Very true. Just the other day, another “copyright is bad” post on the front page. Today its copyright is good because otherwise people might get some use of material in LLMs.
Considering this is hacker news, it seems to be such an odd dichotomy. Sometimes it feels like anti-hacker news. The halcyon days of 2010 after long gone. Now we need to apparently be angry at all tech.
LLMs are amazing and I wish they could train on anything and everything. LLMs are the smartphone to the fax machines of Google search.
> Very true. Just the other day, another “copyright is bad” post on the front page. Today its copyright is good because otherwise people might get some use of material in LLMs.
>
> Considering this is hacker news, it seems to be such an odd dichotomy. Sometimes it feels like anti-hacker news. The halcyon days of 2010 after long gone. Now we need to apparently be angry at all tech.
>
> LLMs are amazing and I wish they could train on anything and everything. LLMs are the smartphone to the fax machines of Google search.
Sorry this such a (purposefully?) naive take. In reality the thoughts are much more nuanced. For one open source/free software doesn't exist without copyright. Then there is the whole issue that these companies use vast amount of copyrighted material to train their models, arguing that all this is fair use. But on the other hand they lock their models behind walls, disallow training on them, keep the training methods and data selection secret...
This tends to be what people disagree with. It feels very much different rules for thee and me. Just imagine how outraged Sam Altman would act if someone leaked the code for Gpt5 and all the training scripts.
If we agree that copyright does not apply to llms, then it should also not apply to llms and they should be required to release all their models and the way of training them.
I think that opens several other cans of worms, but in principle I would support a solution that allows using copyrighted materials if it is for the common good (I.e the results are released fully open, means not just weights but everything else).
As a side note i am definitely not strong into IP rights, but I can see the benefits of copyright much more clearly than patents.
My point wasn't supposed to be that copyright is bad (or that it's good), just that the business logic of fighting the sharing of lyrics is incomprehensible to me.
That aside, I think there's a lot more complexity than you're presenting. The issue is who gets to benefit from what work.
As hackers, we build cool things. And our ability to build cool things comes in large part from standing on the shoulders of giants. Free and open sharing of ideas is a powerful force for human progress.
But people also have to eat. Which means even as hackers focused on building cool things, we need to get paid. We need to capture for ourselves some of the economic value of what we produce. There's nothing wrong with wanting to get paid for what you create.
Right now, there is a great deal of hacker output the economic value of which is being captured almost exclusively by LLM vendors. And sure, the LLM is more amazing than whatever code or post or book or lyric it was trained on. And sure, the LLM value comes from the sum of the parts of its source material instead of the value of any individual source. But fundamentally the LLM couldn't exist without the source material, and yet the LLM vendor is the one who gets to eat.
The balance between free and open exchange of ideas and paying value creators a portion of the value they create is not an easy question, and it's not anti-hacker to raise it. There are places where patents and other forms of exclusive rights seem to be criminally mismanaged, stifling progress. But there's also "some random person in Nebraska" who has produced billions of dollars in value and will never see a penny of it. Choosing progress alone as the goal will systematically deprive and ultimately drive away the very people whose contributions are enabling the progress. (And of course choosing "fair" repayment alone as the goal will shut down progress and allow less "fair" players to take over... that's why this isn't easy.)
Sounds like it was never about copyright as a principle, only symbolic politics (ie. copyrights benefit megacorps? copyright needs to be weaker! copyright hurts megacorps? copyright needs to be stronger!)
More and more authors of nonfiction today make money from a number of channels other than actual book sales. The book only serves as a promotional tool for their personal brand, and is only a collection of previously published blog posts or magazine pieces. This already started due to changing consumer behaviors and declining interest in books, so by the time piracy has come on the scene, the shift had already largely occurred.
Cory Doctorow's works that were released under more permissive licensing still reserved some rights for the author. I believe he used some flavor of a Creative Commons non-commercial license, if memory serves. Point being that the method of licensing his works was still fundamentally based on copyright.
(I think the US copyright system is hugely broken and the social contract needs to be re-negotiated, but I comment here in the interests of facts, not in support of the broken system.)
Recipe blogs are mostly "corporations" even if small ones. Most things you find at the top of Google search results aren't just enthusiastic individuals sharing their personal ideas with you but businesses who work hard to make sure you go to their websites rather than better ones.
The Republicans can end the shutdown today. They don't want to, they're letting it continue and that's their choice. They need 0 Democratic or independent votes to pass the funding bill if they change the rules, which requires a simple majority vote and they have that.
The longer the government is shut down, the better excuse/cover the executive branch has to permanentely layoff and shutdown departments. Democrats are going to lose this standoff either way...
I disagree. I think the fact that Democrats let Trump have a blank check-book for his first term was a massive mistake.
We went from having government shutdowns all the time due to the rampant spending to just covering up the fact that he's blowing through more money than presidents do in 8 years. It leads to a perception that he's actually fiscally responsible.
At this point, either political party pretending to care at all about fiscal responsibility is absolutely hilarious.
The standoff is about a component of the Affordable Care Act that is expiring. Democrats don't have the footing to win this battle - so the longer the standoff holds the worse outcome they can expect. Trump's administration seems to have wanted this shutdown... and Democrats walked right into the landmine.
Can you elaborate on which part you believe isn't true?
The bill already passed the House. The Democrats have a minorotiy in the Senate. 60 votes are required to pass the bill - and Democrats are holding it up, deliberately for political reasons.
Republicans can end the filibuster in the Senate with a simple majority after which they can pass the funding bill with a simple majority. But of course, they won't do that since it opens up other bills to be passed with a simple majority too in the future. So it's not factual.
It’s weird to me that any Democrat would be rooting for the nuclear option right now. After losing so badly last year, after demographic trends continue to degrade their ability to win majorities in either house or in to win in most swing states, and with Trump improving his numbers even in blue states last year. Democrats are currently daring Republicans to eliminate their last resort option to block anything, even though they should not be confident that the DNC will ever be in a position to benefit from this new power later themselves.
I think shutdowns are usually pretty stupid, but I’m not reading for “nuclear option” to be the way this one gets resolved.
While I'd prefer to get the first legislative crack at it, I am extremely convinced that the filibuster is poison to our country.
Congress cannot pass anything meaningful except one omnibus spending bill each year via budget reconciliation, which has arcane rules around flat budget impact after 10 years. Since congress can't do anything, we naturally move more and more of the details of federal governance to executive agencies and executive orders. While not all of this is bad, it has a few horrible effects.
First, the supreme court is more willing to interfere with executive action, amplifying its power as it finds reason to protect actions by the favored party and cancel actions by the disfavored party. Increasing the power of the supreme court shifts federal power towards an unelected branch that is the slowest to adjust to changing voter preference.
Second, as more power within the executive gets concentrated specifically with the president we enable more and more federal action at the whims of exactly one person. This exposes us to the current situation, where Trump is unfettered in how he wields the executive branch rather than guiding it and having its power distributed across the executive branch.
And we've also got the general popular dissatisfaction with congress and the democratic process because they can't get shit done. A country that has an enormously unfavorable opinion of congress is more primed for collapse into anti-democratic governance.
Yes, if we didn't have the filibuster then the GOP could pass all sorts of nightmare legislation. But I'd prefer legislation enabled through the will of the people to this slow collapse into authoritarianism.
> Since congress can't do anything, we naturally move more and more of the details of federal governance to executive agencies and executive orders.
It's really interesting how wildly different this is than parliamentary systems where Parliament is the ultimate authority and the executive's tenure simply ends if they lose the 'confidence' of a plurality in Parliament.
Instead we have a useless legislature as you described, and an executive whose claim to all this additional power is actually quite dubious, yet that executive controls nearly everything. I don't think a swing to the other party changes this, either (not that the DNC is capable of winning elections enough to ever hold the kind of power the GOP now has). I think from now on, the President will rule by executive action, and use creative avenues like rulings from friendly courts to vaguely legitimize this power.
An equally true statement is that republicans didn’t pass a spending bill that could attract 60 votes in the Senate. But in either case, Republicans are the majority in both houses, they write the bills and have the responsibility to write bills that can be passed and signed.
And it’s a small thing, but this is very obviously a hatch act violation and silly me, I feel like the President should be beholden to our laws.
“Congress shall make no law respecting an establishment of religion, or prohibiting the free exercise thereof; or abridging the freedom of speech, or of the press; or the right of the people peaceably to assemble, and to petition the Government for a redress of grievances.”
Did congress pass a law I missed? Government communication isn’t a 1st Amendment issue. When you work for any employer, you are subject to the whims of that employer.
Governments are not limited by the Hatch Act, civil service employees are. And the limitations are mainly around elections, so for example under the Hatch Act a civil service employee may not "use his official authority or influence for the purpose of interfering with or affecting the result of an election", may not "knowingly solicit, accept, or receive a political contribution from any person", may not "run for the nomination or as a candidate for election to a partisan political office", and may not "knowingly solicit or discourage the participation in any political activity" of someone who the civil service employee is interacting with.
The Hatch Act of 1939, An Act to Prevent Pernicious Political Activities, is a United States federal law that prohibits "civil service employees in the executive branch"
>The laws are different when you work for the government.
Which laws? Because every government job I've come across has pretty strict rules on what you can and cannot say in public, or at least they way in which you have to frame it apart from your job.
I think there's two parts of that question. The first is that the law is the hatch act, federal employees may not engage in political activities. But the second thing is, the government is assuming your identity to change your speech and broadcast that speech as if you said it. I'm not a lawyer, but I think that would squarely follow government action to abridge your speech.
> No, you're renting the physical space -- a scarce part of the commons -- from your community.
Property law in the US and most western democracies doesn’t remotely agree with that. Land is not a communal or solely government owned resource, and the govt doesn’t ‘rent’ it out.
Stop paying your property taxes in the US and see how long it takes before the government forecloses. It is effectively rent under a different name. In exchange the government will protect your property ownership rights so that you don't go on vacation and find someone else now gets to claim your home since you weren't there to stop them.
Note: I think this is a good thing and that property taxes are vital to our local communities well-being.
> Just imagine giving sensitive personal, government-issued ID to a corporation to install an app outside Google Play
In Spain, I have to give my NIE (National ID number) and show my government ID just to send or receive a package from FedEx. Why should I have to give up sensitive information just to receive a package?
Hamas to this day still has hostages. People were still butchered and raped by Hamas. The meme that Palestinians are blameless is just not accurate. Not defending either side, just saying this is like the Soviets and the Nazis on the Eastern Front.
> this is not the fault of scientists. This is a byproduct of a severely broken system with the wrong incentives, which encourages publication of papers and not discovery of truth
Are scientists not writing those papers? There may be bad incentives, but scientists are responding to those incentives.
That is axiomatically true, but both harsh and useless, given that (as I understand from HN articles and comments) the choice is "play the publishing game as it is" vs "don't be a scientist anymore".
I agree, but there is an important side-effect of this statement: it's possible to criticize science, without criticizing scientists. Or at least without criticizing rank and file scientists.
There are many political issues where activists claim "the science has spoken." When critics respond by saying, "the science system is broken and is spitting out garbage", we have to take those claims very seriously.
That doesn't mean the science is wrong. Even though the climate science system is far from perfect, climate change is real and human made.
On the other hand, some of the science on gender medicine is not as established medical associates would have us believe (yet, this might change in a few years). But that doesn't stop reputable science groups from making false claims.
If we’re not going to hold any other sector of the economy personally responsible for responding to incentives, I don’t know why we’d start with scientists. We’ve excused folks working for Palantir around here - is it that the scientists aren’t getting paid enough for selling out, or are we just throwing rocks in glass houses now?
Valid critique, but one addressing a problem above the ML layer at the human layer. :)
That said, your comment has an implication: in which fields can we trust data if incentives are poor?
For instance, many Alzheimer's papers were undermined after journalists unmasked foundational research as academic fraud. Which conclusions are reliable and which are questionable? Who should decide? Can we design model architectures and training to grapple with this messy reality?
These are hard questions.
ML/AI should help shield future generations of scientists from poor incentives by maximizing experimental transparency and reproducibility.
Apt quote from Supreme Court Justice Louis Brandeis: "Sunlight is the best disinfectant."
Not a answer, but contributory idea - Meta-analysis. There are plenty of strong meta-analysis out there and one of the things they tend to end up doing is weighing the methodological rigour of the papers along with the overlap they have to the combined question being analyzed. Could we use this weighting explicitly in the training process?
Thanks. This is helpful. Looking forward to more of your thoughts.
Some nuance:
What happens when the methods are outdated/biased? We highlight a potential case in breast cancer in one of our papers.
Worse, who decides?
To reiterate, this isn’t to discourage the idea. The idea is good and should be considered, but doesn’t escape (yet) the core issue of when something becomes a “fact.”
Scientists are responding to the incentives of a) wanting to do science, b) for the public benefit. There was one game in town to do this: the American public grant scheme.
This game is being undermined and destroyed by infamous anti-vaxxer, non-medical expert, non-public-policy expert RFK Jr.[1] The disastrous cuts to the NIH's public grant scheme is likely to amount to $8,200,000,000 ($8.2 trillion USD) in terms of years of life lost.[2]
So, should scientists not write those papers? Should they not do science for public benefit? These are the only ways to not respond to the structure of the American public grant scheme. It seems to me that, if we want better outcomes, then we should make incremental progress to the institutions surrounding the public grant scheme. This seems fair more sensible than installing Bobby Brainworms to burn it all down.
Considering this is hacker news, it seems to be such an odd dichotomy. Sometimes it feels like anti-hacker news. The halcyon days of 2010 after long gone. Now we need to apparently be angry at all tech.
LLMs are amazing and I wish they could train on anything and everything. LLMs are the smartphone to the fax machines of Google search.