mpoteat

I agree that while it might be fun / efficient / impressive to learn the various methods of quickly specifying mathematical diagrams using symbolic entry, it distracts from the underlying message of the lecture.

Experiential salience is important for learning. It would be like copy and pasting code without typing it out yourself re a tutorial.

It's well considered by a lot of people in the industry that a prospective's grades and even which school they went to aren't a good predictor of skill.

Agree. A one off tool that detects pirated links seems odd to pay for.

Either a monolithic tool for handling moderation automation in general seems good, or better yet a marketplace where I can buy automation plugins for Discord in general, both for moderation as well as other administrative or entertainment tasks.

I am unsure whether or not GPUs are good at work involving arbitrary precision numbers. My intuition was that they excelled at doing massively parallel operations for which accuracy is not a high concern.

I would probably look into using some sort of large integer sieve if I were trying to search for an example. Thinking about it though, I am even unsure whether it is decidable that N is widely digitally delicate.

The generation quality observed will reflect the overall quality and personality of the prompt the transformation model receives.

I would've been more interested to see the output associated with high quality internet content rather than as the article puts it, a "top ten listicle". Worth considering what the most likely best sources of amusing pranks would be, perhaps anecdotal stories? Likely better than SEO content marketing stuff.

Why would you assume that years of exposure isn't terabytes of data?

At least we can train language models in hours or days rather than years. If you view it that way we've made quite a bit of progress!

This made as little sense to me as it did when I was talking the Finite Element Methods class during graduate school.

Still don't quite understand why you can't just use Runge Kutta methods to numerically solve these problems. I became quite good at manipulating the symbols to derive variational solutions while having absolutely no idea what any of it meant.

This has been widely discussed on this site and is a pretty contentious topic.

For my view, consider a very bright problem solver, who bursts into tears every time a person watches or criticizes their code. Do you think that's relevant to performance?

Certainly is a balance, but studies like this usually wilfully ignore all of the other aspects of software work, that aren't sitting in a chair alone writing code.

A much more salient and important point is generative models relationship to music. Right now, covers are legally protected as derivative works. I predict commercial usage of generative music to avoid royalties within the next decade.

Philosophically, I think that the process of learning - whether by a human or simulacrom - should be protected. That is, if I write a book, I don't need to seek the permission or pay the authors of every book I've ever read. To me, this is the only sane result for a non-dualist materialist who believes neural networks are in principle isomorphic to human minds.

As a matter of law, I believe it will have to be determined on the case law level some point. I'm not a lawyer and this is not legal advice.

If true, what implications does this have for P = NP? My intuition for a long time has been that P = NP only in the limit of an "infinitely sophisticated" algorithm. Such that as we do more research (or I suppose train larger models), we get closer and closer to the polynomial ideal.

Certainly depends on your utility function and the degree to which the ends justify the means. Is it good to steal from an apothecary to treat your ill wife? etc.

The underlying issue with Pareto optimal scenarios like this is that we end up trapped in a local minima.

I think in a particular case it's important to look at the context. For this, it likely depends on the scale, so in small town Georgiaville it would prevent innovation, but across the world scope its better* for ideas to be free.

*given most mainstream progressive utility functions and non-zero-sum philosophies

The theory is that nobody will invest in e.g. researching how to build a complicated lumber cutting machine if they don't have protections against their competitor copying the results.

Let's say this research costs 1 million dollars, and the company bankrolls this via a loan. They build it and are successful, and begin paying it off.

Their competitor copies their idea, and exploits the fact that they have more capital and less debt to buy up more of the market, and end up the winner.

So, under a zero protection scheme it is optimal to wait until your competitor invests in research, and then just steal it. Which is what happens on a geopolitical level where there is no feasible enforcement possible.

If freely wired neutral networks are DAGs, I wonder how cyclic graphs of neurons behave, or even if that model would be theoretically meaningful, or computationally feasible.

Thank you for keeping Hacker News a clean and a productive environment for rational discourse. I see you a lot across posts and I'm not sure how much automation you use, but it's clear that you put in an extraordinary amount of effort in tending this garden.

Thanks dang.

Sometimes I feel like I'm trapped in a bizarre Escher painting of deployment deployers stacked on an infinite pile of bespoke libraries that all clamor and scream that they are the "one true solution to reduce all complexity".

Why not just use ask cli? Don't answer this I might just start screaming and never stop

I think the idea is that if we continue to print money, the dollar will eventually collapse as the world's reserve currency, and nations will use the yuan or euro instead - having a material effect on American persons life outcomes.

I know increased national debt doesn't directly imply money printing, just throwing some thoughts out there.

Enter a JSON wrapper format for specifying assembly applications, compete with a NodeJS transpiler toolchain a la Babel.

Seriously though, it's been a while since I've had to work with assembly or IDA but JSON is great as an interoperable data format. Sure, it's not maximally efficient, but what needs to be?

Correlation is soft evidence for causation if one signal leads the other. Correlation is one of the only ways we have as humans to understand phenomena through observation.

From where I'm standing the article makes a cogent thesis as to why price is driven by the online activity rather than vice versa. After all, you wouldn't expect the success of a business to be predicated on its listing price - the tail wagging the dog so to speak.

"Red motivation" is definitely the sort of apt polite allusion people would use refer to that subject matter. Chilling!

My push back comes from the claim that DLLs are intrinsically different than npm code when it comes to security / complexity / etc.

In the bad old days you would find some random GLUT binary off of the Internet and stick it in your code, spend hours setting up linking, etc.

Pretending that was any better than what we do now, or pretending apt and npm are different in terms of trust and security is putting on rose colored glasses.

I've followed up and verified that indeed, most all VR displays have a static focal distance. Thank you for taking the time to correct my and others misconception on the topic!

In the future there may be an integrated display technology that physically cannot display an arbitrary combination of pixels at high speed, but can only do so for the space of "images which humans are likely to see" which I suspect is an unimaginably smaller possibility space.

I suppose something like a neural representation of a scene that is directly rendered on the underlying hardware, without first being converted to a raw bit stream.

It is obviously saliently true just via direct experience. Your eyes focus at different distances when looking at 3d objects in VR. You can feel it!

I am not a physicist by any means, and I am not sure what a "light field" exactly is in terms of engineering, but as a everyday skeptic I am generally rather wary of anything Magic Leap puts out.

I just wanted to say that this resonated with me, as a former grad student involved in protein research who is now doing dev work. I hope you're doing well these days!

Isn't modern human intelligence widely considered to be the result of evolutionarily optimizing complex iterated social games?

I respectfully disagree - I believe this is a philosophical viewpoint that shouldn't be presented as a straightforward truth.

Unless you are a dualist, I would say that it's reasonable to view that it is in principle possible to produce an artificial network accurately emulating the function and behavior of a human brain.

If you are a dualist, then there is no further discussion to be had as we are very unlikely to ever be able to prove anything like the existence of a soul.

Apologies if you were speaking to some more subtle nuance that I was unable to pick up.

From my perspective, the only fool proof way of removing all audio watermarks from a conversation is to run individual speaker detection, STT detection, and voice cloning algorithms to "recreate" the conversation from scratch.

Even things like background electrical hum have been used in audio forensics.

Does this mean if a sports fan comes into my forum about programming and starts making sports threads, I cannot moderate that away, or else face legal action? Does it also mean HN could face legal action according to its current rule set?

If only we had a decentralized way of allowing people to host their own content and draw their own users, without having to rely on a centralized platform... Of course, I'm talking about the Internet itself. Mastodon at first glance it's just an abstraction over an already decentralized system of content.

I for one find this fascinating, please work on this!