Introducing the Explainable AI Cheat Sheet, your high-level guide to the set of tools and methods that helps humans understand AI/ML models and their predictions. I introduce the cheat sheet in this brief video:
Today we’re releasing a new chapter on the History of the Web. The rest of the book is about how web browsers work, but this chapter is about why. It’s about the goals and motivations for the web and for web browsers, which shape the way they work.
Next.js 10.2 introduces performance improvements (faster builds, refresh, and startup), improved accessibility, Automatic Webfont Optimization, and more!
I’ve been working in Growth for 10 years now and I wanted to reflect on the top 10 mistakes I’ve seen teams make from either teams I worked on directly or mistakes I’ve seen from talking to Growth teams at other startups over the past 10 years. For each lesson I also linked to an […] The post Top 10 Mistakes In Running A Growth Team appeared first on John Egan .
The trigger for this post is the reinstating of Richard Stallman, a very problematic character, to the board of the Free Software Foundation (FSF). I am appalled by this move, and join others in the call for his removal. This occasion has caused me to reevaluate the position of the...
Weights in the final layer of common visual models appear as horizontal bands. We investigate how and why.
We’re releasing Chapter 5. It introduces layout trees, the data structure at the center of layout. Building, laying out, and rendering layout trees is how every web page you view is rendered!
When a neural network layer is divided into multiple branches, neurons self-organize into coherent groupings.
Next.js 10.1 introduces 3x faster refresh, 58% smaller install size, 56% less dependencies, WebAssembly Image Optimization, improved next/image support, and more!
In product-led Growth, some of the biggest opportunities can require working across product boundaries and partnering with other teams. When collaboration goes well it can deliver big unlocks, help build a seamless user experience, and avoid the trap of your product reflecting your org structure. Collaboration not done well can cause friction, contention, and put […] The post The 5 Rules Every Growth Team Should Follow for Effective Collaboration appeared first on John Egan .
.column-center-outer {width:125%} .left { width: 44%; float: left; } .right { width: 56%; float: left; margin-left: 3rem } .row { display: flex; margin-bottom: 3rem; } } .post-body img { border: none!important; box-shadow: none!important; /* Browser specific implementations */ -moz-box-shadow: none!important; -webkit-box-shadow: none!important; } <p&
The intro chapter is here! Now you get to find out why this book was worth writing, mostly in Chris’s words. Plus you’ll learn a few things about where the web came from and how the sausage is made.
We report the existence of multimodal neurons in artificial neural networks, similar to those found in the human brain.
For the last five or six years, since I bid goodbye to the startup scene and Silicon Valley, I have been increasingly working in public. I have written a book, given around 100 talks (many of which are available on YouTube), published over 20 research papers (all freely available from...
“Traducción a Español” Biologically Plausible Deep Learning (BPDL) is an active research field at the intersection of Neuroscience and Machine Learning, studying how we can train deep neural networks with a "learning ru
Neural Cellular Automata learn to generate textures, exhibiting surprising properties.
Unfortunately, the Covid-19 pandemic has been with us now for more than a year and it is likely to remain in some form for many months to come. Researchers haven’t wasted anytime in looking into the impacts of the pandemic on many, many fields including health care, mental health, the news media, social media, […]
This is the video recording of my presentation and consequent Q&A at last week’s Frankfurt AI Meetup. It seems to have struck a chord with a few people and the feedback I got afterward has invigorated me to restart my efforts on ReWriting the Human Story. You can read most of the articles I used […]
Release The GATE Team is proud to announce two new releases that bring GATE and Python together: Python GateNLP (version 1.0.2) : a Python 3 package that brings many of the concepts and the ease of handling documents, annotations and features to Python. GATE Pytho
We present techniques for visualizing, contextualizing, and understanding neural network weights.
Chapter 4 is out! It’s decked out with HTML parsing and error handling, and starts building out the core data structures of the browser. Plus, it has tips for leveraging implicit tags to make your HTML cleaner and simpler!
Reverse engineering the curve detection algorithm from InceptionV1 and reimplementing it from scratch.
A family of early-vision neurons reacting to directional transitions from high to low spatial frequency.
My ML mentees often ask me some variant of the question "how do you choose which papers to read from the deluge of publications flooding Arxiv every day?” The nice thing about reading most ML papers is that you can cut through the jargon by asking just five simple questions. I try to answer these questions as quickly a
Provalis Research in collaboration with the Mixed Methods International Research Association – Caribbean Chapter (MMIRA-CC) will give a FREE 3-hour workshop on content analysis, a powerful deductive technique for text analysis. The workshop will be held online, Friday, February 26, from 12:00 pm to 3:00 pm Eastern Time. Content analysis approach allows analysts and researchers […]
By visualizing the hidden state between a model's layers, we can get some clues as to the model's "thought process". Figure: Finding the words to say After a language model generates a sentence, we can visualize a view of how the model came by each word (column). Each row is a model layer. The value and color indicate the ranking of the output token at that layer. The darker the color, the higher the ranking. Layer 0 is at the top. Layer 47 is at the bottom. Model:GPT2-XL Part 2: Continuing the
Who is doing interesting work on decentralised content moderation? With Donald Trump suspended from Twitter and Facebook, and Parler kicked off AWS, there is renewed discussion about what sort of speech is acceptable online, and how it should be enforced. Let me say up front that I believe that these...
If you thought machine learning on mobile is already a bit of a stretch, how about machine learning on microcontrollers? Take the Arduino Uno . This has an 8-bit CPU running at 16 MHz, no real floating point support — it does not even have a division instruction — and only has 2KB of RAM. Is it even possible to do machine learning on these devices? The answer is yes , but you need to be smart about the al
With a new year upon us, we also have a new blog! In this post, we would like to take a look back at some of the highlights from 2020, as well as some ideas we have for 2021.
I’ve added a code outline tool to help you see the intermediate-level structure of a web browser. As Chris and I write more chapters, it’ll become more and more useful.
<p>Happy new year from Svelte! In the last month we made progress on Sapper's upcoming release, fine-tuned our <code>SvelteComponent</code> typings, and have seen some amazing apps, sites, and libraries coming out in the showcase.</p> <h2 id="What's-changed-in-Svelte"><span>What's changed in Svelte?</span><a href="#What's-changed-in-Svelte" class="permalink" aria-label="permalink"></a></h2><p>A new minor release replaces the <code>SvelteComponent</code> class with a <code>SvelteComponentTyped</c
In the classic 1986 essay, No Silver Bullet , Fred Brooks argued that there is, in some sense, not that much that can be done to improve programmer productivity. His line of reasoning is that programming tasks contain a core of essential/conceptual 1 complexity that's fundamentally not amenable to attack by any potential advances in technology (such
2020 has been a long year. As it comes to an end we wanted to share a special Holiday Update on our research into zero-bundle-size React Server Components.
Interfaces for exploring transformer language models by looking at input saliency and neuron activation. Explorable #1: Input saliency of a list of countries generated by a language model Tap or hover over the output tokens: Explorable #2: Neuron activation analysis reveals four groups of neurons, each is associated with generating a certain type of token Tap or hover over the sparklines on the left to isolate a certain factor: The Transformer architecture has been powering a number of the recen
Chapter 3 is ready. Read about typography, baselines, and tags! Our browser lines up text with style! My favorite aside is the Michigan Supreme Court case on font sizes.
Neural networks naturally learn many transformed copies of the same feature, connected by symmetric weights.
In the 2020 presidential election, most exit polls showed Donald Trump largely holding on to his support among white evangelicals with approximately 81% support essentially the same as in 2016 (AP VoteCast). Trump saw a drop in the white Catholic vote to Biden, who is Catholic, but gained among Catholic Hispanics. Trump held this […]
