marcinjachymiak
Joined 7 years ago
Karma
119
Submissions
35
Member since
Mar 7
Recent submissions
Comment · 3 months ago
It's not surprising since these are all reasoning models now. Token sampling/decoding isn't exposed because the reasoning models are trained around specific configurations. That's why it's been replaced with "reasoning effort" instead (low, medium, high "thinking" levels).
Comment · 9 months ago
The ISS is in the middle of rolling out upgrades to their panels so it’s not a great comparison.
https://en.wikipedia.org/wiki/Roll_Out_Solar_Array
> ROSA is 20 percent lighter (with a mass of 325 kg (717 lb))[3] and one-fourth the volume of rigid panel arrays with the same performance.
And that’s not the current cutting edge in solar panels either. A company can take more risks with technology choices and iterate faster (get current state-of-the-art solar to be usable in space).
The bet they’re making is on their own engineering progress, like they did with rockets, not on sticking together pieces used on the ISS today.
Comment · 4 years ago
The memory layout can be better this way depending on how the fields of a strict are laid out. It also gives the data better locality.
Let's say you're iterating over position coordinates of some set of objects in a game. With an array-of-structs you can lose ou on cache locality if unrelated fields are nearby. If you use a struct-of-arrays you get cache locality advantages anytime you do something like "iterate over this 1 field for each object"
The latter approach is called data oriented programming btw, and seems to have been around for a while. Andrew Kelley gave a good talk about it recently while using Zig: https://media.handmade-seattle.com/practical-data-oriented-d...
Comment · 4 years ago
There are also other ways to get these returns outside of DeFi, like by staking ETH, selling covered calls, or various other strategies on centralized services
Comment · 4 years ago
I don't know what happens behind the scenes at Celsius.
I do know there are many ways to get those high ROIs in DeFi.
With zero evidence suggesting they are lying about returns, the charitable guess I can make is that they are getting high DeFi returns and giving their customers slightly lower ones after taking a cut.
Comment · 4 years ago
I'm not a fan or user of Celsius, but this is an incredibly inflammatory title with very little evidence to back it up. Claims like that demand greater proof. This is an incredibly lazy article
The author pretty much failed to do any research on DeFi investments (point 3 in the OP). Compound and Aave are just 2 of many places investors place their assets, and are definitely near the lower end of APYs. Badger, which Celsius has already said they used, offers much higher returns. So do other projects like e.g. Convex Finance.
Where do these high APYs come from? Well a lot of it is coming from incentives of these protocols and speculation in those native assets. But the author doesn't even know that, so I won't bother steelmanning his argument
Comment · 5 years ago
A large proportion of readers on HN are also quite lucky in similar ways to OP and can make this work. This lifestyle also requires less luck now with remote work having taken off big time.
Comment · 5 years ago
chess.com has a really cool vision game for picking out squares and moves by name: https://www.chess.com/vision
It seems silly but makes reading chess books and analyzing games much easier. The first time I played with the vision tool for 20 minutes I felt way faster at it already.
Comment · 5 years ago
The trigger for surveillance is $600 which is practically every American. It is certainly about regular people.
Larger sums of many are already tracked, so wealthy people would have already been monitored.
Comment · 5 years ago
It's good to see that OOM errors can be handled in Rust now, at least in unstable.
Zig has a similar approach that is pretty cool. I don't know of any other language that let's you handle it like this https://ziglang.org/documentation/0.8.0/#Heap-Allocation-Fai...
Comment · 5 years ago
You could probably use a low Japanese table for this pretty well.
I think a combination of positions is probably best to get variety in muscles you're using. Reminds me of a theory I heard that hard-backed chairs are actually better because they make you fidget and readjust your body, versus using the soft chair as an external spine
Comment · 5 years ago
Yes, that is the main argument and I agree with it. But bundling it with EA is just inaccurate.
Longtermism seems to me like one of many deadends that utilitarianism takes you down if you embrace it and only it as your moral theory.
A little bit of utilitarian thinking can be fine when you have another theory to tell you the "why". That's why I don't have any problems with EA which is usually quite concrete
Comment · 5 years ago
> Hence, they point out that focusing on superintelligence gets you a way bigger bang for your buck than, say, preventing people who exist right now from contracting malaria by distributing mosquito nets.
This comment and others in the article seem to be making a (bad) case that Effective Altruists don't care about the Global South. I find this bizarre because all EA enthusiasts I know are very much into mosquito net altruism.
GiveWell, an organization very popular among EA folks, has a list of top charities and most of them target the Global South https://www.givewell.org/charities/top-charities.
Comment · 5 years ago
Personally, I am convinced that all of the harms of social media can be traced back to the Garden of Eden.
People just decide to be bad sometimes. The perfect feed algorithm doesn't exist.