advael

22 ms·

331.
by advael 2y ago
I mean, maybe, but this is definitely doing causal modeling backwards Yes, it's possible that there are strong genetic predictors of household income, as a lot of people seem to want there to be for some reason, but when predicting the
332.
by advael 2y ago
Yes please Vastly prefer commercial enterprises that are upfront about how they make money. It would be so sweet if internet businesses charged money for things instead of creating a massive surveillance machine that is also the most advanc
333.
by advael 2y ago
The word "criminal" has always been an overbroad term that conflates several different levels of harm without nuance, but at this point in history I don't even believe that the majority of people incarcerated are there from d
334.
by advael 2y ago
Of course prioritizing shipping more features is going to ship more features. The whole point of prioritizing engineering is to trade off some speed for other things, like your shit not breaking. The incentives in most companies produce cat
335.
by advael 2y ago
I agree with you entirely that intuition constitutes fairly weak evidence of something being true, and isn't the sort of thing we should view as a source of data for empirical methods, unless we're studying the intuition itself as
336.
by advael 2y ago
This post demonstrates a willful ignorance of the factors driving so-called "populist techno-pessimism" and I'm sure every time a member of the public is exposed to someone talking like this, their "techno-pessimism&quot
337.
by advael 2y ago
The model is "open-source" for the purpose of software engineering, and it's "closed data" for the purpose of AI research. These are separate issues and it's not necessary to conflate them under one term
338.
by advael 2y ago
From a legal perspective, yea. If we end up having any legal protection against training AI models, legal liability will be a huge mess for everyone involved. From an engineering perspective, if all you need is the pretrained weights, the
339.
by advael 2y ago
Generally speaking, no. An important part of a lot of benchmarks in ML research is generalization. What this means is that it's often a lot easier to get a machine learning model to memorize the test cases in a benchmark than it is to
340.
by advael 2y ago
This isn't a dilemma at all. If Facebook can't release data it trains on because it would compromise user privacy, it is already a significant privacy violation that should be a scandal, and if it would prompt some regulatory or l
341.
by advael 2y ago
Kinda. Since you can self-host the model on a linux machine, there's no meaningful way for them to prevent you from having the trained weights. You can use this to bootstrap other models, or retrain on your own datasets, or fine-tune f
342.
by advael 2y ago
As an ML researcher, I agree. Meta doesn't include adequate information to replicate the models, and from the perspective of fundamental research, the interest that big tech companies have taken in this field has been a significant imp
343.
by advael 2y ago
Look, absolutely zero people in the world should trust any tech company when they say they care about or will keep commitments to the open-source ecosystem in any capacity. Nevertheless, it is occasionally strategic for them to do so, and t
344.
by advael 2y ago
This reeks of propaganda - I don't accept the nonsensical premise that a terrorist group is somehow forcing a vastly superior military force to kill this indiscriminately, and I especially don't buy that civilians somehow invite t
345.
by advael 2y ago
Guess we just take legal matters on faith then? My lawyer's gonna be devastated
346.
by advael 2y ago
I think this is an important thing to point out. This is the first really compelling test of the legal theory that the overbroad content-ownership boilerplate licenses big user-generated-content hosts use could actually bind people to allow
347.
by advael 2y ago
This is a nice regularization method, and I'm curious how it compares to model distillation as a simplification method. Does it do better? Is it synergetic or redundant with a distillation approach? Does it simplify in a way that also
348.
by advael 2y ago
Explicit keyword seems reasonable too, I just like !! more than the verbosity of casting tbh. Totally aesthetic preference
349.
by advael 2y ago
I'm at least moderately sure it is
350.
by advael 2y ago
A pattern I've seen some people do is to just make an operator bool for the type that checks for a contextually-sensical invalid state that is also the default constructed state. This is absolutely a hack, and has lots of other implica
351.
by advael 2y ago
Yes, it's likely an unmitigable problem with trying to aggregate information in this way. That's why the value proposition of a search engine is that it presents information from primary sources, rather than trying to summarize it
352.
by advael 2y ago
This reading isn't supported by the text
353.
by advael 2y ago
Yea, and use of highly privileged software with the ability to update itself that the organization has no oversight of should be the most suspect. Software is used by nearly every organization for drastically different needs, and I think
354.
by advael 2y ago
At this point I just assume any "cybersecurity expert" that defends Microsoft's nonsense is a cop
355.
by advael 2y ago
The answer to both questions is robust organizational infrastructure. To be frank, I think a minimal linux system as a baseline OS serves most use cases better than a bare metal application, but many applications have self-contained update
356.
by advael 2y ago
Is your implication that only muslims would oppose the IDF's genocide? A lot of Israeli citizens are against it, too, and only a US veto prevented the UN from condemning it
357.
by advael 2y ago
I agree that all dependencies should be treated as attack surface. For that reason, systems for which dependencies can be more tightly controlled are inherently more secure than ones for which they can't. The monolithic and opaque natu
358.
by advael 2y ago
In the current economic environment, something doesn't have to be wise or even feasible to have an "industry"
359.
by advael 2y ago
You can run single-purpose software on bare metal, and many OS-agnostic toolkits for things like user interfaces exist
360.
by advael 2y ago
Most people don't know how to tell what's going wrong with a windows computer A windows computer that relies on cloud services, as an increasing and often nonsensical subset of the functionality on one does, can often only be fixe

More ›