Well at least it seems pretty implausible to me that a machine learning model trained to reproduce human text can in principle generate something that is significantly above human text production ability. If the "pea-sized semiconductor brain" is not a surprisingly shallow problem that you can just solve by interpolating existing research, I don't really see the LLM-approach to AI being the thing that makes something like it happen.
>we do use simulated evolution as a standard technique in machine learning.
Well, not really. For large-scale AI models it's almost exclusively some form of gradient-based non-linear optimization. Genetic algorithms (which is just hill-climbing optimization with extra steps) and genetic programming (which is really cool and not well-understood) do not perform all that well in practice and I'm not aware of any notable applications.
The article is "not even wrong". Most of the argument just circles around an inconsistent definition of what "real world" means.
> Remember "imaginary numbers" from algebra? You could do funky stuff with them like take the square root of negative nine. Imaginary numbers are generally useless for the physical world that we experience. However, they turned out to be very useful in the field of electronics. More specifically, they had predictive power in the field of electronics. Using imaginary number math, one can predict behavior of electrons and the accuracy of such predictions in models can be measured.
Like here, I struggle to understand what the point is. Is predicting the behavior of electronics "useless for the physical world we experience"? And that's ignoring the author's apparent ignorance of all the other ways in which complex numbers are useful in the physical world besides taking negative numbers' square roots.
> Any solution that can convince the Germans, the most privacy obsessed sticklers on the whole planet, has my support by proxy. If they think it's safe enough, it most likely is. Almost no other country has seen the dark side of what you're saying here as much as Germany, first with the Nazis and then in East Germany.
As a German citizen myself I argue that the more correct view of this is that the Germans are going back to their fascistic roots.
> Both Brexit and the Trump election have been significantly impacted by this, and it's not even controversial to observe that.
Yeah and most of that online chatter that made that happen originated from adult citizens that would not be blocked by online identification laws.
> Perhaps, but that doesn't mean
that we should not address the elephant in the room - the seriously degrading impact that social media has on our society.
Well if you agree with my point then that would imply that foreign social media influence operations are not actually the elephant in the room. I think a convincing case that these measures would meaningfully improve the political climate has a much higher burden of proof.
> There was no debate because almost no one in (for example) tech circles is even acknowledging the problem, let alone coming up with a solution. Give me a better solution and I would argue for that instead. The status quo is unacceptable.
This "well we must do something" attitude where people are willing to accept anything that promises to solve some problem is how the worst laws are created. It's easy to forget that things can always be made worse too.
> Real life is not anonymous and consequence free either, why should online life be?
In real life you are largely anonymous if you go to any place that is not in your social circle. If I go to any physical store under the sun and pay in cash I remain completely anonymous. The only exception is staff checking your ID for alcohol etc. if you look underage, but that datum is only stored in one persons brain for as long as they bother to remember. Many people would take issue if the clerk noted every persons name when getting carded.
> what's so great about having complete online anonymity anyway?
Because liberal governments can always become repressive governments. Unless you have absolute faith that your government will never do something you have a moral objection to, you can never be sure that anything you are, believe or do will not be censored or land you in jail in the future. Infrastructure that has a minor benefit under a "good" government, but would serve as a major tool of repression under a tyrannical regime should not be built out of principle.
As a thought experiment, ignoring issues with technical feasibility, would you approve if every person is only allowed to leave the house, under a stiff prison sentence for violation, if they wear a shock collar that can be activated remotely by a law enforcement official? That way, if police want to arrest you, they don't need to use violence and they don't have to chase you if you tried to run. It would make the police's job a lot easier, would it not?
> If we can make sure that every real person can only have 1 social media account per platform, and if we can check that someone is an adult, and if (and only if) we can do that in a privacy preserving manner... then honestly, I don't see why I would be against that. I'm ok with being held accountable for what I do online
To the extent that political discourse is shaped by astroturfing, realistically it'll just give a monopoly of influence to whoever controls/bribes the company or entity doing the verification. There certainly would be technical/cryptographic solutions where there isn't some central entity with a master key, but I doubt that it'll work like that anywhere, especially if it requires a citizen safeguarding his own keys.
> I want to pay that price to prevent the severe outside interference we've seen in elections and in our politics.
Outside interference in the form of legal bribes (lobbying) and sometimes less legal forms of corruption has orders of magnitude more sway over politics than whatever the public may effect in elections.
> You and many others might not be, but it seems like you've lost the argument.
It's ridiculous to imply that there was any serious public debate on this.
Also 95% of those 30k lines changed are fully self-contained inside of the aarch64 directory and of the remaining changes it looks like the majority is just adding "aarch64" as another item into an existing list. There are a few core changes that to me look like they could be done in their own PRs, but also core maintainers get to decide if they want to apply bureaucracy to their own work.
I don't understand that methodology in the first place. Does Anthropic even have some kind of somewhat objective definition to measure and judge "memorization"? Is there any evidence that other LLMs are viable tool to determine that?
If we were perfectly rational, then yeah, more data should never lead to worse decisions. However, it's easy to fall into the trap where being data-driven makes you only work on those things that you know how to measure.
While xor eax, eax only uses 2 bytes. Since there are only 8 registers, meaning they can be encoded with 3 bits, you can pack two values into the <Registers> field (ModR/M).
Making mov eax, 0 only take two bytes would require significant changes of the ISA to allow immediate values in the ModR/M byte (or similar) but there would be little benefit since zeroing can already be done in 2 bytes and I doubt that other cases are even close to frequent enough for this to be any significant benefit overall. An actual improvement would be if there was a dedicated 1 Byte set-rax-to-0 instruction, but obviously that comes at a tradeoff where we have to encode another operation differently (probably with more bytes) again (and you can't zero anything else with it).
I would be surprised if modern CPUs didn't decode "xor eax, eax" into a set of micro-ops that simply moves from an externally invisible dedicated 0 register. These days the x86 ISA is more of an API contract than an actual representation of what the hardware internals do.
We have not been shown any evidence that shows anything approaching this scale of people dead, we have not been told on what basis that specific number of deaths was determined and we have not been told why that supposed evidence can not be made public. 30 thousand killed in the span of two days would make this one of the bloodiest massacres in history. I don't see how that would not result in at least a single photo or satellite image of a street lined with dead bodies, of mass executions, of mass graves or simply just the logistics operation required to dispose of this many bodies. The only thing to back up this number is trusting the probably least trustworthy US administration in history.
> This also re-opens a lot of "party pooper" results in mathematics: impossibility of representing solutions to general quintic (fine print: if we restrict ourselves to arithmetic and roots/radicals).
Solving the quintic in terms of transcendental functions has already been done
It's very common when people start doing a new exercise regimen. Muscles can become significantly stronger on a timescale of months, while an equivalent increase in the strength of tendons happens on a timescale of a few years. Once somebody has a few years of training under their belt, muscle mass gains are way slower and the capacity of the ligaments will have caught up and these issues go away. However, with bodybuilders and strength athletes these problems can easily come back once anabolic steroids are involved and give big muscular gains without the same level of adaptations in connective tissues.
Your view on the law seems a bit alien to me. My opinions on what the rules of the law should roughly look like, are largely independent of who specifically is involved in a legal dispute. Sure I guess if Hitler was being sued and the only way to stop him was this lawsuit by Sony, I would probably concede that on balance it's better to have a slightly worse legal standard around copyright. Otherwise, I think having a law that best reflects my moral views and creates the best incentives for society in general, far outweighs how i feel about the plaintiffs.
As for how I arrive on my views, it's obviously not an entirely rational process, but the rules you get from viewing property rights and self-ownership as fundamental seem to lead to the most preferable outcomes to me. If I were forced to adopt a more deontological philosophy, it's also the one that has the fewest obviously absurd conclusions, though not entirely. From this it's, in my opinion, pretty obvious to be skeptical of copyright law more generally (Ayn Rand would disagree) and therefore I welcome any precedent that weakens it.