Hacker News new | past | comments | ask | show | jobs | submit | MonkeyClub's comments login

Well said.

Particularly, "Data excluded from training by default" is not available in the free and first paid tier.

Google was obviously irked that Microsoft got all this juicy training data since everyone is on their walled git garden, and tried to come up with a way to also dip some fingers into said data.

You should see the reviews in the JetBrains plugin page: https://plugins.jetbrains.com/plugin/24198-gemini-code-assis...

People are all so "shut up and take my money", but the bloody thing won't even sign them in.

But it's still in beta, right? Perhaps it'll start working in a couple more months.


The people they would manage are getting fired, so they're also becoming redundant by consequence.

Project managers typically manage projects, not people

Sure but projects don't complete themselves.

Not with that attitude they won’t.

I once worked in a company where one of my colleagues actively created phantom projects for his department, oversaw them, checked off metrics, went to meetings, and “deployed” solutions for capabilities that the company already had but had forgotten about. He once confided to me that he hadn’t actually made anything in over a decade. He still works there, 15 years later.


I salute your colleague's silent resistance to wage slavery.

The TFA's AI-generated image "Harnessing the most compute in the known universe" that captures the entire globe in a blue hue of AI compute is a bit disturbing, TBH.

I get the same vibe, especially after reading the update where the book author contacts him to clarify stuff.

The whole post reads like hasty clumsy grey marketing.


I wrote the blog post, and I did it on my own freewill, and am receiving no compensation. The main reason I wrote it was to help cement my own learnings from the book. I've heard that the best way to learn something is to teach it, so I wanted to see how much I could regurgitate on my own. Turns out, not a whole lot. It was hastily written, and more of a "brain dump" than anything else. I'm entering a new-to-me field, and wanted a place to document the things I'm learning. If anyone finds it interesting, great. If not, no big deal.

As for the specifics of the model I trained, I would be hard pressed to recall the specifics off the top of my head. I believe I trained a small model locally, but after completing that as a PoC, I downloaded the GPT-2 model weights, then trained / fine-tuned those locally. That is what the book directed. All the steps are in my github repo, which (unsurprisingly) like the author's repo. His repo actually has more explanation. Mine is more or less just code.


But it takes a master to not write the write piece.


> I don’t know that GitHub per se would be a requirement

Indeed.

Although providing a browsable source tree is convenient, we shouldn't default that on Microsoft's private platform (which, after all, monetizes the code stored there by using it for LLM training).

If a project is free software or open source, Codeberg.org is an excellent solution, while there exists a whole host of other web git hosts as well.

Let's take advantage of the field's diversity, lest it narrows down on us abruptly.


> makes you wonder about the future of humanity

Fun anecdote (because I can no longer remember the source):

There's this Roman author deriding the practice of filling one's villa with books for the sole purpose of showing them off at parties, and not reading them.

So the future of humanity, much like its past, will probably be like its present: pretentious faux culture for the many, laborious work for the few, and a big divide in between which educated populists manipulate.


From p. 6:

"But a headnote can introduce creativity by distilling, synthesizing, or explaining part of an opinion, and thus be copyrightable."

Does this set a precedent, whereby AI-generated summaries are copyrightable by the LLM owners?


> If I’m paying good money to host something locally

The thing is, however, that at 2k one is not paying good money, one is paying near the least amount possible. TFA specifically is about building a machine on a budget, and as such cuts corners to save costs, e.g. by buying older cards.

Just because 2k is not a negligible amount in itself, that doesn't also automatically make it adequate for the purpose. Look for example at the 15k, 25k, and 40k price range tinyboxes:

https://tinygrad.org/#tinybox

It's like buying a 2k-worth used car, and expecting it to perform as well as a 40k one.


Taxi driving is an antifragile profession apparently, to a degree that computer programming could only aspire to.


Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: