jstx1

33 ms·

361.
by jstx1 4y ago
Your girlfriend broke up with you and you don't like the subject of your PhD. Those are quite common experiences and very far from rock bottom. So just keep going with your life and things will be fine, because they're kind of fin
362.
Living with Rust Long-Term – Jon Gjengset(youtube.com)
33 points by jstx1 4y ago|6 comments
363.
by jstx1 4y ago
I don't like Julia either but Carbon isn't ready for use - https://github.com/carbon-language/carbon-lang - and even if it was, it would still need an ecosystem of libraries which it doesn't have.
364.
by jstx1 4y ago
Still sounds like a system design interview to me, just a bit less structured.
365.
by jstx1 4y ago
> The future of software is asking smart questions to a mesh of APIs running layered models Seems like a horrible future and I'm concerned that if enough people keep repeating this, it might happen.
366.
by jstx1 4y ago
> But what rules are they discovering The point of ML is that you don't think about the rules explicitly, you let the algorithm figure them out from the data. > how do they become so good at text continuation? Good underlying alg
367.
by jstx1 4y ago
What you're describing is how to turn an LLM into a chat bot like ChatGPT. OP is asking about LLMs which by themselves don't need any reinforcement learning.
368.
by jstx1 4y ago
They take a sequence of tokens and they do their best to continue the sequence in a plausible way. That's the basic idea - given the text so far figure out what's likely to follow.
369.
by jstx1 4y ago
> I have been tired of working from home for a good while now. I’ve felt like it hurt my productivity and I really couldn’t put my finger on why. Same, I just don't care about anything at work, and without any people around me I don
370.
by jstx1 4y ago
But he's an OpenAI cofounder? And also just recently laid off a bunch of ML researchers from Twitter's Cortex team?
371.
by jstx1 4y ago
> Over the longer term, we'll focus on developing AI personas that can help people in a variety of ways. It's easy to imagine a future in which Facebook have a paid monthly subscription to a friend, partner, therapist etc. >
372.
by jstx1 4y ago
Can something like a fitbit be used as an initial screen? Mine detects multiple very short sleep interuptions every hour during sleep and I have no idea how to interpret this. (I'm chronically tired and distracted and I have no idea wh
373.
by jstx1 4y ago
That's a completely different thing though. Python <> Nim is nothing like JS <> TS. And who cares about the syntax if you're using a completely different language, you can pick any typed language at that point.
374.
by jstx1 4y ago
I've only seen it come up in discussions about job-hopping and what the optimal time to stay at one job is.
375.
by jstx1 4y ago
All of this is super interesting but it's been discussed for a while and I'm still waiting for a "do $x, it will improve condition $y through mechanism $z" kind of prescription and I don't think that's coming a
376.
by jstx1 4y ago
> Can one argue that the random initialization of a model is akin to this starting point? It's a starting point but it's random, unlike the starting point of a human being which is very very not random.
377.
by jstx1 4y ago
A child doesn't start from scratch, they have billions of years of evolution encoded in their DNA.
378.
by jstx1 4y ago
You don't need to be dismissive - demonstrating what can be achieved with a lot of data/compute and additional human-supervised learning is progress.
379.
by jstx1 4y ago
Just because a bunch of crypto grifters are finding a new field to dip into doesn't mean that there hasn't been a lot of genuine progress in AI in the past 6 months.
380.
by jstx1 4y ago
> I think it's better to think of JAX as a more general framework for differentiable programming and PyTorch more focused specifically on deep learning/neural networks. I don't get the point of this distinction - JAX was
381.
by jstx1 4y ago
"Someone could make an organization that solves the problems" is hardly an idea.
382.
by jstx1 4y ago
The research is open, you still need 1) to have a good idea for a product that people want 2) lots of talent and resources to actually build and run everything
383.
by jstx1 4y ago
I bet they're still the majority - the people who only use their phone do it through apps, not through their mobile browser.
384.
by jstx1 4y ago
> A Value Qualifier is a proposed solution for a key problem that the company is facing delivered to the hiring managers to showcase applicants’ skills to qualify for the role regardless of background and seniority. Aka your ticket to sk
385.
by jstx1 4y ago
1. No 2. Miss what boat? If they have a legitimate product, the hype cycle doesn't matter too much. And it's not like anyone is competing with them right now.
386.
by jstx1 4y ago
It boils down to "do I need ML?".
387.
by jstx1 4y ago
What the article is saying is that when you develop and build ML systems, your workflow is different from the workflow you would have if you were solving the same problem without ML. The weird part is presenting it as 2.0 which implies that
388.
by jstx1 4y ago
Are they paying more than your current FAANG job?
389.
by jstx1 4y ago
> I've started applying to private companies that do ML research (OpenAI, Meta, DeepMind, NVIDIA, etc.) If you want to do research at a top company, yes, you probably need one. Or at the very least you need to somehow compete with o
390.
by jstx1 4y ago
I also don't know if Rust aligns well with the goal of reducing complexity. Programming in Rust gets really bogged down into thinking about programming language abstractions rather solving whatever problem you need to solve.

More ›