Visual Studio 2022 17.10 is available. This release features a new, single GitHub Copilot extension. To enable these AI experiences, activate your GitHub Copilot subscription today by signing in to GitHub and starting a free trial.
“AI” as currently hyped is giant billion dollar companies blatantly stealing content, disregarding licenses, deceiving about capabilities, and burning the planet in the process.
It is the largest theft of intellectual property in the history of humankind, and these companies are knowingly and willing ignoring the licenses, terms of service, and laws that us lowly individuals are beholden to.
I guess we wait this one out until the “AI” bubble bursts due to the incredible subsidization the entire industry is undergoing. It is not profitable. It is not sustainable.
It will not last—but the damage to our planet and fallout from the immense amount of wasted resources will.
I am not a fan of Copilot (and the like) on principle but I am a fan of the idea of a responsibly trained locally run “AI” assistant. These don’t make us dumber any more than calculators, computers, and GPS do. We will get worse at doing things we offload to the computers like we always have. That’ll make room for other sorts of areas we will get mentally sharp at. Our generation was scolded by elders for not learning all the ways to do arithmetic in our head. It’s the same sort of thing. I know I am not learning the roads where I moved well because I rely on GPS too much. When we choose to want to gain proficiency in an area we can fall back away from the technology. That doesn’t mean we are in any way dumber or less intelligent by using these tools, just differently so. #ai#llm#GithubCopilot Is an overreliance on Microsoft Copilot and ChatGPT negatively impacting our cognitive skills and making us dumber?
I've hooked up Copilot for #Xcode with #OpenRouter.ai’s server today (they provide an OpenAI compatible API server), and played around with phind/phind-codellama-34b & meta-llama/codellama-34b-instruct, and it’s been pretty nice.
Good models, both, and the explanations they provided for some of the Swift code I showed them were more concise than what #GithubCopilot delivered. YMMV.
OpenRouter is good, and using the various models costs only cents per day. Neat service.
#AI, specifically #GithubCopilot, is an incredible tool for upgrading code.
I fixed an architecture problem that resulted in over 700 warnings related to nullability in C#. I was able to clear all of those warnings properly with the help of AI in just a couple hours. Now my code base is far more robust and better at handling errors elegantly.
Shockingly, when you use a glorified pattern-matching algorithm to deal with a pattern-matching problem, you get good results. Who knew?
I've been playing around with it for a few days now, and so far in my personal use with #PHP / #Laravel it seems to produce better completion that GitHub's Copilot.
Plus ça va, plus je me dis que je ne signalerai plus de bugs ni ne proposerai de correction pour ces derniers sur Github. Que leur IA s'entraîne avec en pensant que c'est du bon code. Et tant pis pour l'open source, s'il reste sur cette plateforme. #ai#github#githubcopilot
@arselectronica Just like #GithubCopilot doesn't take away key architecturar planning and design decisions and for the most part just automates away the boring part of turning ideas into code.
And that's good and not a violation of copyright because only human beings have that and machines can't!
> Just as GitHub was founded on Git, today we are re-founded on Copilot. 🤖 See how GitHub’s AI-powered platform vision evolved into a new reality for the world’s developers, and find out everything we announced at this year’s #GitHubUniverse. https://nitter.net/github/status/1722309261680607674?s=20
For years I've been saying that using Microsoft GitHub to host #FLOSS code is dangerous and would backfire. :blobcatcoffee:
@PiTau@rysiek Sadly I've yet to see a better option that does scale and isn't maintained by Stallmanists...
I'd consider migrating away from #GitHub if it provides actual benefits and not just being a pain in the ass to do and basically shift said repos to be slaves.
#GitHubCopilot fun ...
I prompted left (pic) in #perl and it produced the (reasonably correct) code on the right. Interestingly enough, it cut the range of #fastq scores to 33+41 for unclear reasons (I corrected the 41 to 93), and I got myself a #fastq simulator
Every five to ten years, I make the effort to switch my #LaTeX editor to a more modern one. This process has now iterated several times back from when I was a graduate student in the mid-1990s using vi from a UNIX shell to write in plain TeX. On the suggestion of a reader here, I installed #VSCode + #TeXLive + #LaTeXworkshop + #GithubCopilot as an upgrade from my current setup of #TeXnicCenter + #MiKTeX
which I had been using for almost a decade, and am recording my first impressions here (which will most likely be quite naive for existing VSCode users).
The installation had no problems (other than the four hours needed to download the TeX live packages on a slow internet connection). I began experimenting with various features. So far I have mostly played with the user-defined code snippets feature, which can allow me for instance to create an entire corollary environment by typing in a trigger word (I chose "cor") and pressing tab (see enclosed screenshots). Strangely enough I had a version of this functionality 20 years ago during a brief period when I experimented with using Microsoft Word as a LaTeX editor purely for the ability to use Visual Basic macros (though I abandoned this shortly after due to the lack of other LaTeX-friendly features). I could certainly see myself using this feature frequently as a time-saver.
So far the AI-powered Copilot suggestions have been mainly useful for filling out the snippet functionality: after giving a few examples of the snippets I wanted, it was able to suggest more that I could accept, again with the single click of the tab key.
(Incidentally, the screenshots are displaying a paper which I will be putting on the arXiv shortly. Stay tuned...)
One very interesting observation from coding with #AI/#ML:
I've found that I've started subconsciously trying to race #GithubCopilot to type out lines of code before it can guess them.
I used to stop occasionally to sit and ponder what I wanted to write, but now if I do that, the machine tells me what I wanted and it's usually right, so that motivates me to stop pondering and do what I know needs to get done.
#GithubCopilot suggested "I genuinely don't have the spoons to figure it out" in a comment I'm writing explaining why a specific workaround is needed.
The full comment now reads: I have no idea why the unknown-casting here is necessary. TypeScript throws a hissy fit if it isn't there, and I genuinely don't have the slightest clue why.
Kritischen Bug in #Questlog eingebaut, den ich hätte finden können, wenn ich automatisierte Tests hätte.
Bugfix: 2 Minuten
Tests für alle wichtigen Basics (im Backend) in Questlog schreiben: ~40 Minuten
Hätte ich viel früher machen sollen. Aber manchmal braucht man einfach nur die richtige Motivation. 😅
Aber es ist echt krass, wie viel redundante Tipparbeit #GitHubCopilot einem beim Tests schreiben abnehmen kann. Und wie viele sinnvolle Tests vorgeschlagen wurden, die ich einfach vergessen hätte.
Man kann von ML-Assistants (ich weigere mich, das AI zu nennen) halten was man möchte. Aber gerade für nervige Aufgaben, die simpel und repetitiv sind, kann das enorm Zeit sparen.