HN user

planetsprite

682 karma

learnings for machine learning

Posts1
Comments156
View on HN

Hard to weld doors shut on peoples homes if they have shotguns.

This point is always brought up but is never argued to its natural conclusion.

On the surface it seems you are implying that citizens should have powerful guns so they can threaten and potentially kill police officers trying to do something despotic. However if someone did that, the police would just crack down 10x as hard on that individual with more powerful, government purchased guns and arrest them for life. No matter what guns people are allowed to have, it's not a deterrent against despotism on an individual level.

The argument that the point of the 2nd amendment is for the citizenry to defend against government power is ridiculous. Why would a blueprint for a government install protections to make sure it can be violently overthrown, when the rest of the constitution is entirely focused on ensuring the consent of the people is funneled up to government power in a non-violent matter? To argue that that's the point of the 2nd amendment implies the founders didn't believe that the Constitution would be effective in forming a representative democracy.

In the arts, music, graphic design, animation, 3D models, and eventually realistic looking video seem to be the clear next steps.

The key to AI art is that while the universe is infinitely complex, the visual form of most things humans can recognize follow a collection of natural patterns. Anything humans can do can be computed, the only issue is being able to set up the scope for the patterns we seek to replicate. For games like chess, a bit easier, for images that represent label-able things, harder but still manageable. The main problem of making things beyond this is giving a valid scope to the whole problem: what patterns, precisely, do we intend to replicate?

That's because all the top companies in any sector in China are subsidized and heavily supported by the government. The distribution isn't as organic, it's more a matter of in China, crossing the threshold to being the "golden child" and being guaranteed success right after.

One interesting next step to the idea is to indicate a continuity of meaning via the edits. For example, for each randomly selected gif, have the next gif share at least one tag. This allows each vaguely connected gif to create a randomly generated narrative by their connected meaning.

It's hard to get out of that mindset, especially as a young new hire.

In the United States, your job is very closely tied to your livelihood due to high rents, few social support structures, health insurance often tied to your job, etc.

There's a model of thinking taught in schools and universities that teaches individuals to defer helplessly to their superiors, to be subservient to a fault and respect hierarchies as sacrosanct. When someone abuses that hierarchy, one either has to unlearn their programming, or assume the burden of the imposition of value on their psyche.

Peto’s Paradox 4 years ago

Cancer, mathematically, all things being equal, shouldn't simply scale with number of cells, but rather with number of cell divisions.

This differs by species, along with other cell mechanisms that could inhibit cancer that only certain species have evolved.

GPT3 charges for every token read/written. What may be more useful is using GPT-3 not to manually run itself on every row, but to take the task and generate a sufficient function that fulfills the task.

This is true. I had a friend who programmed part of one of SpaceX's satellites, basically working as an unpaid intern. He spent 80 hours a week working on it non stop, sleeping in the lab, all while finishing his undergraduate CS degree.

The same thing with Elon's Hyperloop pod competitions. A completely meritless future technology still saw millions of free man-hours poured into being the creators of the first step into the future. As smart as everyone who participated in it was, the prestige of everything Elon blinded them.

As much as people dislike Musk, he draws excitement and captivation not only to himself, but to whatever he touches. Now for the first time people actually care about the "future" of Twitter, or in a sense, have a vision for its future that isn't defined by its frequent moderation-related controversies.

Is Musk's financial success and unparalleled cultural relevance a science that can be replicated? I mean, as much as people don't like him he has gotten to a financial position where anything he wants seems easily achievable. Is it the fact that he always carries the implied promise of "bold new future that will change your life" more than any other cultural entity?

Obviously Mark Zuckerberg is trying the same thing with Meta, but that's been a failure.

People hate Musk's personality, but no companies occupy the reverence in people's mind as much as SpaceX and Tesla. Even a man for a while much richer than Elon, Bezos, couldn't muster much in terms of cultural relevance for Blue Origin than a PR stunt with William Shatner.

Does Musk have a internal "toolbox" of guiding principles and behaviors that tremendously advantage him in his current position as outgoing tech CEO, or is he just extremely lucky? Is it merely because of his seeming utopian ambitions that make him more than a guy merely obsessed with financial success that he has ended up much more wealthy than if he were a much more grounded, brass-tacks CEO?

Words2Emoji 4 years ago

he caches to redis, I checked the network request on common words.

What do you mean by smart? A well organized moderately smart person is more productive to a company than a disorganized genius. Someone who can hyper focus on their competent domains is more effective than a super smart generalist who spreads themselves too thin.

Fake Books 4 years ago

Same with most AI image generating models. In 10 years 99% of images on the internet will be AI generated. Would it not regressive for the models to train on their own outputs? Should regulation require AI generated media label itself, or is it the responsibility to train detectors which can intuit the difference between the models and reality better than humans can?