xylar

I'm going to get started in #opensource
Into fullstack web-dev, ML & AI, (cautious) Blockchain, etc
Also retweeting useful stuff!
I learn and build things ✨

[xylar @xylardev]

Ain't no way my guy

[xylar @xylardev]

Aaand here we are

[xylar @xylardev]

Re-subbed to Claude Max just to try out Fable and see what's the hype about. Let's see if it holds up

[xylar @xylardev]

Re-subbed to Claude Max just to try out Fable and see what's the hype about. Let's see if it holds up

[xylar @xylardev]

@merishabh_singh Hello, I've been updating my profile on your platform. How do I submit this form if I'm currently still working here?

[xylar @xylardev]

@SRustagi1996

[xylar @xylardev]

@merishabh_singh I've developed and shipped frontends to solve complex problems. My resume is here https://t.co/iG1y1yfDBl

[xylar @xylardev]

@ThakurAbhay342 Hand em mine and the 5k is yours :D
https://t.co/0XUH1Q9v75

I've worked on stuff like making a Wix-like platform and hospital management portals.

[xylar @xylardev]

@merishabh_singh Interested, here's my resume:
https://t.co/iG1y1yf5LN

[xylar @xylardev]

@ThakurAbhay342 Well, holy moly that's a lot of replies

But for what it's worth, for SDE-2: https://t.co/0XUH1Q9v75

I'm a full-stack dev that has worked on healthcare provider portals and building Wix-like no-code apps.

[xylar @xylardev]

@merishabh_singh I've 3 years of experience working with full stack applications such as a Wix-like website builder and hospital management platforms. My resume is attached

https://t.co/iG1y1yf5LN

[xylar @xylardev]

@ikoichi This is very helpful! I'm curious what extra steps a person with little to no following would need to take for this though

[pilcrow @pilcrowonpaper]

*If* you want to contribute to Lucia, migrating OAuth providers from Lucia to Arctic would help me and the project a lot. There are 17 minor OAuth providers left

[Yohei @yoheinakajima]

obsessed w this new learning tool

[murat 🍥 @mayfer]

PSA: you can search through 100k gpt-3 embeddings in an instant on a macbook with a basic for loop.

you'll learn the lessons for getting good results locally. so once it's time to go big you already know what you're doing

don't waste time with vector stores when tinkering

[Andriy Mulyar @andriy_mulyar]

Announcing GPT4All-J: The First Apache-2 Licensed Chatbot That Runs Locally on Your Machine💥
https://t.co/LLLugDQODr

Large Language Models must be democratized and decentralized.

[xylar @xylardev]

@fireship_dev 🫂

[Danny Postma @dannypostma]

"You're only successful because you have a large following."

BS. Here's how you can do the same with 0 followers 🧵👇

[Eliezer Yudkowsky ⏹️ @ESYudkowsky]

Maybe I shouldn't, but one last shot at explaining my position on AGI gatekeeping.

I'm not saying that I know everything known to the high-status inventor or engineer of the largest LLM.

I'm saying that I'm dubious that they have secret knowledge deeply relevant to *alignment*, which they cannot explain, cannot cite a paper about, cannot even gesture at, by which they know that superintelligence poses no threat to humanity.

There is a story almost certainly apocryphal, that when Diderot, a French atheist philosopher, came to the court of Catherine the Great, Catherine asked Euler to debate Diderot on atheism; and Euler said...

"Sir, (a+b^n)/n = x; hence God exists, answer please!"

...and Diderot could not answer, so he went back to France in shame.

This would be hugely dickish behavior by anecdotal!Euler, of course; and socially destructive commons-burning, because sometimes technical experts do need to appeal to arcane technical expertise to say, no, that bridge will fall down, don't build a bridge like that. In reality there is no reason to believe that Euler was the kind of person who'd do such a dickish thing, and Diderot had written on probability and wouldn't have been fooled.

Probably the false anecdote got repeated by people who saw it as a burn on atheists, and did not quite see it the way I would, that I hope you would, about whose reputation was being impugned.

This is not what we have here. What we have here is, "Hence God exists" but they won't write down any actual specific fake equation. Why? Well, because if they tried that, I, like the actual historical Diderot, am not innocent of technical details (even if I am not on the level of the equivalent of Euler), and would tear apart their bluff in two and a half seconds. Clearly the would-be gatekeepers know that, on some level that cannot rise to conscious reflection, which is why they give no details.

There *are* technical details about AI that are important up at the macro level, to the question of whether superintelligence kills you. You need to know that every parameter in the model is being differentiated with respect to the loss function, for example. If you don't understand stochastic gradient descent on that level, you really *aren't* qualified to think about how to align deep learning systems; you need to know where the optimization comes from, at some very basic level. But that really is not *complicated* math.

If you want to talk about macro alignment issues, ideally you'd also have studied - not that my critics would have bothered to learn this - the very basic math of evolutionary biology; and you'd be able to calculate how much selection pressure gets exerted on a mammalian genome per-generation. This helps see how hill-climbing to produce an AI model might be different from evolution's hill-climbing to produce hominids. I can tell you in words that the information bottleneck was "vastly tighter" on genomes, but it's better to grasp the difference quantitatively and see how it's derived.

Am I gatekeeping there? Only at people who tried gatekeeping at me first. And I SPELL OUT what it is that I think is the important technical detail and WHY, and I don't sniff and posture that you can only possibly *really* understand all the implications by spending 6 years studying evolutionary biology. There *are* further technical details in evo-bio but they are not *relevant*, so far as I know. You can look up Fisher's Fundamental Theorem of Natural Selection and Price's Equation and ponder those actually quite simple bits of math, alongside the quite simple math of stochastic gradient descent; and someone who is sufficiently good at seeing the implications of simple math without someone else spelling them out, will already have passed that low gate that I briefly gestured to there.

That's how to not be story!Euler; that's what it looks like to say you know something important and technical and *not* be bluffing.

[Brian Roemmele @BrianRoemmele]

You will own your own AI.

Final testing on a new massively smaller 100% locally running ChatGPT 3.5 turbo type of LLM AI in your hard drive on any 2015+ laptop.

I will have pre-configured downloads and it is massively smaller than most models I have, just 4gb.

Out soon!

[xylar @xylardev]

@G3rgoPasztor 🤔 I need to get new glasses fr