Sinister variations on the positive state of flow

Experiments in reading with LLMs

Chronic boredom causes stress, disengagement, and poor well-being in adults. So why do we glorify it for children?

An overview of the features of the Solveit platform, which is designed to make exploration and iterative development easier and faster.

Chris Lattner on software craftsmanship and AI

A text and code version of Karpathy’s famous tokenizer video.

An email sent to all fast.ai forum users.

Introducing fasttransform, a Python library that makes data transformations reversible and extensible through the power of multiple dispatch.

A friendly introduction to Foundation Models for Computational Pathology

fast.ai is joining Answer.AI, and we’re announcing a new kind of educational experience, ‘How To Solve It With Code’

Pundits say my husband and I are parenting wrong.

Answer.AI is a new kind of AI R&D lab which creates practical end-user products based on foundational research breakthroughs.

We’ve noticed an unusual training pattern in fine-tuning LLMs. At first we thought it’s a bug, but now we think it shows LLMs can learn effectively from a single example.

Moving AI ethics beyond explainability and fairness to justice

Model licensing & surveillance will likely be counterproductive by concentrating power in unsustainable ways

The history of technology suggests that the greatest risks come not from the tech, but from the people who control it

Mojo is a new programming language, based on Python, which fixes Python’s performance and deployment problems.

We’ve released our new course with over 30 hours of video content.

Language is a source of limitation and liberation. GPT 4 pushes this idea to the extreme by giving us access to unlimited language.

Last year, I became captivated by a new topic in a way that I hadn’t felt since I first discovered machine learning

4 videos from Practical Deep Learning for Coders Part 2, 2022 have been released as a special early preview of the new course.

Signups are now open for Practical Deep Learning for Coders Part 2, 2022. Scholarships are available for fast.ai community contributors, open source developers, and diversity scholars.

Practical Deep Learning for Coders part 2, 2022

Prior to 2020, we never expected to homeschool, and now we have committed to it long-term.

Previously, using git with Jupyter could create conflicts and break notebooks. With nbdev2, the problem has been totally solved.

Our favorite tool for software engineering productivity–nbdev, now re-written with Quarto

A complete from-scratch rewrite of fast.ai’s most popular course, that’s been 2 years in the making.

Notes taken whilst preparing a paper on mask efficacy from Nov to Jan 2022. My previous paper on this was written in April 2020 and published in the Proceedings of the National Academy of Science

Following the thread of any seemingly quantitative issue in AI ethics (such as determining if software to rate loan applicants is racially biased or evaluating YouTube’s recommendation system) quickly leads to a host of qualitative questions. Unfortunately, there is often a large divide between computer scientists and social scientists, with over-simplified assumptions and fundamental misunderstandings of one another.

In the west, our ideas of harm are largely anchored to an individual being harmed by a particular action at a discrete moment in time. Yet the harms caused by algorithmic systems are often collective and communal.

Many cultural factors, misconceptions, stereotypes, and obstacles turn people off to math.

Lightning talks by Australian experts on a range of topics related to data science ethics, including machine learning in medicine, explainability, Indigenous-led AI, and the role of policy

While data for social good projects can be useful, there are also pitfalls to avoid.

Things can go disastrously wrong on data science and machine learning projects when we undervalue data work, use data in contexts that it wasn’t gathered for, or ignore the crucial role that humans play in the data science pipeline.

All approved SARS-CoV-2 vaccines provide far more benefits than risks. The very rare risk of VITT from the AZ and JJ vaccines is not due to the spike proteins, but is most likely due to details of their formulation.

Statistical tests need to be paired with proper data and study design to yield valid results. A recent review paper on Long Covid in children provides a useful example of how researchers can get this wrong. We use causal diagrams to decompose the problem and illustrate where errors were made.

From covid-19 to HIV research to the long history of wrongly assuming women’s illnesses are psychosomatic, we have seen again and again that medicine, like all science, is political.

Letter to the editor of the New Yorker on the irresponsible description of a suicide, undisclosed financial conflicts of interest, and omission of relevant medical research and historical context in their recent long covid article

By using better masks, monitoring and improving indoor air quality, and rolling out rapid tests, we could quickly halt the current outbreaks in the Australian states of New South Wales (NSW) and Victoria. If we fail to do so, and open up before 80% of all Australians are vaccinated, we may have tens of thousands of deaths, and hundreds of thousands of children with chronic illness which could last for years.

You can use filters in your mask, your home central air, and your fans to clean the air in your home.

We’re releasing Practical Deep Learning for Coders (2020), fastai v2, fastcore, and fastgpu.

To celebrate the release of fast.ai’s new course, book, and software libraries on August 21st, we’re making available the foreword that Soumith Chintala (the co-creator of PyTorch) wrote for the book

Free, online course from fast.ai and University of San Francisco Data Institute covering disinformation, bias & fairness, ethical foundations, practical tools, privacy & surveillance, the silicon valley ecosystem, and algorithmic colonialism.

You can permanently damage your back, neck, and wrists from working without an ergonomic setup. Learn how to create one for less at home.

DIY masks aren’t just for protecting those around you. They can protect you too, but material and fit matters a lot. Learn what works, based on the latest research.

SARS-CoV-2 does not float in the air. It’s expelled as large droplets, which are easily caught by a cloth mask.

The University of San Francisco Center for Applied Data Ethics welcomes its first cohort of research fellows.

Wearing a mask decreases the number of people infected by an infectious wearer, because it reduces by around 99% the number of droplets that are ejected during speech

Most scientific evidence points in the same direction: keep your droplets to yourself - wear a mask

At the CADE Tech Policy Workshop, members of local government, activists, researchers, and those in industry shared prespectives on how to understand the uses, risks, and opporutnities around government use of technology.

An initiative from STEM teachers in Hong Kong to make reusable masks, for the community, by the community

At the CADE Tech Policy Workshop, experts Renee DiResta and Guillaume Chaslot spoke on disinformation, including the dyanmics that cause it to go viral and attempts towards addressing it.

At the CADE Tech Policy Workshop, Y-Vonne Hutchinson spoke about the role of tech in facilitating mass atrocity, and Catherine Bracy spoke on the need for empathy and collective action.

As governments consider new uses of technology, in public places, this raises issues around surveillance of vulnerable populations, unintended consequences, and potential misuse. There are several principles to keep in mind in how these decisions can be made in a healthier and more responsible manner.

With Jupyter Notebooks and fast_template, we can easily share prose, code, tables, charts, and more!

Overview of 4 part series explaining how to host your own blog without any coding

We show how to use fast_template’s special ‘screenshot’ feature to get high-resolution screenshots

With GitHub Pages you can synchronize your blog with your computer, and write posts with MS Word or Google Docs

An easy and free approach to using an entirely browser-based interface for all your blogging needs

Pretrain a model using labels that are naturally part of the input data, instead of using external labels

There’s a lot more to creating useful data projects than just training an accurate model

nbdev is a Python programming environment which allows you to create complete python packages, including tests and a rich documentation system, all in Jupyter Notebooks

Three months after the launch of CADE, find out what we’ve been up to and how you can get involved.

Unthinkingly optimizing metrics can lead to a variety of grave harms, and what most current AI approaches do is to optimize metrics.

8 important truths about surveillance

Fixing the delegation program using delegates decorator and GetAttr

University of San Francisco is launching a Center of Applied Data Ethics, and Rachel Thomas is the director of the new center.

fast.ai’s newest course is Code-First Intro to NLP. It covers a blend of traditional NLP techniques, recent deep learning approaches, and urgent ethical issues.

Today we are releasing a new course, Deep Learning from the Foundations, which shows how to build a state of the art deep learning model from scratch.