Nick Doiron – Medium

Published in MLearning.ai

·Pinned

Chronological language model logic

A group of researchers recently published TimeLMs, a paper and series of models trained on Tweets from each quarter of 2020 and 2021. The paper is compelling, and illustrates how language models age after training with a beautifully gradient-ed table: ‘Up-to-date language models’ is a topic which I wrote…

NLP

3 min read

Published in Towards Data Science

·Pinned

Training NLP Models on Counterfactual Text

Reinflection experiments in Spanish and Arabic — Early this year, I uploaded a seq2seq model to generate gender counterfactuals in Spanish (el profesor viejo <-> la profesora vieja). Since then, I debugged some issues, made a general-purpose library, and created an initial seq2seq model for Arabic. The goal is to see if passing data through process creates…

Seq 2 Seq

4 min read

1 day ago

Podcast Recommendations Part 6

Autism’s First Child — The story of the first child diagnosed with autism: Donald Triplett from Forest, Mississippi. There are thoughtful tangents into early 20th century eugenics, race and autism diagnoses today, and so on. The podcast’s creators wrote a 2010 Atlantic article about Triplett and a 2016 book (In A Different Key: The…

4 min read

May 7

ML Arxiv Haul #5

Time for another hyper-exciting sweep of ML papers : A very preliminary analysis of DALL-E 2 The DALL-E 2 system generates original synthetic images corresponding to an input text as caption. We report here on…arxiv.org This is a less strictly formatted paper, but the authors have access to DALL-E 2 and are posting on arxiv, so let’s go. Just as CLIP / DALL-E floored image generation people one year ago, DALL-E 2 is a clear winner…

Naturallanguageprocessing

6 min read

Apr 30

Update on ML side projects

This is a relatively exciting week in the ML industry. …

3 min read

Apr 23

Reading Blog: April 2022

I’m back on the road for a bit, hopefully with enough reading to keep me busy but not weighed down. I’ve been making progress on Tombstone, the history of the Great Chinese Famine. That will be in my post next month. …

10 min read

Apr 16

ML Arxiv Haul #4

Oh crud, this is too many papers; I should probably rethink this strategy of pasting an Arxiv link here instead of reading or saving it in a tab. I decided to make a daily routine to finally get ahead on filtering and summarizing. A Conversational Paradigm for Program Synthesis Program synthesis strives to generate a computer program as a solution to a given problem specification. We propose a…arxiv.org

Machine Learning

11 min read

Apr 5

Making Sense of PyTorch

I’m a full-stack software engineer who’s been writing Python since One Laptop per Child around 13 years ago. I see that there are plenty of deep learning libraries in Python: PyTorch, TensorFlow, Keras, JAX. …

7 min read

Apr 3

Reading Blog: March 2022

Stories from the Pacific and Mongolia. Delayed this a bit to see if I would finish a Yugoslavia book, but I’m only halfway. At the end I had updates to previous reads and added a mini-review for A Chemical Hunger. Voyagers: The Settlement of the Pacific (Nicholas Thomas, 2021) This book offers a look at Oceania history through a…

5 min read

Mar 11

Podcast recommendations part 5

I am going to keep doing these, because there are so many mini-worlds out there in podcasts, and on Twitter I hardly ever see anyone talk about these. I follow way too much politics and tech on Twitter. One pro-tip: if you can’t see a playback speed control on a…

4 min read