qayxc

66 ms·

751.
by qayxc 5y ago
> Just like how people are allowed to read websites, but scraping is often disallowed. Hosting code on Github explicitly allows this type of usage (scraping) according to their TOS so I have to ask again - why the sudden complains? Are w
752.
by qayxc 5y ago
Well, GTP-3 isn't open source, yet the Twitter-sphere was ecstatic about its code generation features. It isn't even new at all [0], [1], [2] - it's just getting exposure because people start actually using it. GPT-3 still fo
753.
by qayxc 5y ago
Copyright is different from licensing and is far more complicated that most people think. > Why take a risk for some glorified text snippets? Name an example where that was actually an issue. And I'm talking explicitly about single
754.
by qayxc 5y ago
And who checks that the code snippets didn't come or were derived from code licensed under a different license?
755.
by qayxc 5y ago
The exaggerations won't stop now, will they? First of all, it's not as if CoPilot spits out verbatim replica of training data on every other prompt. Secondly, the consequences of accidentally copying code by means of using this to
756.
by qayxc 5y ago
On top of that, I get the feeling that way too many programmers have a poor understanding of the actual value of their work. A single function, even if particularly clever, elegant, or performant, doesn't have much value on its own. Su
757.
by qayxc 5y ago
> ... will you defend it just because I claim it is a tech preview? That's a straw man argument and you know it. Code snippets are in no way shape or form comparable to entire software products and CoPilot neither installs anything
758.
by qayxc 5y ago
> As human beings we have limitations. That's a fair point. ML models don't seem memorise all the code they've seen either, it seems. Plus while the argument of human limitations applies to the vast majority of people, wha
759.
by qayxc 5y ago
> Today copilot does what it does. It's also not a commercially released product yet, but a technical preview, so uncovering and addressing issues like that is exactly what pre-release versions are for. I'd say it succeeded gre
760.
by qayxc 5y ago
> But in this case it is obvious that the AI isn't writing the code - at least not all the time, it is instead choosing what to copy - verbatim. I still don't see any problem with that. If it's larger sections (e.g. entire
761.
by qayxc 5y ago
Why and how? I'm honestly interested in an answer here. What exactly is the difference between a machine learning patterns and techniques from looking at code and people doing it? Is every programer who ever gazed at GPL'ed code g
762.
by qayxc 5y ago
If it's large sections, that can be fixed by either licence attribution or result filtering. That's at best a technical issue. What way too many people claim, however, is that the machine isn't even allowed to look at GPL&#
763.
by qayxc 5y ago
I don't understand the argument here. Yes, sometimes code is returned that is a verbatim reproduction of the training data. This can be prevented if need be. What I really don't understand is how some people are complaining about
764.
by qayxc 5y ago
The primary reason is that every organisation focuses on their own personal gains (mostly: power and influence). Carbon taxes are stupid because it's just another metric that can and will simply be gamed by the industry without any net
765.
by qayxc 5y ago
It even goes both ways: most species change their environment to better accommodate their needs (beavers dam rivers, termites build mounds, forests modify the soil and create their own microclimate, etc.). So it really is a complex system o
766.
by qayxc 5y ago
> What do you mean, Nature isn't kind to humans? Because we aren't immortal? Something like 90% of Earth's surface is outright deadly to humans: about 71% is covered in water, in which humans cannot survive; then there&#x2
767.
by qayxc 5y ago
Pain is a vitally important function of the central nervous system. Children born without the ability to feel pain (congenital insensitivity to pain and anhydrosis/HSAN IV) rarely live to puberty. There's very little to be gained
768.
by qayxc 5y ago
Even worse: if simulations are used, you now have two problems - formulating correct incentives and protecting against abusing flaws in the simulation.
769.
by qayxc 5y ago
> Are we approaching (or have we already passed?) a singularity in the human race based on our ability to product fault tolerant ball race bearings and machine screws that never snap? I guess not [0]. I don't know of any artificial
770.
by qayxc 5y ago
A complete specification also must include assumptions about external systems and limits. Here the assumptions are infinite stack (or item count < stack size limit) and a malloc that never fails. Incomplete specs are just as bad as buggy
771.
by qayxc 5y ago
> it's definitely not impossible a priori and especially not because of Gödel. You forgot something there: you know the underlying system and about its existence. In the absence of this knowledge it's indeed impossible to kno
772.
by qayxc 5y ago
> I'm going to assume it will also regurgitate malicious code. Well if it wouldn't replicate or even randomly generate malicious code, it would imply that CoPilot would somehow be able to solve Halting Problem or - at the very
773.
by qayxc 5y ago
The danger is that CoPilot becomes "AutoPilot", because as the author points out, you get used to it way too quickly. Searching the web still implies extra effort on your side and (hopefully) some scepticism towards the code you f
774.
by qayxc 5y ago
There are two types of products: the ones that people complain about and the ones that nobody uses or cares about; based on a famous Bjarne Stroustrup quote. The fact that people post about it, point out flaws, leave GH out of spite, etc. o
775.
by qayxc 5y ago
You can just disable JavaScript and the site works fine.
776.
by qayxc 5y ago
NVIDIA used CUDA to establish industry-wide vendor lock-in on GPU compute. Google uses TPUs to try and establish an industry-wide vendor lock-in on deep learning. Same old, same old.
777.
by qayxc 5y ago
> I continue to be surprised GitHub shows examples that wouldn't compile/run correctly. Why is that surprising to you? CoPolit doesn't actually know how to code, it just generates symbols that match learned patterns. Somet
778.
by qayxc 5y ago
The RTX 3090 is a beast compared to what researchers had available to them just a few years ago. Don't try to chase SOTA - that's a fruitless endeavour. 24GB of VRAM is plenty for CV and you can train some excellent models with it
779.
by qayxc 5y ago
The irony of it all is that their code will find its way into the next Common Crawl release anyway and that's used to train GPT-3, which in turn forms the basis of OpenAI Codex, which is the product that CoPilot builds on... So hosting
780.
by qayxc 5y ago
It's a reflex with some people and that's OK. Not everyone has the patience and ability to discuss their objections in a public forum while their rights are being violated (in their view). Some people have a passion and a very str

More ›