Lichtso
Joined 10 years ago
Karma
760
Submissions
282
Member since
Sep 21
Recent submissions
Comment · 1 day ago
Comment · 1 day ago
The reversibility of transactions is often brought up but I don't think that is the main differentiator, or any at all really.
First, both cryptocurrencies and traditional finance systems have chosen to and have chosen not to reverse or freeze criminal transactions in the past. Decentralized blockchains have chosen to do so at least once [0]. The traditional finance system has chosen not to many times and even large sums [1] [2]. It really comes down to how much is stolen and how much influence the victim(s) have.
No, instead I think the main difference is the speed of transactions. Some blockchains operate within milliseconds while bank transactions take days. Even instant transfers are actually settled on a way longer cadence in reality.
That means attackers can do a fast sweep on blockchains, which is not possible with bank transactions because everything happens in slow motion and humans have way more time to react and step in.
[0]: https://en.wikipedia.org/wiki/The_DAO [1]: https://www.turkiyetoday.com/turkiye/german-bank-e100m-disap... [2]: https://en.wikipedia.org/wiki/Bangladesh_Bank_robbery
Comment · 1 day ago
The Youtuber "Emergent Garden" has run similar experiments in Minecraft, though with smaller swarms (if you can call it that at all).
Comment · 2 days ago
Comment · 3 days ago
You would be surprised how hard it is to visualize arbitrary graphs. Basically, we can only render small highly regular ones in a neat way. Everything beyond that turns into a ball of wool.
Graphs in general are high dimensional structures and don't lend themselves to 2D projections. That is not a problem with coding the layouting algorithm. There simply is no good geometric arrangement for most graphs / networks because they are not planar [0], or even close to it (with few intersections).
Comment · 12 days ago
It is.
What I think the parent post is missing is that breakthroughs in science don't immediately translate into breakthroughs in engineering and even less so into economically viable products.
Comment · 15 days ago
If I am not mistaken a weighed fold / reduction [0] over a sliding window (kernel) is a convolution [1], right?
While the algorithm presented here is sequential, this smells like it could be done in parallel.
In fact the "cum_aggs" stack is a prefix scan [2].
For the sum of values (and by extension also the average), if you have the input prefix scanned (which can be done in parallel) then you can answer any (yes any, dynamic size, not fixed size window) subslice reduction in constant time.
[0]: https://en.wikipedia.org/wiki/Fold_(higher-order_function) [1]: https://en.wikipedia.org/wiki/Convolution [2]: https://en.wikipedia.org/wiki/Prefix_sum
Comment · 27 days ago
Minus the NPU target, if you want Mojo in Rust there is https://github.com/tracel-ai/cubecl
Comment · 29 days ago
There always has been a lot of stress, but there used to be relaxed moments as well. The issue is that we don't calm down anymore.
Relevant video by Kurzgesagt – In a Nutshell: https://www.youtube.com/watch?v=Mo1A45ShcMo
Comment · 1 month ago
LLMs are not like other tools. LLMs are running a model, and that model is just provenance / attribution washing on a unprecedented scale.
Comment · 1 month ago
On macOS you can delete them manually (though it requires booting into recovery mode and temporarily disabling System Integrity Protection).
Comment · 2 months ago
Technically soruce available now with the promise of accepting contributions (thus becoming fully open source) early next year. But since it is an Apache 2 license (like the rest of the LLVM project) you are already allowed to fork and contribute to your own fork right now, just no upstreaming.
For many the closed source nature of the compiler was a knock-out criterion. We will see if Mojo can gain traction now or if it has missed its window of opportunity.
Comment · 2 months ago
People often forget that architecture in (central and northern) Europe is build under the assumption that the climate has more cold extremes than hot extremes. Thus, buildings were constructed to both retain and collect as much heat as possible to save on heating in the winter. Now, if you just slap an AC onto that it will have to fight all of the passive heat collection. In other words, you would have to also close off all south facing and roof windows for it to work.
Furthermore, the way the last mile distribution of power works is often underground (with the cables directly in the sand, no hose or tube around it) which makes it expensive to upgrade to handle a much higher peak consumption. Only lucky thing is that you mostly need AC when the sun is shining, thus local production via photovoltaic cells is a good solution that sidesteps the power grid problem.
Comment · 2 months ago
Audio glitches like the one at 45:15 (https://youtu.be/4S8I22ybG2c?t=2715) make me think even the interview video might be generated.
Exactly, that is my point. It is less about reversibility and more so about speed. Being slow can be an advantage (for security).