advael

25 ms·

421.
by advael 2y ago
Most animals are bad at learning stochastic or even unreliable patterns, even when doing important things like threat assessment
422.
by advael 2y ago
During the cold war and the GWOT, people were profiled based on their race, religion, suspected opinions, or association with people suspected of {commun,terror}ism. A ton of these correlations were spurious, sometimes corruptly motivated,
423.
by advael 2y ago
It's crazy how much institutional decay there is basically everywhere in this era. Like I grew up hearing stories about institutions being crazy effective compared to the ones I actually encountered. Sure, humans have always had folly,
424.
by advael 2y ago
I think most people who say "I have nothing to hide" are just mimicking a slogan they no doubt heard someone they respect or at least think talks well say Another common refrain, especially since Snowden, is just glib defeatism (&
425.
by advael 2y ago
I like this, and it seems a lot of people are thinking this way lately A common theme is that optimization usually does mostly bad things, while maybe arguably from someone's perspective doing one thing really well. I particularly like
426.
by advael 2y ago
I find it really unfortunate that a "security through obscurity" mindset has remained deeply embedded in conventional wisdom despite its well-understood flaws I entirely blame tech companies for this one. While organizations like
427.
by advael 2y ago
I'm not close enough to anyone you'd probably consider famous to claim to know the inner workings of their mind, and you keep adding more weirdly circuitously specified conditions. At this point I'm not sure what point, if an
428.
by advael 2y ago
I think the wisdom that headcount doesn't make a product faster has managed to reach some people, but somehow not the wisdom of don't burn out your people. Like no matter how much you talk about burnout, there is no adequate valve
429.
by advael 2y ago
This is incredibly useful advice that essentially everyone who gets to make decisions in this space hates This article makes sense to me as a developer and an occasional manager of developers and a former student of psychology and honestly
430.
by advael 2y ago
Sure. There are people who can do that. I think it's a hard skill to master but definitely one that can be performed and improved somewhat reliably for people who manage to get the hang of it initially and care to work at it, and which
431.
by advael 2y ago
I mean, that seems to be exactly how he's defining "open web" here, actually. That which is - in the dichotomy presented by these two quotes - "the open web" is free game for any use, and he defines things that use
432.
by advael 2y ago
Like, fair. I like the EFF's work a lot, but I think they primarily function on the level of public policy and sometimes mounting strategic legal action, and I was more asking about whether there were ground level orgs in operation, un
433.
by advael 2y ago
I mean while I agree that CPUs can be hungry, a lot of traditional beefycore proc tech in this time period has gone into energy efficiency and miniaturization whereas a lot of GPU tech has been all about more and more cores. Research and en
434.
by advael 2y ago
I realize it's kind of a game of musical chairs when companies work so hard to shut down various forms of {labor,equality,ethics} activism in tech, but I found myself wondering at the end of this whether there are groups that are activ
435.
by advael 2y ago
I mean if we count autonomous taxis why not the entirety of the recommendation-engine-driven internet, including ad networks, social media feeds, streaming services? Or like object detection and tracking on cctvs? Does AI mean just LLMs and
436.
by advael 2y ago
This article is a great example of a really frustrating thing people are doing, which is pretending that all this AI business started two years ago I get it, GenAI became popular as a consumer-facing product and tech industry PR blitz in th
437.
by advael 2y ago
Generally speaking they are considered "unelectable" and thus mostly exert influence indirectly if at all. Especially in the United States, even many "progressive" causes that have become relatively mainstream (in terms
438.
by advael 2y ago
What, like not project authority? Admit that they are lost, confused, powerless, don't know something, aren't in control? Break the empire society kayfabe? Yes, absolutely. I view this as one of the criteria by which I assess emot
439.
by advael 2y ago
I think most people do that all the time. Projecting authority is one of the most important skills in a world dominated by human institutions, because it's an effective means of manipulating most humans. Sad but true
440.
by advael 2y ago
I both agree that it's pointless to argue about definitions and think you've presented a definition that fails to capture a lot of common usage of the word. I don't think it matters what the dictionary says when we are talkin
441.
by advael 2y ago
On HN? Killing time, reading articles, and getting nerdsniped by the feedback loop of getting insipid replies that unfortunately so many of us are constantly stuck in In general? Slowly dying mostly. Talking. Eating. Fucking. Staring at mic
442.
by advael 2y ago
Well, first of all, the stated purpose of RLHF isn't to "improve model accuracy" in the first place (and what we mean by accuracy here is pretty fraught by itself, as this could mean at least three different things). They ini
443.
by advael 2y ago
Your post and the one I originally responded to are good evidence against something I said earlier. The mere existence of LLMs does clearly change the landscape of epistemology, because whether or not they're even involved in a conve
444.
by advael 2y ago
As long as we're doing unsolicited advice, this revision seems predicated on the assumption that we are writing for a general audience, which ill suits the context in which the posts were made. This is especially bizarre because you th
445.
by advael 2y ago
I think you'll find I'm quite horseshit at optimizing for persuasion, as you can easily verify by checking any other post I've ever made and the response it generally elicits. I find myself less motivated by what people think
446.
by advael 2y ago
This is a great example of a common problem, which I've taken to calling "intent-solution conflation". Garbage collection is intended as a way to improve memory safety in a programming language. It doesn't fully accomp
447.
by advael 2y ago
RLHF to me seems more as a PR play than anything else, but inasmuch as it does anything useful, adding a second LLM to influence the human that's influencing the LLM doesn't solve any of the fundamental problems of either system.
448.
by advael 2y ago
1. Deception describes a result, not a motivation. If someone has been led to believe something that isn't true, they have been deceived, and this doesn't require any other agents 2. While I agree that it's a stretch to call
449.
by advael 2y ago
The problems of epistemology and informational quality control are complicated, but humanity has developed a decent amount of social and procedural technology to do these, some of which has defined the organization of various institutions.
450.
by advael 2y ago
This is a really bizarre thing to do honestly It's plausible that there are potential avenues for improving language models through adversarial learning. GANs and Actor-Critic models have done a good job in narrow-domain generative app

More ›