🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

6.) The minmax function doc.rust-lang.org/std/cmp/fn.min…

Chris Olah posted a HN comment saying that he's been trying for the last 10 years to find a use of topology in ML to no avail. This is surprising, for us topology is definitely one of the pillars. E.g. interpreting RL through the lens of control theory makes some of these

Traceoid builds on concepts from harmonic analysis. It is surprising that the Fourier transform, a fundamental concept in harmonic analysis, has been used in Babylonian astronomy in 700 BC.

It’s strange to focus on prompts and not on the real thing: the partition function (attention but better) and energy minimization.

A. A. Markov when you violate the Markov property.

I wish I could mute a million

This esotericization of ML and and brain dynamics is counterproductive. While the two use different building blocks, they have the same goal: energy minimization. Yann LeCun was right all along.

Scary = we do not understand it. AI will stop being scary when it becomes interpretable. Energy-based models are the way forward. This is the Traceoid vision.

If C++ can be as safe as Rust, why is C++ not a popular backend language? Checkmate atheists.

Yann is really right about this, and it is funny that no one besides us at traceoid.ai is working on energy-based models (or anything else besides what is out there right now). Backprop and autodiff are cool and all but they are just stepping stones. Embrace the

There is no verification without Interpretability. Embrace the trace!

Hate to repeat myself, but yeah, there is no AGI without approximating the partition function.

What is the best local file search solution on macOS? I care mostly about PDF and text image search.

Who is the ultimate sigma male? Patrick Bateman or Émile Borel who introduced the sigma algebra?

GPU Mode is a GPU mood.

YC still associates with this fascist?

AGI cannot be achieved with backprop. It's funny how there are so few people working on fixing that part of ML, despite it being so fundamental.