. In the end Germany got a lot of economic growth from cheap gas for decades and managed to get off Russian gas very fast. The European nuclear industry on the other hand is still heavily reliant on Russia.
Broad German anti-nuclear sentiment can be traced back mainly to Chernobyl and the exit plan that was followed in western Germany was decided after Fukushima. Eastern German reactors lacked containment and were shut down after the reunification. Contrary to an often heard claim western German nuclear was not replaced by fossil energy sources, but more than compensated by the growth of renewables. You could certainly point out that coal plants should have been shut off first, but that was even less possible politically at the time. In the end you have to shut off nuclear reactors because of growing safety risks caused by material fatigue and new ones have questionable economics.
Looking in absolute terms France is essentially phasing out nuclear power. But can't bring themselves to accept it politically.
They have an enormous fleet nearing end of life and are making political noise over a tiny number of plants they haven’t even taken the final investment decision on.
They are stuck arguing how the mindbogglingly large required subsidies should be financed.
In other words: They are betting on renewables as much as anyone else, they just can't bring themselves to stop wasting money on nuclear power due to political reasons.
It is more representative of the situation over the past 30 years to say that France is not Europe. But even they had plans to cut nuclear energy to 50%. [0]
As you can see in the first link, there was a 472 TWh production of nuclear Nuclear power in 2024, in contrast to your claim that it was phased out. It didn't really changed that much - the absolute change in coal or renewables combinded is larger.
This chart shows that nuclear produced >900 TWh in the EU in the early 2000s and the latest data point is 619 TWh (2023). That's a >30% reduction with a clear trend. I did not claim that every nuclear reactor in Europe was turned off.
> because they can only solve things that are already within their training set.
That is just plain wrong, as anybody who spent more than 10 minutes with a LLM within the last 3 years can attest. Give it a try, especially if you care to have an opinion on them. Ask an absurd question (that can be, in principle, answered) that nobody has asked before and see how it performs generalizing. The hype is real.
I'm interested what study you refer to. Because I'm interested in their methods and what they actually found out.
The crux is that beyond a bit of complexity the whole house of cards comes tumbling down. This is trivially obvious to any user of LLMs who has trained themselves to use LLMs (or LRMs in this case) to get better results ... the usual "But you're prompting it wrong" answer to any LLM skepticism. Well, that's definitely true! But it's also true that these aren't magical intelligent subservient omniscient creatures, because that would imply that they would learn how to work with you. And before you say "moving goalpost" remember, this is essentially what the world thinks they are being sold.
It can be both breathless hysteria and an amazing piece of revolutionary and useful technology at the same time.
The training set argument is just a fundamental misunderstanding, yes, but you should think about the contrapositive - can an LLM do well on things that are _inside_ its training set? This paper does use examples that are present all over the internet including solutions. Things children can learn to do well. Figure 5 is a good figure to show the collapse in the face of complexity. We've all seen that when tearing through a codebase or trying to "remember" old information.
I think apple published that study right before WWDC to have an excuse to not give bigger than 3B foundation models locally and force you to go via their cloud -for reasoning- harder tasks.
beta api's so its moving waters but that's my thoughts after playing with it, the paper makes much more sense in that context
What you think is an absurd question may not be as absurd as it seems, given the trillions of tokens of data on the internet, including its darkest corners.
In my experience, its better to simply try using LLMs in areas where they don't have a lot of training data (e.g. reasoning about the behaviour of terraform plans). Its not a hard cutoff of being _only_ able to reason exactly about solved things, but its not too far off as a first approximation.
The researchers took exiting known problems and parameterised their difficulty [1]. While most of these are not by any means easy for humans, the interesting observation to me was that the failure_N was not proportional to the complexity of the problem, but more with how common solution "printouts" for that size of the problem can typically be encountered in the training data. For example, "towers of hanoi" which has printouts of solutions for a variety of sizes went to very large number of steps N, while the river crossing, which is almost entirely not present in the training data for N larger than 3, failed above pretty much that exact number.
It doesn't help that thanks to RLHF, every time a good example of this gains popularity, e.g. "How many Rs are in 'strawberry'?", it's often snuffed out quickly. If I worked at a company with an LLM product, I'd build tooling to look for these kinds of examples in social media or directly in usage data so they can be prioritized for fixes. I don't know how to feel about this.
On the one hand, it's sort of like red teaming. On the other hand, it clearly gives consumers a false sense of ability.
Indeed. Which is why I think the only way to really evaluate the progress of LLMs is to curate your own personal set of example failures that you don't share with anyone else and only use it via APIs that provide some sort of no-data-retention and no-training guarantees.
This is "can" has exactly the same meaning as in "UB can make your programms faster". You could replace it with "it does, at least with clang". LTO is, in this regard, the same as UB, and unlike guaranteed optimizations, such as the single member optimization, or the empty base optimization.
Concretely, here, the UB-exploitation in question in this case is assuming that the "this" pointer in C++ is aligned and non-null, meaning it's a pervasive annotation throughout C++ codebases, not an edge-case.
Relying on LTO to "discover" this annotation through interprocedural analysis -- based on my experience of looking at LTO in practice -- will not be as comprehensive, and even when it works it accomplishes its task in an achingly-slow and expensive way.
Somewhat separating owning and non owning memory in the type system goes a long way. Also a much better standard library and a stricter typing discipline.
The fact that it's mostly backwards compatible means you can reproduce almost all issues of c in c++ awell, but the average case fares much better. Real world C++ does not have double-frees, for example. (As real world C++ does not have free() calls).
It also was Wales who bought up the topic, not the journalist. If he considers it a stupid topic he does not want to talk about, why is it the very first thing he talks about?
reply