make3

9 ms·

211.
by make3 8mo ago
yes but he's 30 not 90, and knowledge and experience continues to accumulate through life, which can certainly compensate
212.
by make3 8mo ago
Amazon and Spotify and Netflix are all enormous examples of not just giving you what you asked for (->recommendation engines). They were there all along, just less egregiously. Even PageRank (the secret sauce in the original Google searc
213.
by make3 8mo ago
Ageism also just one of these shitty unproven biases, like sexism, which is self-realizing by applying pressure to people who fall out of the mold even slightly. He's 30 something, not 90.
214.
by make3 8mo ago
I think that the style itself is very clear and has its advantages, it's hated only because it's from LLMs, which are not liked when used without judgement (which is often the case). So, someone who falls on the side of not comple
215.
by make3 8mo ago
or maybe he's just very good?
216.
by make3 8mo ago
"this design falls flat" he says, as if it was an objective fact and not a personal preference
217.
by make3 8mo ago
model size directly affects latency
218.
by make3 8mo ago
have people tried Antigravity
219.
by make3 8mo ago
it's still a bad metric and OP is also just being loose by repeating some marketing / LinkedIn post by a person who uses bad metrics about an overhyped subject
220.
by make3 8mo ago
still, how many ppl do they kill per mile compared to humans?
221.
by make3 8mo ago
It would be very easy for them to switch the various (compute) cost vs performance knobs down depending on load to maintain a certain latency; you would see oscillations like this, especially if the benchmark is not always run exactly at th
222.
by make3 8mo ago
There's a million algorithms to make LLM inference more efficient as a tradeoff for performance, like using a smaller model, using quantized models, using speculative decoding with a more permissive rejection threshold, etc etc
223.
by make3 8mo ago
It's very clearly a cost tradeoff that they control and that should be measured.
224.
by make3 8mo ago
Your work exploits people's addictive propensity and behaviours, and gives corporations incentives and tools to build on that. Insane spin you're putting on it. At best, you're a cog in one of the worst recent evolutions of c
225.
by make3 8mo ago
Exactly, that sounds to me like a TikTok vs NPR/books thing, people tell everyone what they read, then go spend 11h watching TikToks until 2am.
226.
by make3 8mo ago
Overleaf basically already has the same thing
227.
by make3 9mo ago
I assume that the differentiable renderer is only given its position and viewing angle at any one time (in order to be able to generalize to new viewing angles)? Is it a fully connected NN?
228.
by make3 9mo ago
We could definitely use RL to add blank, invisible "<thinking>" tokens whenever the model thinks it should. Or just allow it to say "Hmm." lol. Easy to test from a technical perspective is all I'm saying, and
229.
by make3 9mo ago
There's no way there's enough data for you to get a model that is anywhere as strong as mainstream languages. If your model is struggling with parentheses, that means it's not even the level of GPT-3 for a mainstream language
230.
by make3 9mo ago
That's easy, AlphaZero had a perfect simulator of the world it existed in (chess, super easy), so it was insanely easy to run simulations of that world ad infinitum, and learn from it. It's simply not the case for the real world,
231.
by make3 9mo ago
With Chain of Thoughts (text thinking), the models can already use as much compute as they want in any language (determined by reinforcement learning training)
232.
by make3 9mo ago
LLMs in their current form rely heavily on the vast amount of human data that's available, to learn from it as a first step (the second step is RL). We're not building a language for LLMs just yet.
233.
by make3 9mo ago
If you look at the list, you'll see that you're incorrect, as C and JavaScript are not at the top. Seeing all the C languages and JavaScript at the bottom like this makes me wonder if it's not just that Curly brackets take a
234.
by make3 9mo ago
I wonder what "94.18% of quality" means
235.
by make3 9mo ago
that's in 2019
236.
by make3 9mo ago
What you're arguing about is the field completely changing over 3 years; it's nothing, as a time for everyone to change their minds. LLMs were not productified in a meaningful way before ChatGPT in 2022 (companies had sufficiently
237.
by make3 9mo ago
I don't know if I'm the only one, but 30 strong friends sounds insane to me. You're a lucky human to have the personality and skills to achieve that.
238.
by make3 9mo ago
"of what I would've seen"
239.
by make3 9mo ago
it's not a crime but applying post processing in an overly generous way that goes a lot further than replicating what a human sees does take away from what makes pictures interesting imho vs other mediums, that it's a genuine repr
240.
by make3 9mo ago
It's just the "cognitive load" UX idea, with extremely non-technical people having extremely low limits before they decide to never try again, or just feel intimidated and never try to begin with. It's the Apple story al

More ›