Michael Marino

Grad student, AI engineer, and reader of things. Aude aliquid dignum.

[Michael Marino @mjmarinoo]

"He seemed too submissive to Paul, but then the Sardauker had never been prepared for such happenings as this day. They'd never known anything but victory which, Paul realized, could be a weakness in itself."

[Michael Marino @mjmarinoo]

@DrJimFan @johnschulman2 is the talk available online somewhere?

[Michael Marino @mjmarinoo]

"Before we begin ruthlessly deconstructing the amateur life, let’s pause for a moment to give it its due. The amateur life is our youth. It’s our hero’s journey. No one is born a pro. You’ve got to fall before you hit bottom, and sometimes that fall can be a hell of a ride." - SP

[Dimitris Papailiopoulos @DimitrisPapail]

For anyone starting out with transformers as language models, this is *by far* the most complete, to the point, and non-BS expository document.
https://t.co/OV6pZMs3Vy

major extra points for having zero figures of chaotically intertwined boxes

[Michael Marino @mjmarinoo]

@therobotjames https://t.co/C7ewpmUdkD

[Michael Marino @mjmarinoo]

@tunguz @TraderRobJ lol I disagree with the original tweet, but this one sounds more realistic

[Michael Marino @mjmarinoo]

@emilymbender Whether or not industrial entities are publishing papers isn't related to whether or not cultural biases of the masses give people a different impression of industry vs. academic experience that relates to rampant anti-intellectualism... and the sentiment of the original comment.

[Michael Marino @mjmarinoo]

@smiteri There’s certainly plenty of mathematicians working in non-academic contexts. Math is much broader than what people are doing with it in academia.

[Michael Marino @mjmarinoo]

@emilymbender ...which is kind of funny -- as if the computer I'm typing on would have been possible without the computational/electromagnetic theories underlying its design.

[Michael Marino @mjmarinoo]

@emilymbender I think even for CS/Math/Engineering people, there is a general bias against people whose experience is primarily academic. Seems associated with general anti-intellectualism, as if theoretical results are not worth pursuing...

[Michael Marino @mjmarinoo]

The thing about heuristic devices is that the device itself can't tell you whether or not you should be applying the device. To assume a particular heuristic is always the solution is to make a fetish out of a tool.

[Michael Marino @mjmarinoo]

@egrefen I find her criticisms thoughtful in general. It’s fairly obvious LLMs have a lot of utility and will reshape a lot of tech in very interesting ways but that doesn’t mean there aren’t significant theoretical concerns with their usage that should be a part of the conversation.

[YK aka CS Dojo 📺🐦 @ykdojo]

As a developer, if you have a lot of trouble understanding or building something, remember - that's probably when you learn the most.

[Arvid Kahl @arvidkahl]

The hard truth is, in most cases, the solution to "how do I get better" is to become more skilled and put in the time. No shortcuts, just hard-earned results.

[Michael Marino @mjmarinoo]

This one sentence encompasses so much of the route technology is likely to take in the coming years. Guessing it’ll be some flavor of clearly amazing but potentially terrifying.

[Ben Tossell @bentossell]

AI is huge. Think bigger.

Instead of X company + AI feature, rethink everything from first principles with AI in the picture from day zero.

[Michael Marino @mjmarinoo]

@tibo_maker Trust myself.

[Michael Marino @mjmarinoo]

@concodanomics I’d really like some of your takes on “modern monetary theory”.

[Michael Marino @mjmarinoo]

@therobotjames Super awesome thread.

[Michael Marino @mjmarinoo]

@shaneparrish Thinking in Systems by Donella Meadows. Key takeaway: the key to understanding and anticipating system level behavior is in feedback mechanisms and their interactions. Surprises are usually due to a significant impact of a higher order interaction that isn’t being considered.

[Stanislas Polu @spolu]

A list of predictions for 2023 for the field of LLMs🧵

[Michael Marino @mjmarinoo]

The danger of this phenomenon, both in AI and in people, is difficult to exaggerate.

[Sam Altman @sama]

@thatfollowed it does know a lot, but the danger is that it is confident and wrong a significant fraction of the time