causal

32 ms·

781.
by causal 2y ago
Considering all the high level departures, this makes the most sense to me. Their valuation largely rests on this mystique they've built that says they alone are destined to unlock AGI. But there's just no reason to believe they h
782.
by causal 2y ago
Saw someone on HN call it NopenAI
783.
by causal 2y ago
Ehh, there's a lot of space between "desperately in need" and "wanting to seize billions in equity".
784.
by causal 2y ago
Yeahhh they might have lost some awareness of how utterly dystopian their own tech can be
785.
by causal 2y ago
That graphic comparing benchmark averages is really nice, wish things were presented so clearly more often. That being said, I think this definitely tilts things in Molmo's favor by including so many benchmarks that seem to favor Molmo
786.
by causal 2y ago
The type of test definitely matters but 15 hours ain't too shabby either
787.
by causal 2y ago
I think a better piece of advice would be: Ship your projects, don't worry about finishing them. I'm more likely to be bitten by the need to perfect than the lack of effort to finish. And I still often wait too long to ship and ge
788.
by causal 2y ago
Huh, that doesn't really make a lot of sense to me. Why turn them off after being sighted? Because AIS tags aren't matching reality perhaps?
789.
by causal 2y ago
Nice, thanks for sharing such a complete example. So I take it there isn't really any kind of verification of AIS message validity. I suppose that would be difficult to do, so it makes sense that AIS would be easy to spam.
790.
by causal 2y ago
Right, this article is lumping together a bunch of different things that are maybe loosely related because of anime as a style. The unstated thesis might just be surprise that this once-nerdy stuff has become mainstream at all and therefore
791.
by causal 2y ago
Making the remote editing extension closed is particularly frustrating, as you have little visibility into what it's doing and it is impossible to debug obscure errors
792.
by causal 2y ago
Thanks for sharing! I'm trying to understand what bogus AIS tracks means - are you saying these are AIS messages injected into the network somehow, or actual AIS radio signals that somehow ended up transmitted from the wrong place? I&#
793.
by causal 2y ago
Config generally feels more complex though, since there isn't a central parameter server anymore. The colcon build system also just feels more complex now, which I thought was already impressively complex with catkin.
794.
by causal 2y ago
Slow-moving successor, which the community isn't exactly going wild over. It offers modest improvements in exchange for a painful upgrade process, and many of the original issues with ROS1 remaining unsolved. The other half of the joke
795.
by causal 2y ago
Upvote them, definitely something that makes HN special.
796.
by causal 2y ago
I trashed Arc immediately after install when I found out having an account was mandatory. That seemed so silly, like toothbrushes-requiring-wifi absurd. How much moreso now.
797.
by causal 2y ago
Claude definitely seems more "emotionally intelligent", but even just for 1-shot coding tasks I've been pretty bummed with 1o... like it will provide lots of output explaining its reasoning, and it all seems very sound, but t
798.
by causal 2y ago
I don't know what I'm doing wrong but I've been pretty underwhelmed by o1 so far. I find its instruction following to be pretty good, but so far Claude is still much better at taking coding tasks and just getting it right on
799.
by causal 2y ago
Love the irony of someone mocking science fiction being an accidental prophet
800.
by causal 2y ago
It's even more subtle than that. Things just start decaying when talent leaves- you often can't trace it to a single engineer or piece of knowledge that's lost. You'll find reasons why things broke- but you won't se
801.
by causal 2y ago
My first few attempts at getting it to work with an existing codebase have not been impressive. Perhaps 1o is best suited to difficult problems that can be stated in only a few sentences.
802.
by causal 2y ago
It's trivial to come up with prompts that 4o fails. If it's hard to come up with prompts that 1o succeeds on but 4o fails, that implies the delta is not that great.
803.
by causal 2y ago
If anything this highlights how important Vienna was pre-WW1 and how much that status has changed. I suspect history articles of the future could likewise point to Manhattan as having hosted many famous people simultaneously, and anyone liv
804.
by causal 2y ago
"Break the first law of thermodynamics with this one simple trick!"
805.
by causal 2y ago
Good problem to take on, you're absolutely right that there is a lot of demand for reliability. Curious how effective the learning + tuning really is. I presume that's the reason for the limited selection of models - i.e. only som
806.
by causal 2y ago
$0.05/request is a staggering amount. Their demo to order a burger is funny considering that a long chat session could easily end up costing more than a burger.
807.
by causal 2y ago
I had a friend who developed the same schizophrenic symptoms described in this article, and responding with any kind of acceptance was difficult when he was starting to threaten and act against people who he imagined were after him. Last I
808.
by causal 2y ago
Thanks for sharing your experience. I'd like to point out that there are other victims too- the wife suddenly being accused of murderous intent is very much a victim too, and in some cases the person hearing the voices acts first. I ha
809.
by causal 2y ago
Taking an open source project and making it closed source no less. Poor taste IMO
810.
by causal 2y ago
I'm also thinking long-context LLMs are going to make this advice seem pretty archaic in a few years. They're so good at reading code and extremely useful for asking questions of a code base. That said, I completely agree with the

More ›