simiones

This is precisely why (programming language) types are poor model of physics units, despite often being touted for this exact use case. 3m is not the same thing as "the value 3 of type meter". It is the multiplication of the dimensionless scalar 3 with the special "m" constant for meters.

That's why pow(3m, 2) = 9 m^2, and not `the value 9 of type meter`. Of course, you can define the type `square meter` as well, and define `pow -> meters -> positive integer -> square meters`. However this quickly becomes overwhelming once you start doing more complex expressions with multiple types. What is the type of `pow (3kg^2 * m/s, 3/2)`?

Edit to add: also, there is a simple fact that "sin(pi/2 kg)" is just not defined, in programming or math or physics or any other useful system. It's definitely not 1kg, just like sin ( (pi/2) * 2) is not sin (pi/2) * sin (2).

An important nitpick: Wikipedia requires reputable secondary sources, not primary sources. You can't cite a politician's personal social media, for example, you have to cite reporting on those statements. This is part of Wikipedia's defense against novel research - it's supposed to be a record of what is commonly known, not a replacement for, say, Reuters or arxiv.

It has been happening since people have existed. It used to be called propaganda before these, and rumor spreading since before the written word even existed.

> All those methods are applied with the specific goal of improving output quality and are applied to the extent that they do this

Yes, that's the goal that was used, but they are quite simplistic and crude methods, not some specifically designed function, with carefully fine tuned parameters or something. So, if a basic function like top_k can improve model utility, it's not impossible to imagine that watermarking could also happen to do so, or at least not have a significant negative effect. So whether the effect is deleterious or not is an empirical question, not something we can assume ahead of time.

> You're basically saying that if we train LLMs on watermarked text they'll be really good at producing text that looks watermarked

No, you're misunderstanding how the training works. If we train the model's output so that it minimizes the error function after the watermark is applied on it, the model will learn how to produce the best output it can given the watermark. It will produce better text that happens to be watermarked, not "more watermarked text". Same as if you train the model on minimizing `top_k_error(input) = |top_k_sampling(model_output(input)) - desired_output(input)|`, the model will learn to produce better output under top_k sampling, not learn to produce output that's "looks more top_k".

My main point is that sampling with a modified distribution compared to the one produced by the model is already being done, and it is generally found to increase quality, not decrease it. So there is no reason a priori to assume that the watermarked distribution would be lower quality than other schemes for altering the "raw" output distribution (such as top P, top K, temperature, etc).

My second point is that the training of a model by definition maximizes the fitness between the final output function and the training metrics. So, if the model is trained with the watermark applied, the training process will minimize the function `model_error(input) = |watermarked_sampling(model_output(input)) - expected_output(input)|`, by definition. This means that a model trained in this way will perform better when sampled using the watermaked_sampling method than if using, say, top_k sampling.

No one uses a pure random function over the whole probability distribution described by the LLM's output. For example, there is exactly 0 probability that the chosen next token by any common API or even local LLM runner would be a token whose final value is "0.0001" if there exist at least K tokens whose value exceeds "0.7".

Also, as long as the same sampling strategy is used during training as the one used during inference, then the LLM will actually do much better with the biased sampling strategy than it would with a fair one - because that is what it was trained to optimize.

An arena is a type of allocator that allocates efficiently and never deallocates (`arena.free(ptr)` is a no-op). That is basically what defines it. Since there is no free(), doing allocation via a simple pointer bump is the most natural implementation.

Not sure if they tried to do this in the UK, which I believe has very different standards to the USA for what it means for something to be a legitimate religious practice. In the USA I believe they actually succeeded in obtaining various religious exemptions, such as wearing a collander on their head in driver's license photos.

I'm personally atheistic, and thus have no religious practices whatsoever. However, just like it's legal to build churches, to ring church bells, to wear visible religious imagery, it's also legal and highly protected to talk to others, even strangers, about your religious beliefs (or lack thereof).

Even this is too simplistic. While C compilers normally use the OS ABI, there is nothing that requires them to do this, and other languages don't. Of course, when you need to call functions provided by the OS, you have to do so following the OS ABI; but calls between functions written in your own language, even in different libs, don't need to follow this same rule.

I'm pretty sure that would fall under freedom of religion rights, and it couldn't be banned. Especially as proselytizing is a well known and well established religious practice.

I think they meant "unknown numbers" as "numbers that are not saved in my phone", not hidden numbers.

"Solution" seems rather euphemistic for making people literal slaves of AI.

A change can introduce regressions in any large project even if 100% of unit tests pass. Unit tests test individual units, regressions can happen at many levels. Especially if we treat performance degradations as regressions.

It is a well known fact that projects with unit tests never have regressions.

Well, even that doesn't really work, because optimized allocators typicallly don't allocate exactly the amount that you ask for, they allocate some larger chunk to help with both alignment and fragmentation.

So, most likely, there are two sizes in reality: the size of your user data that you care about; and the size of the memory chunk in which your user data resides, that free() cares about. So, unless you're willing to go for an API like this, you can't rely on the consumer:

  int* ptr_to_dest = NULL;
  size_t size = 10 * sizeof(int);
  size_t allocated = malloc(&ptr_to_dest, size);
  if (allocated <= 0 || !ptr_to_dest) {
    //handle allocation error
  }
  //... use ptr_to_dest, size
  free(ptr_to_dest, allocated); //careful not to pass size here!

> The US can win the opening military phase pretty decisively.

Can they? Iran seems to have destroyed them pretty decisively, after the literal first day of casualties (which were happening while Iran thought it had just reached an agreement that would prevent a war, per the Omani negotiators at that time).

But the point of the conjecture is not that "some equilibrium price exists and can be reached". The whole point is that even monopolists will actually sell this type of good directly at the marginal cost, at the fix point. That is the whole prediction; and/or that no one would buy if they did otherwise.

People generally hate to see scams stealing money from naive marks. Both SpaceX and Tesla are obvious scams at anything close to their current valuations (despite having real, functional businesses in there! ), that Elon keeps pumping up with ever more ridiculous claims (50 million humanoid robots by 2026/7! Tesla Roadster out by 2024, pre-order now! Starship point-to-point for business travel by 2030! Datacenters in space!).

Why 50,000 years of technological development and not 50 billion?

> (False that life dies in minutes in space; plus the engineers can invest in even greater error correction than radiodurans.)

What form of life can actually survive (that is, continue living, not just go into some dormant state from which it can later resurrect like a tardigrade) in space for more than a few minutes?

Not in rocky bodies, except maybe for Oxygen (which is commonly found in the form of oxides, very rarely as a gas). Carbon and Nitrogen and Hydrogen and similar elements are mostly found in gas clouds and star that are not really conducive to any form of life or even really fit for extraction by a probe. Maybe some gas giants could be targets for a process of this kind?

> A lot of computational power can be thrown at the problem in this time.

We know plenty of actually very simple problems that we would not be able to solve from now till the Sun explodes even if we all of the computers on earth were only doing this. Say, solving a worse case instance of the Knapsack problem for n > 250 items easily fits into this.

If you want to be precise, the complexity class EXP best fits my description (though NP-complete would also be enough). That is, if discovering the required technology to build a self-replicating interstellar probe requires solving an instance of an EXP problem that is not ridiculously tiny, then we would likely never be find out how to build this probe. And no amount of AGI would change this, in this case.

I'm not suggesting that I have any reason for this ridiculously precise scenario to be correct. Just that some technology being allowed by the laws of physics should not be taken to mean that it is actually implementable in any imaginable amount of time, even if we were to assume singularity-style AGI.

I hate this assumption that many sci-fi enthusiasts seem to make, that as long as something is not ruled out by currently known physics rules, it doesn't matter that we have no idea how it could be built, there will be some way in some plausible future.

When we see currently insurmountable problems in creating a piece of technology, it's absolutely possible that we'll never be able to build it. Even if it is theoretically constructible, there is no reason to believe that the way to build it would be found before, say, the sun runs out of hydrogen.

The thing is, while the universe is full of metals, it's not that full of the materials needed to sustain life (as we know it, at least). You can find metals and other inorganic compounds on virtually every asteroid, moon, and planet, and many comets even. But water and nitrogen and carbon are significantly rarer.

Plus, life can't survive more than a few minutes in space without metal encasings and electronic life support; whereas metal alone only requires life at a much longer time scale. So, while it may be possible to build a fully inorganic self-replicating fleet, it's certainly impossible to build a fully-organic one with any technology or chemistry we know about today at least.

Per the article, the NTHE "movement" are doom sayers, not doom pushers. That is, the NTHE people are people who predict that humanity and most complex life on Earth will be killed off by global warming by 2030 (used to be 2026), through a somewhat convoluted but unavoidable series of consequences of the polar ice cap melting completely (in their theory). There is no indication that the NTHE people want this outcome to happen, it's just the outcome they think is already unavoidable.

Now, the author's position is both more and less nihilistic. The author also seems to think that human extinction is unavoidable, but he thinks the timeline is much more dragged out. He also seems to think that this is actually the preferable outcome at this point, given that global warming is our own fault, and that the only thing we should do is try to make it so that the planet is as well as it can be after we are finally gone.

If I need to take notes by hand, I definitely prefer cursive writing (though I have my own variations & preferences, for example I write almost all capital letters in the print versions). I never saw a reason to learn to draw a more type-like font by hand, so that definitely feels much worse whenever I decide to do it (e.g. for filling in certain forms).

Of course there are, even among the canonical gospels. The length of Jesus' ministry is uncertain (1 year in Matthew, Mark, and Luke, but 3 years in the later John). What Mary and Joseph did after his birth also conflicts among the gospels - according to Matthew, they fled to Egypt, and only returned after Herrod had died (4 BC) while according to Luke, they went to the Temple in Jerusalem to cleanse Mary, which would have happened 40 days after the birth. Some of the identities of the disciples differ between the Synoptic gospels and John (particularly, there is no Bartholomew, no Matthew, no James son of Alphaeus, and no Simon except Simon Peter). The various miracles and acts of Jesus, the exact contents of the sermon on the mount, the parables that he shared - there are many other differences between all the canonical gospels.

Of course, they all agree on a broad narrative - Jesus was born of a virgin Mother in Bethlehem, lived in Nazareth for some time, wandered for 40 days in the wild where he was tempted, then returned and was baptized by John the Baptist, collected a number of disciples while preaching in Galilee and performing various miracles, then during Passover entered Jerusalem, came into conflict with the Jewish religious authorities and/or the Roman authorities, had a last supper with his disciples, and was crucified, died, was buried in a tomb, and was resurrected.

This is about it in terms of the major plot points that are common. Major acts, such as expelling the money lenders from the Temple or the resurrection of Lazarus, are not part of the shared narrative - they either occur at different times (in John, expelling the money lenders is one of the first acts of Jesus, at the start of his ministry, not one of his last public acts at the end), or not at all (the story of Lazarus is only present in John).

I would say that there are indeed major narrative differences between the Gospels.

> For our thought experiment, let us suppose that science observes God doing something at step 1 (systematic observation). Now, what is science going to do with it? It's going to throw it out at step 2 (search for a regularity), because there is no regularity - unless God does the same miracle repeatedly.

> But if it makes it past that step, the next problem comes at step 3 (forming a hypothesis). Under current thinking, God will never be the hypothesis. But in our thought experiment, God is actually the cause!

> And even if God were to be the hypothesis, the next problem comes at step 4 (testing). How could you test the hypothesis? "Uh, God, could you do that again, and please sign it this time"? I don't see how you could do the experiment, even in principle.

This sounds very convincing at first glance, but it completely disregards the actual history and practice of science.

In particular, for your insurmountable problems with steps 1 and 4, we actually have lots of scientific theories about singular events that we can't reproduce. Perhaps the most famous and well studied of these is the Big Bang.

If there were some god who could perform supernatural events, and presuming it didn't then explicitly try to obscure them from us, we could still study the consequences of these breaks in regular physics. If loaves multiplied for a feast, we may not be able to reproduce the event, but we could still use scientific means to confirm whether enough material had been present in the area to produce the total observed number of loaves of bread that was reported; and we could study the waste from the feast to corroborate eye witness testimonies with physical evidence (bread crumbs, stool samples, weight of the participants before and after, etc).

And historically, for step 3 - even though today gods or God are not commonly accepted hypothesis, they/he were for a long time in the history of science. However, every time something was thought to only possibly originate from supernatural causes, a natural physical/biological mechanism was later found. After this happened enough times, scientists started disfavoring this hypothesis, on the basis that it was not turning out to be useful or necessary.

Not really. Sure, sometimes long established archaeological conclusions get overturned by new digs, new sources, or just new methods or new evidence of similar events in other places and so on.

But one very important thing to add here is that there is no bias against the naturalistic narratives of the Bible in mainstream archaeology / Bible scholarship / historiography. Huge swaths of the Bible are considered firmly historical, with many important Bible characters being well attested to and many events in the Bible being accepted as historical facts. For example, the prophet Elijah and King Ahab are widely considered to be real people. The events of the exile to Babylon are (now) considered as well to be real historical events.