cs702

36 ms·

541.
by cs702 2y ago
Too many Waymo robotaxis arrive at the overnight parking lot at around the same time, after their shift, creating congestion in the parking lot. It takes those robotaxis around an hour to self-park, during which they're moving forward,
542.
by cs702 2y ago
In other words, not cheap :-P
543.
by cs702 2y ago
Just when one would think AI hype is fizzling out. What does "training all 35K staff in generative AI" even mean? I have no clue, but I'm pretty sure it won't be cheap.
544.
by cs702 2y ago
> The bug in the legal system is that by filing this motion/claim, Disney can... For large corporations and rich individuals that are in the wrong, this isn't a bug of the US legal system. It's a feature . Dragging the
545.
by cs702 2y ago
Thank you. Yes, I'm not at all surprised to hear that. The biggest challenge I see for you guys is that your best customer prospects, i.e., those organizations which need this most urgently and are willing to pay the most for it are th
546.
by cs702 2y ago
Congratulations on launching! Trellis looks amazing ... but only if it works well enough, i.e., if the rate of edge cases that trip up the service consistently remains close to 0%. Every organization in the world needs and wants this, like
547.
by cs702 2y ago
Interesting. The authors claim this outperforms Ring Attention for distributed computation of self-attention over multiple GPUs. Distributing computation is necessary whenever context is too long for self-attention's computation to fit
548.
How Close Are the Planet's Climate Tipping Points?(nytimes.com)
10 points by cs702 2y ago|1 comments
549.
by cs702 2y ago
*"trust but verify"
550.
by cs702 2y ago
The key take-away, for me, is that I should "keep my guard up" on any video call about money or other important matters , even if other participants on the call are colleagues, friends, or relatives. There are no guarantees of au
551.
by cs702 2y ago
Well, I understand how it works, and I still find it freaking amazing. The quality is... impressive. On the flip side, the ability to deep-fake a face in real time on a video call is now accessible to pretty much every script kiddie out the
552.
by cs702 2y ago
At first, I couldn't believe anyone had actually asked about sorting inputs and outputs independently of each other, such that each sample input is no longer paired with its corresponding output. WTF? Then I read the second paragraph,
553.
by cs702 2y ago
Unless I'm missing something big, this looks like a significant deal for independent developers of self-driving AI software: GPUDrive enables them to run driving simulations with hundreds of AI agents on consumer-grade GPUs at 1M FPS,
554.
by cs702 2y ago
> the example you gave Ah, got it. I was just trying to keep my comment short!
555.
by cs702 2y ago
Thank you. By the way, I agree with you that "a good future for all of humanity" would be a fantastic goal :-) The multi-trillion dollar question is: How do you specify that goal as an objective function?
556.
by cs702 2y ago
You misunderstand. The phrase "the multi-trillion dollar question" has nothing to do with acquiring trillions of dollars. The phrase is idiomatic , indicating a crucial question, like "the $64,000 question," but imply
557.
by cs702 2y ago
Thanks. I view your comment as orthogonal to mine, because I didn't say anything about how easy or hard it would be for human beings to specify the problems that must be solved. Some problems may be easy to specify, others may be hard.
558.
by cs702 2y ago
Indeed. The reward function we're using in RLHF today induces AI models to behave in ways that superficially seem better to human beings on average, but what we actually want is to induce them to solve cognitive tasks, with human pri
559.
by cs702 2y ago
Programming has a clear reward function when the problem being solving is well-specified, e.g., "we need a program that grabs data from these three endpoints, combines their data in this manner, and returns it in this JSON format.&quot
560.
by cs702 2y ago
These three observations from the OP are consistent with my own: * Many people who spend well below their means experienced financial precariousness earlier in life. * They may no longer be poor, but the stress of deprivation is never far f
561.
by cs702 2y ago
Actually, you get one free year of credit monitoring. After the first year, you'll be asked to pay for monitoring.
562.
by cs702 2y ago
> The article also says that USDoD tried to sell the data for only $3.5 million, so they value it at roughly $830/person. When I divide 3,500,000 USD by 2,900,000,000 people, I get $0.0012/person. How do you get $830/perso
563.
by cs702 2y ago
Well, that's embarrassing. I watched the lecture a long while ago, and somehow my puny little organic brain misremembered the lecturer. Thank you for the correction. And apologies to Daskalakis!
564.
by cs702 2y ago
Cool and interesting. Thank you for sharing on HN. As Nash proved, under very general conditions (e.g., payoffs are finite), in every game there's always at least one equilibrium, i.e., at least one fixed point. Alas, as Papadimitriou
565.
by cs702 2y ago
> The word you’re looking for is speculators. Buffett is an investor. Ponzi participants are speculators. Yes. I refer to "investors" because that's what most speculators call themselves.
566.
by cs702 2y ago
As always and as ever, the blame for overpriced assets goes to... investors themselves. Who else? Investors of all stripes, driven by greed and FOMO, are the ones who have been bidding prices up, either by not cashing out or by buying in at
567.
by cs702 2y ago
> ideological archeology A short, memorable, oxymoronic, and yet accurate description of these efforts. I like the term so much that I'm going to start using "ideological [scientific field]" to refer to similar pseudo-
568.
by cs702 2y ago
Cool and interesting. The authors study how an artificial replication mechanism evolves to cope with emerging faster-replicating parasites in a spatial setting. Extinction occurs in about half the cases due to quick emergence of the fast-re
569.
by cs702 2y ago
I expect kids will figure how to inject prompts into Taco Bell's order-taking LLM in no time.
570.
by cs702 2y ago
This is not about Yale the university but about Yale the manufacturer of door locks. Homeowners using the company's first-generation "smart" locks have been given two choices: * Option 1: Don't pay up to upgrade your loc

More ›