Hey!! Replying on mobile so I'll be sparse, but ya! I basically agree with you, I read your comment in good faith and I knew most of this but I appreciate it anyway. Good + informative comment, esp for the shoutout to give directly.

I used to identify as an effective altruist for basically all the same reasons you are pro-EA!

I am someone who is quite pro-EA in the sense of "charity donations can go a lot farther if you apply them thoughtfully, I love it when fewer children die from malaria", which in my mind is/should be the core Thing of EA, and I deeply deeply resent the AGI people sucking all the oxygen out of the room and torching piles and piles of goodwill the way they do. So please don't read the following as a defense of the AI doomers and longtermists, I find them very tedious and frustrating:

1. The reason EA types don't tend to touch on climate change is because of its core Thing which is trying to find and address problems that are "important, neglected, and tractable". Stopping/reversing climate change is both Important* and Tractable, but it's really not Neglected-- there are tons of really really smart people working on it, it's far from low hanging fruit, and that means the floor for meaningfully contributing to those efforts is pretty high, diminishing returns and all; Altruism in this area will be less Effective. This is definitely a bullet to bite but I think it basically makes sense.

2. I think EA was at one point trying to do stuff about nuclear threats? Maybe it still is, idk, my understanding was that it ended up really not being very tractable, like, how do you know you're reducing the overall threat of nuclear war and proliferation? What does that look like? The most tractable part of that that I can think of (getting existing nuclear weapons dismantled) is not very directly solvable with money, which is the primary tool EA uses (since the original point is making a donated dollar do as much good as possible)

3. Now you might ask, well, how does AI extinction risk rank in important/neglected/tractable, I mean for one how tractable could "AI alignment" possibly be? To which the AI people will say "shhhhhhhhhhhhhhhh"

* (The AI people will say it's not that important because climate change isn't going to drive humans extinct, let alone wipe out all life on earth/the galaxy the way they say AI will, it's just going to make lots of people die and make life worse for everyone who survives.)

Anyway even if EA never does anything tangibly good ever again (unlikely imo, the We Want Less Kids To Die From Malaria faction is still going strong), I think they at least get credit for GiveDirectly, which lets you give direct cash transfers to people living in extreme poverty around the world with no strings attached, so they get to decide where it's best spent, rather than someone half a world away dictating that to them. I highly recommend anyone and everyone donate money to GiveDirectly if they have some spare cash lying around.

In conclusion, sorry if you already knew all this or it's an illegible wall of text, it turns out I was in an infodumping mood, what can ya do.

This is 100% right. It's crazy seeing "effective altruism" people center around "AGI" (which is a stupid word btw) for no reason whatsoever

Yeah I definitely think AI doomerism/boosterism is a distraction from both climate change and nuclear threats.

A bit of both.

The best models still absolutely fucks up code in the same way it does prose, but GitHub's Copilot gets it right 90% of the time. If you're writing in a new language, you can end up with an MVP waaaay faster than you would've before. If you have a method signature with descriptive names, it can usually fill it out for you. If you're a product manager, instead of hiring a team of datascientists, you might just hire one.

I think it's fair to say the supply-vs-demand of programming expertise changed a lot over 2023, purely because of these tools. That's my bread-and-butter and it's depressing.

With Google and other search engines turning to garbage, communities/support becoming increasingly dependent on Discord, and StackOverflow also turning to garbage (with AI answeres), so I found myself just giving up and checking ChatGPT and Copilot instead.

A lot of my coworkers are vocally dependent on these models, and suddenly I was one of them too.

AFAICT, this kind of replacement isn't happening wholesale for artists. People are using AI models in asset-creation pipelines and for decoration and marketing materials, and people are using models instead of commissioning art. But AFAICT there's not the industry-wide "ask ChatGPT rather than your coworkers or docs" that I'm seeing.

This makes sense, because "How do I unwrap this observable?" has one correct answer, whereas you really can't find a model that can, say, "Create concept art for an 80s-inspired retro desk fan prop in our South African inspired animal people world".

TLDR:

• Current models still mess up, but get things right very often.
• Programmers are definitely becoming dependent on these models.
• Programmers are definitely being replaced, and my understanding is that artists aren't being replaced as widely yet.

AI is replacing programmers way, way before it's replacing artists

aren't large language models messing up codes much in the same way they mess up prose? or do you mean other types of NN.

(did you even mean that sentence to imply imminence in the first place)

fartificial intelligence