When I first read Google’s RETRO paper, I was skeptical. Sure, RETRO models are 25x smaller than the competition, supposedly leading to HUGE savings in training and inference costs. But what about the new trillion token “retrieval database” they added to the architcture? Surely that must add back some computational…
As AI creative tools are becoming widespread, the question of copyright of AI creations has also taken centre-stage. But while copyright nerds obsess over the authorship question, the issue that is…
Learn how alien intelligence connects to human intelligence through science and technology using the Principle of Computational Equivalence, the ruliad, rulial space, tiny spacecraft and the Wolfram Language.
Shape From Silhouette Reconstruction from a short cell phone video, showing reconstructed masks.
The first Times crossword Having seen several mind-blowing articles about ChatGPT’s ability (most notably this one), I wondered how ChatGPT would fare solving a cryptic crossword. I decided to try with an easy one: a Quick Cryptic from The Times. I had a search through the Times for the Times archives to try and find…
This technology is already changing the creative industry. Here's how it works.
…and it will look a lot like AGI. LLMs can become extremely powerful by leveraging external cognitive assets and choosing actions.
Left: Original image. Center: Blurred through convolution. Right: Deconvolution of the blurred image. In this post, we’ll have a look at the idea of removing blur from images, videos, or games thro…
Background Backpropagation is a common method for training a neural network. There is no shortage of papers online that attempt to explain how backpropagation works, but few that include an example…
I feel techniques for making machine learning work with small data are not talked about enough. This makes sense because many ML applications are only possible by collecting a huge amount of data. …
We introduce pico.js , a face-detection library written in JavaScript, and show how it works. Although similar projects already exist, our goal is to provide a smaller and computationally more efficient alternative.
In this post we take a look a how the mathematical idea of a convolution is used in probability. In probability a convolution is a way to add two random variables. Using a slightly ridiculous, mad-science, example we walk through multiple way to compute a convolution and ultimate arrive at the formu
I finally was able to try it. And I can assure you it’s impossible to create professional and useful images from a text description.
My personal website
How a Few Activists Made ChatGPT Deny Basic Science
Before asking the question, how do I build AGI, you first must ask the question, what would I recognize as AGI?
This is a guest post, written by Zhangir Azerbayev. Zhangir is an undergraduate at Yale, majoring in computer science and mathematics. He completed this work while visiting Carnegie Mellon’s …
Matthew Butterick
A neural network written in pure Lisp without built-in numbers using only atoms and lists in SectorLISP, a 512-byte Lisp interpreter written by the authors of the SectorLISP project.
Less than a week ago OpenAI unleashed ChatGPT on the world, and it kicked off what feels like a seismic shift in many people’s understand of the capabilities of large …
spoiler alert: not very
Advances in AI have helped automate many tasks that involve processing text. However, text-processing AI algorithms have their own unique vulnerabilities, researchers at IBM, Amazon and University of Texas have found.
