baryphonic

(American here.)

I don't think I'd support a ban if ByteDance was a European company or Indian or South Korean or Japanese. China is a unique threat given the totalitarian turn they've taken over the past decade combined with the fact that no Chinese company is truly private in its day-to-day operations. All Chinese companies must have CCP influence as a matter of Chinese policy. It would be like if T-Mobile (the US mobile division of Deutsche Telekom) was required to have the influence of the German government including the monitoring and reporting of phone calls to senior party officials.

>Fisher became founding Chairman of the University of Cambridge Eugenics Society, whose other founding members included John Maynard Keynes

Keynes was the leading economist of the 20th century. He has some ideas I think are dubious, and his followers have doubled down (I still can't believe people believe in fiscal multipliers greater than 1). Nevertheless, it would be an incredible cheap shot to label Keynes a "eugenicist" when criticizing his economic theories.

2006 was peak "Great Moderation," where there hadn't been any serious financial crises since the 80s. The consensus wisdom was that many of the policies of the 80s & 90s, particularly around inflation and reducing uncertainty, had made financial crises almost obsolete.

It seems dark today given that we know the outcome, but I'm sure at the time, Comic Sans seemed appropriate for a set of tools that they thought likely would never be used. Or maybe it indicates a certain hubris undone within about 18 months.

Each compiled executable has a one-to-one relation with its source code, which has an author (except for LLM code and/or infinite monkeys). Thus compiled executables are derivative works.

There is an argument also that LLMs are derivative works of the training data, which I'm somewhat sympathetic to, though clearly there's a difference and lots of ambiguity about which contributions to which weights correspond to any particular source work.

Again IANAL, and this is my opinion based on reading the law & precedents. Consult a real copyright attorney for real advice.

Going a step further, weights, i.e. coefficients, aren't produced by a person at all – they're produced by machine algorithms. Because a human did not create the weights, the weights have no author. Thus they are ineligible for copyright in the first place and are in the public domain. Whether the model architecture is copyrightable is more of an open question, but I think a solid argument could be that the model architecture is simply a mathematical expression – albeit a complex one –, though Python or other source code is almost certainly copyrighted. But I imagine clean-room methods could avoid problems there, and with much less effort than most software.

IANAL, but I have serious doubts about the applicability of current copyright law to existing AI models. I imagine the courts will decide the same.

Its literally manipulation of those who have money to spend them on product they otherwise wouldn't, has absolutely 0 relationship on quality on the product (in extreme cases it goes directly against it).

This is an extremely strong claim. Certainly you'd concede that some ads contain truthful information. Like there exists at least one ad that is true. So then how is it "manipulation" for someone to post that information in a public space?

We jumped from "billboards are ugly" to "ads are categorically evil," and based on some pretty strong assumptions.

Word of mouth, unbiased reviews (yes, they cost something to keep the interference away but save you tons of money and time down the line).

Okay, so how do you get the first person to buy your product if advertising is illegal? The base case would seem to require it. Same goes for "independent reviews." How do you find the independent reviewer? And this is ignoring getting a critical mass of customers for word of mouth to even work.

s/defaults/segfaults/ # stupid autocorrect

I fail to see how this parade of horribles will happen. Under the Chevron regime, any random person could still sue, and provided that the lawsuit survived an initial motion to dismiss, then any questions involving an administrative agency policy would defer to that agency's interpretation of their own policy and the law authorizing that policy.

The only change now is that the agency will have to demonstrate to an independent Article III court that its policy is correct and compatible with the authorizing law. Stare decisis will still control the lower courts once new precedents are set, and people will have meaningful appeals again.

There might be some disruption in the short term, but in a decade or two, I expect the new normal will be fine, but with the benefit that people can meaningfully appeal self-aggrandizing administrative state rulings.

Galileo also couldn't explain the lack of an observed parallax effect between opposite seasons given the ideas about optics at the time.

When Kepler's model arrived, it was so much better at predicting the positions of all planets except Mercury than any previous model that it was clearly superior. Galileo's was bad at predicting and just contradicted the accepted observations of the day.

IMO Galileo should be better remembered for objects of different masses falling at the same rate and the original idea that all motion is relative (when observing from an internal frame).

I find this bizarre as well. The EU Parliament has no initiative, possessing only the power to approve or veto legislation proposed by the Commission. And then when they occasionally exercise their veto power, the Commission can just resubmit substantially the same legislation as before, as is happening here.

This also means the Parliament has no independent power to repeal previous law that it might have regretted passing. It must again wait for the Commission to propose repeal. I can count how many times an unelected administrative bureaucracy has proposed removing its own power on zero hands.

The whole thing strikes me as a sham democracy.

Not to be too pedantic, but ALJs are technically Article I judges (or more precisely judges over Article I tribunals). The Constitution lists two different sources of judicial power, the first in Article I section 8 and the second in Article III. Article III courts have judges with life tenure, protection of salaries and are subject to review only by other Article III appeals courts including the Supreme Court. Article I courts have judges with fixed terms of office, and Congress can cut their salaries. All Article I courts are subject to review by Article III courts.

There's an open controversy about how much deference the Article I courts in administrative agencies are owed by Article III courts, arising mostly from Chevron v NRDC. That decision requires Article III courts to defer to Article I courts' interpretations of their statutes and even their administrative rules except in extreme circumstances. Several justices on the Supreme Court find Chevron deference problematic, but it currently is the law of the land.

I wonder how this AI interprets customers who express anger and frustration through sarcasm

I guess Big Pharma was better at using lobbying dollars than Big Mask.

Interesting. I liked the 1Password 6 UI, was frustrated with 1Password 7 and have been loving 1Password 8 so far. Version 7 seemed really clunky when I needed to do certain workflows.

I tend to be anti-regulation unless the benefits of the regulation can be justified above and beyond laissez-faire, which in practice makes me quite anti-regulatory with some notable exceptions like the FAA.

I think the root of this problem is that private equity managers often have no skin in the game and tend to fail upward. It would be one thing if a private equity manager had a record of taking a company private, streamlining it and making it sustainable for decades to come. Even with the human cost of that, it's preferable to a company going prompt bankrupt. But too many managers seem to take these things private and then run them straight into the ground. Toys'R'Us, for instance, or Red Lobster as quoted in the article. In the latter's case, the management tried boneheaded promotions like "unlimited shrimp," which would be a bad idea even in a zero interest rate world.

I'd propose instead some sort of mandatory filing on the part of private equity managers that is publicly accessible and searchable and shows the track record of a private equity manager, with links to all of the other managers they've worked with. Then, when a PE investor proposes to take a company private, they're required by the SEC to demonstrate that their management isn't tainted by a chain of incompetence.

Who knows if that would work, but it might increase the skin in the game somewhat. The status quo seems to be HBS grads performing with mediocrity at best and having no real accountability.

Neat tool.

Would be cool if this CLI could have a flag to read .gitignore and exclude all of the contents automatically.

Also it might be cool to have different strategies for detecting secrets, e.g. Kolmogorov complexity as other comments have noted.

If people were avoiding FOIA by having candid discussions on private servers and deliberately misspelling words all while telling the public a contradictory story, that is strong evidence of lying.

If the private communications matched the public ones and there were no efforts to obfuscate, then the best conclusion would be they just called it wrong.

Internally they had dissent but dismissed it and went so far as to misspell things so FOIAs would not find things.

I hadn't heard this yet. Unbelievable. And yet all of the sites I found it on from a quick search have at some point in the past been branded "fake news." In fact, one source, the New York Post, was falsely branded "Russian disinformation" on the eve of the 2020 election and suspended from Meta and Twitter, only for its story to be verified subsequently when it had minimal consequence.

When I read the title, my immediate thought was "because 2s complement!"

Your explanation is more thorough.

I agree with you. I'd add that this kind of snarky sniping nonsense causes rifts between people who might otherwise agree with one another.

I laughed out loud when I read that. Would expect nothing more from Gizmodo.

EDIT: cite source

I have yet to see any evidence of this. To the contrary, we have seen for years that advances in LLMs require orders of magnitude more parameters/power for each generation. Neural architecture search has been underwhelming. RLHF seems to have regressed models like GPT3.5+, rather than improving them. And recently, researchers concluded that multi-modal models require exponentially more data for each extra "mode" being added to the model.[0]

Even in my own experimentation, I've tried to get some AGI-like behavior, and it just isn't there. I have convinced GPT4, for instance, to generate XML source for an SVG, but it looks nothing like what I describe.

I'd argue that these models don't generalize well at all, and I'd bet that, like with Moore's Law, advances in AI will require continual discovery of incremental new innovations and occasionally new architectures.

[0]https://arxiv.org/abs/2404.04125

I'll be honest, I don't hear a close similarity. If I were to pick a celebrity, I'd say it sounds more like Rashida Jones, though Rashida's voice is slightly higher.

Rebase workflows are awful and unintuitive. Leave the rebasing to the git wizards who actually know what they're doing, in no circumstances should this be part of your day-to-day work.

Hard disagree. I hardly consider myself a "rebase wizard," but I've been a near-exclusive practitioner of rebase workflows since I can remember. I find rebasing much more intuitive than workflows with merge commits. Squash merges are fine, but with proper intuition, they appear like a special case of rebase.

In my experience, the resistance to rebasing comes down to fears about "rewriting history" and false intuitions about how git works. I usually allay the former by pointing out that squash merges - which almost everyone approves of - also rewrite history. The latter issue seems to arise from arrows in popular git visualizations pointing in the wrong direction, e.g. in Gitflow.[0] In git, the child commit points to its parent, because each node is immutable. The git data structures are extremely simple (hence why git is so named), consisting of blobs, trees, commits, tags and references. Once you understand how these work in practice, rebase becomes intuitive.[1][2]

IMO, the only thing unintuitive about git is the CLI. Translating the graph operation I want into the commands is sometimes a challenge. Maybe that makes me a (frustrated) wizard after all?

[0] https://nvie.com/img/git-model@2x.png

[1] https://speakerdeck.com/pbhogan/power-your-workflow-with-git...

[2] https://eagain.net/articles/git-for-computer-scientists/

Negligence is a significantly lower standard than actual malice, and presumably a person who has been defamed would be seeking significant actual damages anyway (the sum of lost wages for a SWE following a false SA allegation, for example, would be quite large even in the absence of punitive damages).

That's the only point I was making.

In the US, getting someone convicted of libel or settling for it is pretty solid proof. It means the person being libeled likely could prove that the person making the claims knew they were false and maliciously spread them anyways.

Proving actual malice is only a requirement if the plaintiff is a public figure per New York Times v Sullivan. People who are not public figures have a lower burden of proof, though it is still substantially more stringent than in the UK.

This is an interesting theory. Is there evidence that it's happening? Is Big SEO unreasonably effective at lobbying the Justice Department?

I wasn't responding to the article; I was responding to the claim that Google's results are bad because of all the SEO. It's a claim I've heard from Google apologists including some people I know at Google. I think it's nonsense both for the reasons I stated and for the reasons enumerated in the article.

I'm glad you shared this.

My priors before reading this article were that an uncritical over-reliance on ML was responsible for the enshittification of Google search (and Google as a whole). Google seemed to give ML models carte blanche, rather than using the 80-20 rule to handle the boring common cases, while leaving the hard stuff to the humans.

I now think it's possible both explanations are true. After all, what better way to mask a product's descent into garbage than more and more of the core algorithm being a black box? Managers can easily take credit for its successes and blame the opacity for failures. After all, the "code yellow" was called in the first place because search growth was apparently stagnant. Why was that? We're the analysts manufacturing a crisis, or has search already declined to some extent?

What I don't understand about this explanation is that Google's results are abysmal compared to e.g. DuckDuckGo or even Brave search. (I haven't tried Kagi, but people here rave about it as well.) Sure, all the SEO is targeting googlebot, but Google has by far more resources to mitigate SEO spam than just about anyone else. If this is the full explanation, couldn't Google just copy the strategies the (much) smaller rivals are using?