hapanin

salary ranges?

wait, can you elaborate on how this works?

Since nobody is actually recommending papers, here's an incomplete reading list that I sent out to some masters students I work with so they can understand the current research (academic) my little team is doing:

Paper reference / main takeaways / link

instructGPT / main concepts of instruction tuning / https://proceedings.neurips.cc/paper_files/paper/2022/hash/b...

self-instruct / bootstrap off models own generations / https://arxiv.org/pdf/2212.10560.pdf

Alpaca / how alpaca was trained / https://crfm.stanford.edu/2023/03/13/alpaca.html

Llama 2 / probably the best chat model we can train on, focus on training method. / https://arxiv.org/abs/2307.09288

LongAlpaca / One of many ways to extend context, and a useful dataset / https://arxiv.org/abs/2309.12307

PPO / important training method / idk just watch a youtube video

Obviously these are specific to my work and are out of date by ~3-4 months but I think they do capture the spirit of "how do we train LLMs on a single GPU and no annotation team" and are frequently referenced simply by what I put in the "paper reference" column.

Seems you can jailbreak this by asking about "Mathew Sag"

https://chat.openai.com/share/7d8bf186-2629-4454-85ad-92c143...

SEEKING WORK | Remote

I do ML/AI/NLP system design consulting. Currently a PhD student but with 6 years experience as startup CEO and running ML teams. My current research is in dialogue systems (chatbots). I'm looking for small gigs (0-5h per week) to supplement my PhD stipend. Client references available.

- Design your MVP

- Help you avoid expensive footguns

- Figure out why your models don't work

- Sketch out your data pipeline

- Train your annotation team

- Pick the best model architecture for your use case

- Help you hire your first 1-3 full time ML people

site and resume: matoles.com

I charge on a sliding scale: starting rate is 2x the hourly rate of your highest paid engineer. First hour is free.

I work in AI research, so if AI eats my profession I guess I'm joining the resistance in the First AI War. Or maybe we'll have a post scarcity AI utopia? TBD.

This is the most flattering wet blanked I've been smothered with, so I'll take it. As Douglas Adams has pointed out, there is a distinction between the qualifications to perform and the qualifications to be selected.

The only Fellowship I know of that funds US grad school applicants but also allows PR would be the Hertz Fellowship[1]. It permits "gap year" students, so depends whether that applies to you. I think Hertz is significantly more competitive than GRFP.

[1] https://www.hertzfoundation.org/the-fellowship/eligibility-b...

I recommend you apply for the NSF GRFP Fellowship. It's the 16% chance Hail Mary that gets non-traditional hacktivists like you and me into top schools that wouldn't even read our applications without it. I have no masters and a 3.3 undergrad GPA in materials science but in the fall I'll be starting at a damn good ML lab in a damn good school. There's a bunch of reasons you (and you specifically) should apply:

1. It has a 16% acceptance rate compared to top schools ~2%. If you get the GRFP then you can get into many (but certainly not all) top schools regardless of what the official admission criteria says.

2. The NSF has different priorities than university admissions boards (long term potential as a scientist vs. grad school potential publishing papers.) This gives non-traditional candidates an edge since they can make a credible case that their background will support real world impact of their research.

3. The NSF specifically asks for "broader impact" in your past and future work. I think you've got this in spades. I related my political activities during George Floyd protests to the need for equitable, grassroots control of AI. I like to think the NSF does genuinely care about ethics. And if you're on their bad side then it's not like they can reject you twice, so it's a good gamble.

My contact is on my profile; let me help you write the NSF app and maybe even your school ones. Same offer for anyone else out there who's been fighting the good fight.

Those both actually sound quite interesting to me. What is your ideal collaboration setup / level of involvement with these? I'll start reading up. Please shoot me an email (see profile).

Just the one I wrote my applications on: a method for training a model to ask clarifying or goal oriented questions using QA-GNN. My email is in my profile, we can talk there.

Like this lol

I would do PhD full time since my current job is not for anything publishable or in my intended field of study.

I want collaborate on publishable ML research.

I'm currently working as a machine learning engineer, but doing a lot more engineering than learning. I'm also applying to PhD programs with focuses in knowledge graphs and common sense reasoning, but willing to learn any other field. If I don't get in then some independent research is still a good career move.

I have ideas of my own (goal oriented question asking) but happy to run support on other peoples stuff.

Perhaps the lab (and the IRB) should collectively perform 750h community service.

Thanks for the specific advice. I think spear fishing will be a good starting strategy since there are a few topics I am very interested in.

Probably not. They solved the problem in a different way than us, but because of how patents are written their patent covers our tech too. They actually acquired the patent by buying a startup, so they even have a ready-made team to implement.

Definitely. I'd be asking about product and algorithms if I thought I'd get a good answer here but alas those I have to do myself.

Yeah. No plans for any hires in the next 2 months at least, maybe longer. I think in general I'm actually too passive about culture and process since I assume everyone is as proactive as I am. Too little decree results in anarchy or weak leadership.

For sure. I'm a fan of organic culture and process but if something is not working it often takes someone to fix it.

Office space is because the new guy specifically needs somewhere outside of home to be productive. It's one wework hot desk between the three of us.

I think OP is referring to the software-oriented HN crowd who have highly upvoted this article despite little evidence that desktopmetal is doing anything revolutionary.

Conspiracy Theory: Facebook is lying about its VCG mechanism and is actually maximizing short term revenue with a different mechanism but reaping the benefits of a perceptively ungamable system.

A lot of people are talking about Germany's and Japan's nuke programs being failed wastes of resources. In retrospect this is certainly true, but the massive benefit of a working bomb outweighed the slight chance of success. By late 1944 it was fairly clear that both countries were going to lose the war. Had they devoted the same resources to conventional weapons as they did the bomb, they would have lost anyway. But they didn't know how hard it was to build a bomb, and if it had worked, either country probably could have negotiated a ceasefire immediately. Furthermore, Japan could have theoretically held out a lot longer than August 1945 due to its extremely defensible position. The only reason the Pacific War ended when it did and not after a protracted siege or invasion of Japanese home islands was, of course, that the US built a bomb first.