>This is not at all what I would expect because it's trivial to change the training data to replace Claude with Kimi.
Wait what? The reason you wouldn't expect it is because if it was distilled, it would be easy to get rid of self identification? Is that any less true of a non distilled model? I suppose there's lots of ways to interpret it, but the idea that self-identifying as Claude is affirmative evidence that it's not distilled seems to get the weight of the inference exactly backwards.
The article feels like a long way of saying that there's a significant corpus of underwater acoustic data, a complicated bureaucratic process for procuring it and mobilizing analysis of it, and not finding a smoking gun. But it using the saga as a springboard to talk about the various technologies and processes used to power the investigation in the ways that different lines of evidence could be reasonably hoped to converge and how the fact that they haven't might point point the investigation in a new direction.
I do think the framing of the article seems like it at least has a whiff of an implication that somewhere in there, they're going to tell you that they found something specific, that at least offers a tea leaf of additional new data. But if you're reading the whole article, waiting for the grand reveal, there isn't one.
I agree and I think it was a mistake to jump to a new name. My understanding of the kind of whole big picture conception of Alphabet was that it redefined Google as a set of related enterprises in a very open-ended way such that you could think of them as practically one per letter.
I'm pretty sure the announcement post for Alphabet even was titled g is for Google. Bard was a perfectly suitable, even good, entry as names and branding go. I think Gemini was a problem not only because it plays into the negative perception of Google not being committed to their projects, but also because one of the singular messages of Alphabet was that g is for Google and other letters were for other things. And Gemini obviously is going to be one of the more important Google ventures in a long time, but pretty self-undermining as brand coherence goes.
Not but the most important thing in the world, of course. And who knows, they'll probably rename Gemini in 18 months anyway.
Thanks for the link, and boy is that a whopper. Do my eyes deceive me or is there literally a shot of the computer looking at them from its computery perspective, through its screen, with some layer of visual digital artifacts over it at the 14 second mark?
I love this on a number of levels. First, a great question, and more than half the battle imo is asking the right kind of question. David Cain has touched on a version of this also [1]. And I very much am a person that can get lost in my own organizational overhead and know I'm being very out of balance between thinking and doing, and in search of an overarching framework or limiting principle.
And I don't know that this solves everything, but it does bite off a piece of it, in a way that I think I agree with. The further out, the more circumstances can change, and the closer up, the more practical reasoning matters (they seem to use reasoning to mean something close to practical logistics). There's probably more to it, of course, but it's a good question and good start to an answer.
I got the same impression especially from reading the about page. And it wasn't the em dash this time but the triplet of one word sentences.
But just to comment directly on it and offer something constructive I appreciate a non systemd option that looks nice and hope this gets enough of a user base that it counts as votes in favor of that kind of paradigm.
For me right now it's not going to any article, but a kind of general landing page for the Eureopean Commission's "Press Corner" which lists a bunch of articles:
Wikipedia is just about the last bastion of institutionally vetted knowledge intended for and generally respected by the public, and in an age of maximalist misinformation, it was only a matter of time before people tried campaigning to undermine public trust in it.
If you think the 2020 election was rigged or that Covid is not real, or that climate change isn't happening, or that the earth is flat, or that the moon is made of cream cheese, chances are you strongly dislike the Wikipedia entries on those subjects and believe it's biased against you.
And these aggrieved parties can find each other online and start workshopping narratives of why their causes were unfairly targeted and propagandized, and from there you're off to the races with uncountably many whackamole narratives that create an increasingly self-sustaining alternative reality where it's just obviously the case that Wikipedia is evil. Which, I think is a symptom of Wiki's success in the face of failed pushes to spread misinformation on it as a platform.
I think fact checking sites had a good heydey from say 2008 to maybe 2016, but a combination of underqualified editors in some cases, Enlightened Centrists who have confused ideas about truth, and Bannon-style propagandists who flood the zone and argue that they are biased, were able to undercut those fact checking sites as a conventionally accepted model of information vetting.
Wikipedia continues to be resilient against those efforts because they have well defined processes for reliable sourcing that are strongly enforced and almost purpose built for a moment like this, so the misinformation campaign can't operate from the inside to the degree that it would like to. It doesn't mean there's no issue with intentional PR campaigns or other campaigns flying under the radar but it feels like a pound of grievance from misinformation spreaders for every ounce of legitimate complaint, banking on people not because able to compare the relative scales of the different categories.
I don't understand why you think your question in reply to theirs revealed that theirs was naive.
If anything, I think it actually reduces the quality of discussion because it tries to say that dynamics in China are equivalent to the next you would find in pretty much any country which is is vague and lazy as analysis goes, and goes against the HN recommendation that conversations get more precise over time.
>it's the choice between "prosecute this one or nobody"
Even that assumes a normal of being lucky that anything is prosecuted, ever. So it's good but against a low bar rather than rising to the bar parent commenter suggested.
Just saying I would buy a chat control legislation calendar, where each month of the calendar has significant meeting days of deliberation bodies, elections/nominations of people to relevant boards, as well as historical dates of previous attempts to pass chat control.
Jk, but the skepticism is inevitable. I think we can be dubious about how AI mobilizes global capital while also appreciating tutoring as one of its best targeted use cases.
There's evidence for stunningly intricate, marvelously impressive biological function, but there's no evidence whatsoever for internal mental states.
What's happening is we borrow from a pre-existing vocabulary with connotations to conscious activity and we use it to describe automatic biological reflexes, and then some people lose track of whether it was an analogy or whether it was literal and start claiming plants can "feel."
I think this is a huge error which collapses the most urgently important distinction at the heart of the entire issue. Plants do complex, interesting things, and we borrow from an already existing vocabulary to make analogies. So people will say plants "feel" or "communicate" or "see".
But those are automatic biological reflexes, and using words with connotations to conscious awareness to describe them makes people think they're the same thing. But a plant "feeling pain" is repurposing words about consciousness to mean new and different things. And I think people are seduced by it because it feels like being open-minded to new possibilities.
I remember that. The one thing I would add is I think the usage was much more general purpose. "Free stuff" sites were a big deal and huge source of traffic and .tk was widely shared on those. You could have a banner with ads and have the domain for free.
I'm also left to center and I agree with you on all three! I think left-wing examples are subtler and hard to identify. Even among among Democrats left of center types and liberals, the examples I cited are minority opinions, but likely to be systematically found on the left, a phenomenon which won't be negated by the self-report of one person correctly tracking the facts (and to be clear, I think you are correctly tracking the facts on each of these!).
>Factuality is orthogonal to political leaning generally.
It certainly can be orthogonal, in some notional sense, and in many cases that explanation is good enough. But in practice there are too many contrary cases to ignore, and there's often an integral relation between factual veracity and polarization, especially with respect to American polarization of politics. Global warming, the results of the 2020 election, the percent spent of federal budget spent on foreign aid have factual answers and right wing affiliation can be predictive of (1) not agreeing with the facts and (2) treating factual corrections as "liberal bias".
I think left wing versions exist also but are less systematic: 2004 election results, efficacy of plastic recycling or dangers associated with nuclear power are cases where I think left wing partisan affiliation probably predicts being wrong on the facts.
And meta-narratives about the relation between factual information and partisan bias are themselves as likely to be polarized as anything, complicating the ability of people to do good analysis, or of accurate analysis to be trusted by people committed to certain meta-narratives that would deny the possibility of factual knowledge predicting polarization.
Amazing, and I love every pixel of it. We know the MLB is famously understanding with creative repurposing of their data, so I wonder what live data you are using and if the rationale for use is something like small scale hobby/fair use?
I know when it comes to historical data, projects like the Sean Lahman Database have to go through quite a bit of trouble to reproduce "clean room" versions of historical data that are legally fine to use. I have to imagine there's a lot of complications when it comes to live data for anything that even has a hint of being more than a hobby project.
Wait what? The reason you wouldn't expect it is because if it was distilled, it would be easy to get rid of self identification? Is that any less true of a non distilled model? I suppose there's lots of ways to interpret it, but the idea that self-identifying as Claude is affirmative evidence that it's not distilled seems to get the weight of the inference exactly backwards.