md_

11 ms·

61.
by md_ 3y ago
OK, fair enough. I guess my point was a little bit broader: anyone who knows anything about The New Yorker will think you're ignorant for having written what you did. Your comment here sort of confirms that. By all means, you're n
62.
by md_ 3y ago
Eh. Agreed re: table stakes. But I think it's quite valid to suggest users should carefully think about the tradeoffs between being subject to legal disclosure and being subject to compromise. Basically, do you trust the FISA courts, o
63.
by md_ 3y ago
> the 30-under-30 thing is Forbes, I have always considered them reputable, perhaps I am mistaken. No, Forbes is a shithole. https://www.entrepreneur.com/growing-a-business/did-someone-... The New Yorker is not a sh
64.
by md_ 3y ago
> Oh that's definitely true; though I wasn't saying bribe; Did you know those "30 under 30" lists, you have to apply to be on those. It helps if you're rich, but you don't technically have to pay. It's
65.
by md_ 3y ago
This. I understood the project as less about the flashy ten-year vision, and more about the same thing that drove Google to build Chrome and Android: not wanting to face lockout from competing platforms. I don't think this has been _su
66.
by md_ 3y ago
I think your intuition that gpt4 could function as you described is frankly highly unrealistic.
67.
by md_ 3y ago
Well, my point was that explicit memory and "skill development" (as one kind of what is known as "implicit memory") are fairly different. No matter how many times you retrain GPT4, it isn't going to learn to drive a
68.
by md_ 3y ago
Oh, sure, all of that. My point is that it's a tough argument to say that there were more wars because of labor displacement. There was probably more political upheaval--that I agree with. But there were a ton of European wars prior to
69.
by md_ 3y ago
I don't think the difference between human memory and retraining a DNN is purely about the latency to retraining. They seem to be fundamentally different things, no?
70.
by md_ 3y ago
I agree that people underestimate the impact of the first two industrial revolutions while overestimating the impact of the third, if that was your point. I'm not yet prepared to agree that the scale of bellicosity in the 20th century
71.
by md_ 3y ago
No. A DNN w/ BERT/nnlm embeddings (I'm between models right now). And a sort of context-based classifier so that it can react to new things you "like" without model retraining.
72.
by md_ 3y ago
I see these takes a lot, but I struggle to take them seriously. As a trivial example: LLMs don't learn new skills on the fly. A key aspect of human behavior--implicit memory--simply doesn't exist for these models. That seems like
73.
by md_ 3y ago
I think it's pretty surprising, and counterintuitive, how powerful generative language models are. It turns out that a lot of intellectual tasks are sort of weakly simulatable by just doing a great job at next-token prediction. I don&#
74.
by md_ 3y ago
I agree that "AGI" and "labor transformation" are two different issues. I do want to pick on what I think is a bit of hype around the potential to eliminate whole classes of work, though. First, in the weeds: I keep seei
75.
by md_ 3y ago
Seems like a long-winded way of saying, "no". ;)
76.
by md_ 3y ago
Amusing. I, too, wrote my own ML-powered newsreader. (Not linking here because I don't want to de-anonymize my HN handle.) I guess this is a thing people do. ;)
77.
by md_ 3y ago
Passkeys (as password replacements) generally require user verification (like a PIN or fingerprint) to unlock the secret. I don't think very many people know what "security keys" or "FIDO" are, to be honest, so you&
78.
by md_ 3y ago
This is an extraordinarily bad take. :) - The biggest threat almost all users face is in the form of remote attackers: password phishing, database leaks + password reuse, tricking users into installing malware, etc. Local attackers are so
79.
by md_ 3y ago
I grew up in agricultural rural America in the 80s and 90s. Tons of pickup trucks and, later, SUVs! Subjectively, though, I'd say: * The real farm families would have a pickup for pulling a horse trailer, moving mulch/manure/
80.
by md_ 3y ago
I must be getting different webpages when I click those links than you do. Do you think it’s the cia?
81.
by md_ 3y ago
> I agree that the issues you listed are indeed real and challenging, but I also think that the media's heavy reliance on the "intelligence community" making claims as evidence when there is none is a more pressing one. I
82.
by md_ 3y ago
I remember. I don't think it was because of some unpaid interns in the mail room or whatever. I think this kind of cloak-and-dagger conspiracy theorizing trivializes the real, challenging, human issues of editorial independence, the co
83.
by md_ 3y ago
I assume Musk will get distracted by something else--a bathroom-humor pun off the name of one of his companies, say, or realizing that 4/20 is just around the corner again, or the implacable, unconquerable, immutable existential sadnes
84.
by md_ 3y ago
Hah, funny. This is how policy changes happen in the Musk era: a juvenile troll, followed by an embarrassingly transparent attempt to cover it up. The joke is on all of us for giving this silly man our attention. I regret that I even know a
85.
by md_ 3y ago
I think it's funny that all the Americans/Russians commenting here seem to believe there are only two public broadcasters in the world: RT, and NPR. And that Twitter was merely being unfair by not labeling NPR as "state affil
86.
by md_ 3y ago
Many countries have publicly funded/editorially independent broadcasters. Of the below, none seem to be listed as "state affiliated": https://twitter.com/bbc https://twitter.com/deutschewelle
87.
by md_ 3y ago
https://www.theguardian.com/world/2000/apr/12/julianborger "Cable Network News (CNN) and National Public Radio, (NPR) denied that the "psy-ops" officers influenced news coverage and said t
88.
by md_ 4y ago
Most of them change their minds when they realize their customers--viewers, subscribers, advertisers--don't want to see all legal speech. It's amusing that Musk's incentive appears to be fear of what his viewers will see -
89.
by md_ 4y ago
We were young, and stupid. ;)
90.
by md_ 4y ago
Fifteen years ago at Microsoft, most people worked ~40 hour weeks, there were shuttles between Seattle and Redmond, a good gym on campus, and (mostly) private offices. You'd come into work at 9, wander off campus to a nearby restaurant

More ›