> I think the distinction with the Chinese models (or with any of the other models) is that they aren't particularly vocal and obviously active about their politics
Try asking Chinese models about Taiwan independence or Falun Gong or the Dalai Lama or Tiananmen or the Hong Kong national security law or ASIO’s investigations into Chinese interference in Australian politics
> Feels like a distinction without a difference. What is any intelligence but a sum of its knowledge?
In humans, there is a standard distinction between fluid intelligence (ability to solve problems in the absence of background information) and crystallised intelligence (having more facts and learned skills in your head)
I think it will make less of a difference than you think.
Do you think people who have 0-2 kids would be signing up to have 10+ if only we could grow babies in vats? I doubt it. By contrast, >10 kids is actually quite common in communities like Kiryas Joel.
Most of the expense of children isn't the physical burden of pregnancy, it is the time and effort and expense in raising them. Making pregnancy easier won't make a huge difference to the birth rate, because for most women the burden of raising the child weighs much more heavily on her mind than the pregnancy itself does. Babies in vats would be an enticing option for wannabe single dads, gay male couples, women with medical issues – but the actual increase in "births" from those groups taking up that option is likely to be relatively modest. And natural biological reproduction is extremely cost efficient, due to millions of years of evolution optimising it – even after growing babies in vats becomes technically feasible, it is going to take a lot longer before it can compete on price with doing it the natural way.
Of course, you could outsource the raising of your kids to AI robot nannies – but, that raises serious concerns about how well-adjusted they'd be, about the ethics – and apart from that, who would want to? Most people want to have kids because they actually want the emotional relationship of being their parent – absent religious conviction pushing them to do otherwise, they'd rather have a few they can invest in as individuals, than a large number with most of the actual work of parenting outsourced to AI.
> Of urban college educated professionals, do the conservatives still have more kids than the progressives?
Look at Amy Coney Barrett - Supreme Court justice, before that a constitutional law professor then federal appeals court judge, five biological children plus two adopted, and an ultraconservative Roman Catholic.
I think you’ll find the “women with successful professional careers and >=5 kids” demographic has a strong conservative skew.
> No one has any good ideas about the root causes nor the solutions.
I think if you look at ultra-Orthodox Jews and the Amish, the answer is obvious - the easiest way to maximise fertility is to convince people that God exists, has commanded them to reproduce, and to minimise engagement with cultures that don’t share that commitment; the root cause of low fertility is most people in the current population don’t believe that; the long-term solution is their demographic replacement by people who do.
Any population decline is likely to be temporary; by the 23rd century, I think people will be back to worrying about overpopulation
> Maybe I'm wrong, but it looks like the paper just has a misleading title, and this is about having children, not fertility.
I think the confusion is that “fertility” means different things in medicine vs demography
In medicine, fertility is about your ability to have biological offspring, not about whether you personally choose to make use of that ability
In demography, fertility is about how many children people actually have. The completed TFR (total fertility rate) of a population is the lifetime average number of children per a woman.
From a demographic perspective, the vast majority of differences in fertility are due to socioeconomic and cultural factors, medical infertility and the availability of treatments for it makes only a very small difference to overall birth rates
> The right will successfully reproduce their political alignment sometimes, of course, but they also effectively act as the breeder population for the left.
I think the right has been evolving higher memetic immunity, which is causing this strategy to become less effective over time.
Increases in homeschooling and private religious schools – school vouchers in the US really help with that. Reduced rates of cross-political friendship, dating and marriage. Increased geographic sorting based on ideology. Social media echo chambers. The "right-wing media ecosystem" (see e.g. Libs of Tiktok) is a lot more engaging than it was 40 years ago. Internet filtering (some religious groups pressure even adults to install it.) Increasing political pressure on universities to moderate their politics reduces their effectiveness at transmitting left-wing politics to students, meanwhile right-leaning alternative tertiary institutions are growing.
Also, odds of political defection is partially determined by personality traits, which are partially genetic. This creates selective pressure to reduce the frequency of defection-promoting alleles in right-leaning populations across generations, which is a genetic rather than memetic factor predicting that conservative retention rates will rise over decades to come. See https://pmc.ncbi.nlm.nih.gov/articles/PMC3125629/ which discusses this with respect to genes for religiosity, which is heavily overlapping with (albeit not quite the same as) political conservatism.
> I mean when your IPO is kind of hitting the ground face first, it might be better to concentrate on that than on your hobbies.
It is still trading above the IPO price. It could have gone better, but other IPOs have done significantly worse. It succeeded in turning Musk into the first ever US dollar trillionaire, and while he has lost that status again, he’s not far off winning it back-odds are decent he will, sooner or later; and I doubt anyone else will be joining that club for a while.
> And if you don't know which parts were generated by LLM, you can't safely reuse the code.
I speculate this could be a real issue in future copyright infringement lawsuits.
The plaintiff bears the burden of proving that the code they claim is copyrighted by them actually is copyright. If it is known that large parts of it were generated by LLM, they’d need evidence to demonstrate sufficient human input to establish copyrightability. If they’ve kept highly detailed traces of the development process, that could be rather straightforward; if they haven’t, it could be really difficult.
Now, that’s true in the US, which never accepted mere “sweat of the brow” as a basis for copyright; the UK courts have, and most of the Anglosphere follows the UK on this more than the US.
The other factor: when dealing with an (almost) trillion dollar corporation, even if you’ll win the legal argument, they may bankrupt you with legal fees before the argument is ever properly heard.
But I suspect the precedents on this topic are going to be established by lawsuits involving far smaller actors.
(IANAL and I speculate only for myself, not any present, past or future employers.)
Alternative take: if a SPCX-TSLA merger proposal is publicly announced, that will create market enthusiasm (quite possibly irrational) about “unlocking synergies” which will temporarily pump both stocks, thereby making Musk (and all the other insiders) richer, even if only temporarily.
Plus, the two firms already cooperate heavily, and Musk wants them to cooperate more, but being two separate public companies adds a lot of legal friction to that cooperation (each board has to review and sign off on things independently) and legal risks (shareholder lawsuits alleging they are cooperating in ways contrary to shareholder interests)
So I think, from Musk’s viewpoint, it is a very logical next step, which means it probably will happen sooner or later
> The implementation does not often differ run by run.
If you use a cluster, or even multiple clusters, and they have non-identical hardware, then two consecutive runs could end up being routed to nodes having different GPU models with slightly different floating point behaviour, or even software differences (e.g. newer GPU offers some feature usable to speed up calculations which older model lacked; same code can use the feature when it is available, fall back to slower alternative if it isn’t). The larger your scale, the greater the odds it will happen
> That is not a problem for LLMs, because in practice floating point inaccuracies (in particular after exponentiation) prevent values from being exactly equal
Any two tokens ending up with the exact same logit is very unlikely, but not impossible; and as the number of output tokens grows, the odds that it will happen eventually gets higher and higher.
I suppose, to ensure determinism, rank by logit then token ID, so you still have a deterministic winner even if occasionally two tokens get precisely identical logits.
> It would be extremely heavy handed but the administration has sanctioned the International Criminal Court judges
That's sanctioning specific individuals for specific acts they performed which the US claims contravene its interests and those of its allies.
I don't agree with the ICC sanctions, but it really can't be compared with the proposal "sanction any company, even US domestic entities, which use a Chinese-developed open source model".
In fact, I think part of what enables the US to sanction them (under US law) is the fact they are neither US citizens nor residents; if they were US citizens living in the United States, I don't think the President would have the legal authority to impose those kinds of sanctions.
They could sanction Hetzner–because it is a German firm based in Germany. I don't see how they could sanction a US firm based in the US whose owners and staff were US citizens.
Also, the 5th Circuit Court of Appeal decision Van Loon v Treasury (Nov 2024) is relevant–it held that IEEPA (the law used to sanction ICC officials) couldn't be used to sanction the Tornado Cash smart contract system, since open source code wasn't "foreign property" under IEEPA.
> They could ban payment processors from processing payments to any hosts of GML 5.2
Can they actually though? Do they have legal authority to tell a payment processor that it has to block transactions of a legal US company, just because the company is hosting a Chinese-developed open source model? I’m sceptical
And what about companies (e.g. AWS) that let you “bring your own model”?
> GLM export controls incoming? I predict Commerce will force OpenRouter, HuggingFace to take some open models down within the next few months.
I’m sceptical they could find the legal framework to do this even if they wanted to
They have legal authority to (a) prevent export of US goods/services; (b) ban imports of physical goods; (c) ban transactions (including purchasing services or license agreements) with foreign firms
But I’m not aware of any legal authority which lets them ban US firms from running a Chinese-developed open source AI model in the United States, if they are at arms length from the vendor, and aren’t using it for government contracts or regulated applications
Possibly they could order HuggingFace/etc to suspend Chinese accounts. But if someone in the US (or a third country) downloads the model from China then reuploads it to a US server, completely independently of the vendor - where is the legal hook to prohibit that?
> I sort of agree with this in the abstract. The problem is that concretely, these LLMs are being used to decide whether you receive healthcare, government benefits, or whether your job/agency gets cut. So they have already had real world consequences due to DOGE, Insurance companies, and other uses.
But you can't assume that the AI's bias in a policy debate is the same as its bias in unrelated use cases.
For example, it is totally plausible an AI might display subtle bias against minority applicants in hiring, while simultaneously acting as a zealous advocate for affirmative action when asked to debate policy or politics.
Try asking Chinese models about Taiwan independence or Falun Gong or the Dalai Lama or Tiananmen or the Hong Kong national security law or ASIO’s investigations into Chinese interference in Australian politics