mattalex

7 ms·

31.
by mattalex 3y ago
The problem was mostly that the only guy that was really backing the project (Christian Ude, SPD), was replaced with his successor (Dieter Reiter, SPD) who just didn't have the drive necessary to maintain the project. The entire design
32.
by mattalex 3y ago
The US could have required stock in Intel instead of taking nothing: Intel gets money, US gets influence over Intel. Just like it works everywhere else.
33.
by mattalex 3y ago
Not in the grand scheme of things: you can get fpga dev boards for $50 that are already useable for this type of thing (you can go even lower, but those aren't really useable for "CPU like" operation and are closer to "a
34.
by mattalex 3y ago
It's actually a lot worse than that: Just redesigning LLMs to have separate input channels for prompts and data doesn't solve the problem either, since this would be impossible to train. Effectively you would need to filter all in
35.
by mattalex 4y ago
probably. But you can make it sufficiently hard: If you force the bacteria to make sufficiently large jumps in mutation to escape whatever antimicrobial you design, odds are that you force it into something either incompatible with life or
36.
by mattalex 4y ago
Instruction tuning mostly relies on the quality of the data you put into the model. This makes it different from traditional language model training: essentially you take one of these existing hugely expensive models (there are lots of them
37.
by mattalex 4y ago
It's against openai ToS. Whether this holds up in practice is its own thing, but it's better to not give anyone a reason to shut the project down (even if only temporarily)
38.
by mattalex 4y ago
> If you want to "fund" public transport, you should do it through increased taxation on cars (especially luxury cars) and air travel (in particular private aviation), but ultimately public money doesn't necessarily need t
39.
by mattalex 4y ago
Just that in this case the players are so powerful they can use their wealth to define the rules themselves.
40.
by mattalex 4y ago
There are tons of teams working on historical document analysis, just look at e.g. ICDAR. Working with historical documents is just very hard do to poor data quality and availability.
41.
by mattalex 4y ago
> Sounds like they are taking steps to make sure whoever was chosen for layoffs were not chosen for racist reasons. that would have been illegal either way. Really there are two options here: Either they did something questionable and wa
42.
by mattalex 4y ago
> It's absolutely insane to me that the OS that actually allows users to install third party app stores and third party apps had an antitrust decision against it before Apple, who locks down there entire OS to users, did. Just that
43.
by mattalex 4y ago
He also has a hard drive based NAS, this is an additional local buffer for current work.
44.
by mattalex 4y ago
I think that's a little too jaded. Much more likely they do this now for brownie points and quietly kill the program as soon as nobody's looking. You also have to keep in mind that the cost of onboarding a new developer is orders
45.
by mattalex 4y ago
Unpaywall has very little to do with scihub: This isn't about giving access to works that aren't publicly available, but instead about giving links to existing free uploads of the material by the author (i.e. on arxiv or his/
46.
by mattalex 4y ago
> keep in mind that he is literally the only one who talks about the company that way from everyone whose ever worked there (many, many have left for various reasons since). That's not true. There's been lots of criticism comi
47.
by mattalex 4y ago
In those places the wages are also considerably lower. The 2/3 are not the people buying, it's the people inheriting existing property: First-Time buyers only make up 33% of home owners.
48.
by mattalex 5y ago
Does it matter if you have two factors in your vault, but secure the vault itself with two factors? I secure my vault with 2 factors, therefore any attacker would 1. have to access my vault in the first place 2. find the first factor 3. f
49.
by mattalex 5y ago
Use latexml this is what's running under the hood of ar5iv https://ar5iv.labs.arxiv.org/ and should be able to compile every Latex document to html
50.
by mattalex 5y ago
As long as your name isn't google, transfer by pigeon will always be faster (regardless of africa, silicon valley or downtown New York). The reason people use the internet for data transfer has little to do with transfer speed and ever
51.
by mattalex 5y ago
In standard mathmatics you cannot identify isomorphic objects without the use of an informal argument or abuse of notation: The issue is that the two equivalent objects aren't actually identical, they just behave the same in some way.
52.
by mattalex 5y ago
Barely: the majority of energy tabulated as "hydroelectric power" comes from pumped hydroelectric storage. However, computing this as hydroelectric power is very naive: Option a) the power to pump was generated using fossile fuels
53.
by mattalex 5y ago
Arguably this is already illegal under article 21 of the EU's charter of fundamental rights https://ec.europa.eu/info/aid-development-cooperation-fundam... Specifically the social origin, genetic features, and pro
54.
by mattalex 5y ago
I still think it's absolutely insane that for aviation we require everything to be formally verified, while cars don't have any such thing and can brick themselves via over-the-air updates at any time. And that ignores the even mo
55.
by mattalex 5y ago
> Did we finally learn how to extract more energy from a fusion rector than we supplied to operate it? No. The issue is that most people only report the gain over the plasma (i.e. how much energy was put into generating the plasma) rathe
56.
FreeCAD now has a legal non-profit association(twitter.com)
7 points by mattalex 5y ago|0 comments
57.
by mattalex 5y ago
> It makes no sense to me how bitcoin is still number one cryptocurrency despite it's garbage fundamentals. Failed at everything it was supposed to do. It doesn't matter that bitcoin failed in everything it set out to do (and m
58.
by mattalex 5y ago
The antminer S19j (the j is important) consumes about 3.2kW at 90TH/s
59.
by mattalex 5y ago
At what point does it just become easier to prove the semantics with e.g. VeriFast or frama-c?
60.
by mattalex 5y ago
Why would it be a problem if they were related to his day job? Keep in mind the libraries he was inspired by are open-source apache licensed code. Your argument only makes sense if you assume the code was related to proprietary tech and not

More ›