oivey

30 ms·

241.
by oivey 3y ago
I agree. I think there will be lots of hard problems where an LLM is unlikely to be successful for quite a while. For high surety software like encryption or numerical software I think this is especially the case. Sure, you can have an LLM
242.
by oivey 3y ago
No, it doesn’t. A rusty car still drives.
243.
by oivey 3y ago
I’m not really sure what your comment has to do with carrying capacity. The damage humanity has done to the environment has more to do with a minority of the population using an insane number of resources.
244.
by oivey 3y ago
Other species don’t build technology. Technology changes the carrying capacity. Humans also have environmental impact that varies far more than animals. A bird in Africa and a bird in America probably have similar environmental impact, but
245.
by oivey 3y ago
I think in some sense requiring winter tires is optimal, but I think the heavily car centric transportation introduces other issues. Say you require winter tires, someone loses their job, and now they can’t afford to get the tires or the ma
246.
by oivey 3y ago
What are you even talking about? The person you replied to didn’t say Americans. Were you trying to use Bezos and Musk as representative of the top 1% of the worldwide population? That’s even worse.
247.
by oivey 3y ago
He’s not taking about the richest 1% of Americans. Even if he was your examples are disingenuous. Musk and Bezos aren’t representative of even the richest 1% of Americans.
248.
by oivey 3y ago
Did you mean non-GC languages? GC languages by default have at least memory safety because the GC frees for you when an object is detected to no longer be accessible.
249.
by oivey 3y ago
What’s your plan for that? I was going to say maybe put every single American in prison, but even that probably wouldn’t work.
250.
by oivey 3y ago
I don’t think the choice of using floating point numbers is particularly inefficient. If frameworks were fixed point by default it would be tricky to get the dynamic ranges right all the way through a network. The math in the training assum
251.
by oivey 3y ago
You could never extort MGM with gift cards.
252.
by oivey 3y ago
Yes, and at current rates it would take many, many generations for extinction to occur. Do the math with whatever birth rate you want to use. It’s not like birth rate is likely to stay constant over such a long period, either.
253.
by oivey 3y ago
No, even with lower birth rates there are going to continue to be many, many humans on earth.
254.
by oivey 3y ago
The Mojo one is already doing some pretty horrific low-level stuff with fairly manual SIMD. That’s why it was faster than Julia in the initial example, and the edge is lost when a couple of posters did similar things for Julia.
255.
by oivey 3y ago
Yeah, and then I can buy some insurance / hedges against the spot prices of electricity, day trade my air conditioner, and complete the financialization of my life.
256.
by oivey 3y ago
For most of these type hints or static typing do not at all help. I haven’t went through all of them, but in the first quarter or so almost all of them were due to leaky compiler weirdness in CPython.
257.
by oivey 3y ago
Maybe they used that experience decide that it would be cheaper to just build a huge rocket? The ISS also had the benefit of the shuttle, while US flights to LEO were in turmoil when the SLS was funded. It also wasn’t clear in hindsight how
258.
by oivey 3y ago
You’d probably have to figure out on orbit manufacturing, then, so I don’t think that is at all clear.
259.
by oivey 3y ago
You’re right, I misread static builds as typing. Python’s dependency management system sure does suck, although I think the problem could be solved without static builds. Overall they are a nice feature, though. I think the memory managemen
260.
by oivey 3y ago
I wish my bugs were mostly preventable via static typing and memory safety checks. For most ML and data work algorithm design and logic errors are the primary source of bugs. Type errors are usually the trivial things you fix in minutes.
261.
by oivey 3y ago
This Nature article claimed that two of the Chinese labs attempting replication provided XRD measurements that didn’t at all match what the Korean lab had. I’m not a pro in this field, but XRD gives you insight into the crystalline structur
262.
by oivey 3y ago
What is not true? I don’t see what this is supposed to address in my post. I cited Jython and IronPython because that’s what the person I was responding to mentioned. GraalPy looks neat, but is experimental/young still. Notably, it has
263.
by oivey 3y ago
Static type declarations don't make Java fast. The compiler does. Dynamically typed languages with no type declarations can be very fast if the compiler can infer the types. That's not to say that Python will ever get there. My un
264.
by oivey 3y ago
If subprocesses die (segfault maybe) it isn't uncommon for them to not be cleaned up and/or cause the parent process to hang while it waits for the zombie to respond. That's one I experienced last week on Python 3.9. A thread
265.
by oivey 3y ago
You have to manually set up shared memory with its own API that has its own limitations, right? I thought some seamless integration was a new feature, but AFAICT, transfers between multiprocesses still leads to things being pickled and copi
266.
by oivey 3y ago
There seems to be some confusion here. The GIL is not an invariant of Python that makes your code thread safe. Python is not altering the deal. You can still use threads to write concurrent code in Python today, and you'll still run in
267.
by oivey 3y ago
People stay on CPython due to the performance of C extensions and the vast ecosystem based on them. The fact that people have stuck with CPython isn't at all evidence that they like the GIL or that it doesn't lead to significant t
268.
by oivey 3y ago
That was the opinion of a handful of vocal posters. The overhead of using multiprocessing and/or some network service is extremely high for a lot of applications.
269.
by oivey 3y ago
The near miss behavior is very much like overfitting. Mario is simple and deterministic enough that it doesn’t matter, but think about a scenario like a self driving car. A calculated near miss turns into a crash if the other car’s driver i
270.
by oivey 3y ago
I agree with you that this perhaps the most useful engineering skill. I think schools try to teach this fairly young with supplemental book reading on top of lectures. That’s the teacher telling you where to find the information, but forcin

More ›