jefftk

31 ms·

571.
by jefftk 3y ago
> They purposefully did not precondition any of the batteries. Where are you seeing that? It looks like they use P3 for the testing, and while I'd like it if P3 was clearer about their testing approach it sounds to me like they do
572.
by jefftk 3y ago
But it doesn't interleave? You get a sequence of pairs.
573.
by jefftk 3y ago
The best explanation I've read is https://www.lesswrong.com/posts/KXHMCH7wCxrvKsJyn/openai-fac... and https://www.lesswrong.com/posts/KXHMCH7wCxrvKsJyn/openai-fac... along with Zvi&
574.
by jefftk 3y ago
ConsumerReports just covers the American market, so in addition to Renault (which doesn't sell in the US [1]) it excludes all cars that the listed makes sell outside the US. [1] https://www.hemmings.com/stories/ren
575.
by jefftk 3y ago
Neat that you'd found it! I think part of why people didn't care was that you didn't realize (or didn't post) that the random gibberish was verbatim training data?
576.
by jefftk 3y ago
> MIDI is absolutely horrible for ML. It lacks very necessary information such as articulation etc which are important to make sense of music. This depends enormously on the instrument. Consider someone playing a piece live on a keybo
577.
by jefftk 3y ago
What do you like better about the XML version?
578.
by jefftk 3y ago
Huh? What I'm proposing if you want to make it realistic (which may or may not be good UI) is: * Everything fades from when it was drawn * Everything is only drawn with the sweep * Simulate a continuous motion for the planes, and use t
579.
by jefftk 3y ago
> this DMCA process that expects the claimant to issue a good faith report, under penalty of perjury. This is a common misunderstanding, but if you read the text of the statute it's clear that the "penalty of perjury" o
580.
by jefftk 3y ago
On the other hand, if you want maximum realism, everything should fade continuously from when it is drawn. The underlying technology being imitated here is a slowly decaying phosphor.
581.
by jefftk 3y ago
Nice! Minor nits: * It takes me out of the immersion a bit when planes move after having been drawn. It would feel more realistic if the blips were "painted" by the sweep and then static until the next sweep. * To make it a bit
582.
by jefftk 3y ago
This is a screenshot of https://twitter.com/jakezward/status/1728032634037567509 hosted on Mastodon. We pulled off an SEO heist that stole 3.6M total traffic from a competitor. We got 489,509 traffic in October a
583.
by jefftk 3y ago
Note the zones, though: the AI was far smarter in the environment where it did the former.
584.
by jefftk 3y ago
Yes, if you only need consent for the cart then matching the notification to the action is nice.
585.
by jefftk 3y ago
Shopping carts require cookies if they go beyond what is "strictly necessary to provide a service explicitly requested by the user". This is quite a high bar, and most shopping cart implementations do not meet it. For example, if
586.
by jefftk 3y ago
No. Cookie banners cover any use of client size storage. That includes things like shopping carts and personalization (purely first party), analytics ("first party"), and ads (usually "third party"). Part of what's
587.
by jefftk 3y ago
> you had biplanes; then you had triplanes. You could have followed that farther, and added more wings - but it wouldn't have improved things. But people did try! https://en.wikipedia.org/wiki/Besson_H-5 htt
588.
by jefftk 3y ago
Except that it's a big open question how to control an AGI. So the choice may not be between which entity controls it, but whether it is created by an entity that is able to control it at all.
589.
by jefftk 3y ago
This is a lot like other kinds of security: when there's a hazard out in the wild you sometimes need to make people aware of all or part of the problem as part of fixing it. I would expect making it illegal for people to talk about the
590.
by jefftk 3y ago
I think the current level of caution in LLMs is pretty silly: while there are a few things I really don't want LLMs doing (telling people how to make pandemics is a big one) I don't think keeping people from learning how to hotwi
591.
by jefftk 3y ago
The laws here are in a pretty sad shape. For example, did you know that companies that synthesize DNA and RNA are not legally required to screen their orders for known hazards, and many don't? This is bad, but it hasn't been a p
592.
by jefftk 3y ago
I agree that this sort of control is a downside, but I don't see a better option? Biology is unfortunately attacker-dominant, and until we get our defenses to a far better place, giving out free amoral virologist advisors is not goin
593.
by jefftk 3y ago
What's the issue with including some amount of "model is aligned to the interests of humanity as whole"? If someone asks the model how to create a pandemic I think it would be pretty bad if it expertly walked them through the
594.
by jefftk 3y ago
I don't know, but she's also on the board of EVF UK [1], which is the largest effective altruism organization. [1] https://ev.org/effective-ventures-foundation-uk/
595.
by jefftk 3y ago
Or roughly every 136 years.
596.
by jefftk 3y ago
Spoiling the joke, but the Emergents are very clearly dystopian and NoMoreNicksLeft knows it ;) Definitely give the book a read! One of my favorites.
597.
by jefftk 3y ago
> If your goal is money and you are a manager then in any organization (not just Google) it's likely that income is going to be O(number of reports). Income linear in the number of reports? Increasing with reports, sure, but way
598.
by jefftk 3y ago
> There's also a HN selection bias: a story about FAANG is a lot more likely to be upvoted than a story about some small startup. And an HN selection bias in tone too: a story about how great it was to work at a FAANG is probably
599.
by jefftk 3y ago
Another way to reduce bandwidth costs!
600.
by jefftk 3y ago
Or people would set lots of videos to load in background tabs but only get around to watching some of them.

More ›