Hacker News Reader

Comment by alharith | Hacker News Reader

Microsoft didn't buy Github out of the goodness of their heart. They also didn't buy it because of it's significant revenue, a couple hundred million is probably a marketing budget to Microsoft, not something that drives the bottom line. Microsoft also didn't buy it to score appearance points with the socially conscious dev community that populated most of Github, although it was a nice PR story for a bit. You could say the ICE thing has driven all that goodwill away.

Microsoft bought Github for the data set. To mine code to train models to write code to put developers out of business. Microsoft is coming for all of your jobs. They aren't investing billions in AI research and Open AI for philanthropy. The sooner people realize this the better.

Sure it's tin-foil, but it's a hill I am willing to die on.

View on HN

> train models to write code

I've seen "deep dream". You'll get models that produce stuff that locally looks like code but doesn't actually make sense. Code-like extruded product. People will waste careers on the "steering" and "minor fixups" necessary to make it actually work which will be orders of magnitude larger than the work needed to just write the software in the first place.

That's the cost of giving up symbolic AI in favour of pattern recognition.

I'm not scared of a "model" "trained" on the long tail of Github code any more than I'm scared of someone who can search stackoverflow faster than I can.

Anything sufficiently advanced to produce working code from vague design documents would be functionally indistinguishable from true AI. Barring some kind of major breakthrough in the field of AI (not machine learning, actual AI), this is basically a pipe dream. At best they'll produce something that given a very detailed and rigid set of instructions can produce a working program... which is exactly what a compiler does.

I shudder to think of the kind of code that a machine-learning product trained on the full set of public GitHub repositories would produce. It might be better than what Frontpage used to spit out, or Word, when you used the save as HTML option, but I'm not sanguine about that sort of thing ever putting anyone out of work.

Are they working on something like that? I'm sure that there are people doing so. Will it eventually trickle in some limited way as a productivity enhancement into Visual Studio or VS Code? Probably. Is it going to be some kind of magical no-code silver bullet that obviates the need for programmers? Not a snowball's chance in hell.

Nah. Microsoft bought Github among doing other things (WSL, becoming top member of the Linux Foundation etc.) because they want to replace GNU/Linux with their MS/Linux which could be mostly Open Source and mostly free, but I'm sure will be engineered in a way that gives Microsoft full control of it. There is no need to put Linux developers out of business when Linux developers themselves already use a MS dev system (VSCode) on a MS platform (WSL). Just wait for the day applications will start to nag users with "For best user experience, this Linux application should be run under WSL".

AI is a good thing if it takes all jobs.

The bad thing is the non-stop messing with standards and EEE. Because it makes existing jobs unbearable.

Just wanted to mention that you can already mine the data that would be useful for something like you mention, like so: https://console.butt.google.com/marketplace/details/github/g...

I think the OP thing is bad.

What you're suggesting (if they could pull it off) I would welcome. Either it turns into (an even worse than now) corporatist dystopia and there's turmoil before a reset, or it opens up for "everyone" and we're closer to post-scarcity.

But most likely they'll automate some stuff, because this isn't real AI, and I can adapt faster than they can and do stuff their models can't.

Reply on news.ycombinator.com