HNHacker News
TopNewBestAskShowJobs

Kinrany

4,362 karma · joined September 30, 2017

hackernews at kinrany.dev
submissionscomments
Kinrany··on Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint
I disagree completely and I'd be happy to be convinced otherwise
Kinrany··on Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint
This sounds like the same kind of error as writing "0.10 cents" because it's less than a dollar when the number is in dollars regardless of how big or small it is
Kinrany··on I Don't Like LLMs
It's common to talk about complex machines exhibiting unexpected behaviors as if they had minds of their own, starting with cars. There's no reason to stop doing that with AI when they do that more often than any other kind of machine.
Kinrany··on AI safety is mostly a sex cult
GP isn't claiming to read the author's mind, they're very clearly claiming that one definition matches the author's actual words.
Kinrany··on AI safety is mostly a sex cult
"Mostly" in the title implies otherwise.
Kinrany··on AI safety is mostly a sex cult
> Can "ASI will inevitably kill us all" be falsifiable before we create ASI?

No, but there are decent proxies, like "Alice who is X standard deviations smarter than Bob will win >50% of the time at game Y even given a disadvantage that would make her chances <5% if she was playing against someone equally smart".

I've thought about this analogy for less than five minutes, so it's far from perfect. Point is, ASI being made of silicon isn't very central to the problem. There are lots of analogies that are falsifiable, and "ASI will kill us all" can be decomposed into a simple and bulletproof composition of falsifiable beliefs, even if it is too specific to be itself directly falsifiable.

Kinrany··on Backups Aren't Simple
There should be a standard for backups, one that makes it possible for a (potentially distributed) system to talk to a backup system and to save or restore a snapshot.
Kinrany··on Ask HN: AI Panic, you buying it?
So we should probably have a contingency for that, just like we plan for other bad things like natural disasters that may or may not happen, right? Right?
Kinrany··on Ask HN: AI Panic, you buying it?
It's very clear that people will absolutely put the silicon brains into the most powerful harnesses they can come up with just to see what happens.
Kinrany··on Ask HN: AI Panic, you buying it?
There are at least a few bad guys in the world whose kill switches do not seem to be working even in the absence of futuristic tech.
Kinrany··on Ask HN: AI Panic, you buying it?
The demand for human-level intelligence is obviously at least the same order of magnitude as all office jobs combined.

I'm personally skeptical of LLMs getting there without at least one new invention. But it's also empirically clear that the pace continues to increase, even if we ignore that exponential growth is the default in economics in general.

Kinrany··on Ask HN: AI Panic, you buying it?
What does "based on reality" mean to you?
Kinrany··on AI 'kill switch' may need to be mandatory, Anthropic co-founder tells BBC
The analogy breaks down when the wolf in question may very well eat the whole village: even if the boys who cry wolf are right half the time, every surviving village would have a history of no wolf ever coming to eat them
Kinrany··on When anyone can build software, who decides what not to build?
TFA responds to exactly this: the state that accumulated in those 3-5 years is what cannot be thrown away.
Kinrany··on Measuring the sloppiness of code
Number of iterations solved correctly, on a very large set of iterations, seems like a very good metric. Better than anything else because it measures what we actually care about, not some proxy.

The only caveat is that it's the same model doing an iteration and then using that iteration as a starting point for the next step. So the model is allowed to write absolutely insane solutions, as long as it can read them back, even if no one else can.

One thing that could be done is to use a separately developed baseline coding model B to evaluate the outcome of each iteration. For model under test X to pass an iteration, not only should it be able to solve starting from the previous solution, but so should B, starting from X's previous solution.

Kinrany··on Investing in mothers? The long-run impact of a universal child care
Children being cared for solely by their parents isn't "natural": children used to be raised by entire villages
Kinrany··on When did French fries become the standard side to serve with a hamburger?
I'm curious why they lost to McDonalds
Kinrany··on Aging brains blend memories together instead of just forgetting them
I don't know the mechanics! Just the observation that memories recalled many times tend to change more.

It kind of makes sense in that we do not retain the raw information, it's always interpreted, and it needs to be updated to be in line with other things we know.

Kinrany··on Sony makes bold claim about game ownership
That is, it's bad that these situations happen and that's why they should be regulated away when possible. But if you can't personally change the status quo, the next best thing is to accept the game and respond in kind.
Kinrany··on Sony makes bold claim about game ownership
> And, if you don't agree with the terms, you shouldn't click "I Agree."

Nah, it's perfectly reasonable to click it if you know you're being transparently lied to in a way that won't hold up in court.

Kinrany··on Sony makes bold claim about game ownership
In the rest of the civilized world adults are expected to have common sense and be truthful when it comes to formal transactions.
Kinrany··on Sony makes bold claim about game ownership
As long as we're in copyright world, it ought to be structured as three separate parties:

1. Licensor, the company who makes the game.

2. Licensee, the person buying a digital copy of the game.

3. Safekeeper, Steam or another platform that stores the digital copy on behalf of Licensee. Similar to how depositories store securities and guarantee that they can be neither lost nor duplicated. For a separate fee perhaps.

Kinrany··on Aging brains blend memories together instead of just forgetting them
We do know that memories get rewritten every time they are remembered
Kinrany··on The creator of Jujutsu has joined ERSC
It doesn't require that everyone switch at the same time, unlike other VCSs.
Kinrany··on Good Culture Is the Biggest Productivity Hack, Not AI
Yes, my objection is to the use of "market forces", which are widely assumed to favor efficiency. What we're talking about is, pedantically, something else, something like "capital allocation game theory". A different game that is not known to favor anything in particular, as far as I know.

TFA argues that good culture is important for productivity/efficiency, so any counterpoint would need to come from a game that does say something about efficiency.

Kinrany··on AI-Written Code Is Still *Your* Code. Are You OK with That?
You can write and calculate much faster with a typewriter and a calculator than you would do the same by hand, and you can use both even if you don't know handwriting or the algorithm for calculating logarithms by hand.
Kinrany··on AI-written code is still your code
That's not too different for crediting e.g. a typewriter as the tool used to author the text. That is, not ridiculous, but not new.
Kinrany··on AI-Written Code Is Still *Your* Code. Are You OK with That?
You could ship code without understanding it, it just needs to be treated as throwaway, unreliable and potentially malicious. This is OK for some tasks, e.g. best-effort scraping of a frequently updated website that's actively hostile to automation
Kinrany··on Good Culture Is the Biggest Productivity Hack, Not AI
Infinite growth isn't a myth in general, the world economy does grow over time and for obvious reasons. But of course this doesn't necessarily translate to individual companies growing at the same rate all the time.
Kinrany··on Good Culture Is the Biggest Productivity Hack, Not AI
Right. So these are all examples of market forces being irrelevant. This isn't the market showing that culture doesn't matter: when the winner is predetermined, culture or anything else couldn't _possibly_ matter, as long as it is not so critical that getting it wrong kills the company in the most literal sense.
← PreviousPage 4 of 34Next →