View profile

Mostly Harmless AI

Mostly Harmless AI
By Alejandro Piad Morffis • Issue #3 • View online
🖖 Welcome to the third issue of Mostly Harmless AI! Every Sunday, I’ll send you a curated list of interesting bits about AI all around the spectrum, from cool new research and tools to news and discussions, and everything in-between.

🗞 What's new
Recent news about or related to the technological, scientific, and social aspects of Artificial Intelligence.
The geek news of the week is probably the landing of Perseverance on Mars. This has been a long project for NASA, concluding in a 6 months flight that ended a couple of days ago with a flawless landing, but starting years ago. And it couldn’t be possible without lots of AI. Perseverance is full of firsts, including the first helicopter flight on another planet, and a bunch of new equipment devoted to detecting any possible trace of (even extinct) life on the Red Planet.
Thinking about this, it becomes clear that if we ever meet an alien civilization, that First Contact is gonna be carried out by our robots meeting their robots. AI is going to be our ambassador to the Cosmos, so we better get it right. And getting it right is not just a technical problem, it starts and ends with ethics: how do we instill our own values on a sufficiently advanced AI. And judging by how we are treating our own people, we seem to be failing at this: in what amounts to a one-season drama series by now, Google just fired another top-level AI Ethics researcher.
📚 For learners
Online resources for learners at all levels of expertise: online courses, YouTube videos, blog posts, free eBooks, novel research, and more.
If you’re interested in deep learning, in terms of learning resources it’s hard to get it much better than with MIT’s Introduction to Deep Learning. 10 lectures in video format, totally free, for you to download and binge-watch this Sunday! Ranging from the basics of Tensorflow to vision, sequence, and reinforcement learning, and ending with practical sessions and shareable code.
MIT Deep Learning 6.S191
🔨 Tools of the trade
Apps, libraries, online services, and tools, in general, that you can use to solve AI problems.
The state-of-the-art in NLP today is, hands down, the Transformer architecture. And the state-of-the-art in transformer implementations is HuggingFace’s transformer library for Python. They implement over 40 different architectures, in Pytorch and Tensorflow. And they recently published a web interface where you can search for models and datasets and quickly find the best implementation for your problem.
Hugging Face – On a mission to solve NLP, one commit at a time.
🍿 Recommendations
Podcasts, books, TV, and cinema; recommendations that will spark your interest in AI or make you think deeply.
Artificial Intelligence is the most powerful technology of the near future. It will change (and already is changing) the way we approach almost all areas of our lives. One of the most pressing issues with AI is getting it to learn our key values so that it doesn’t end up destroying society as we know it. This is called the alignment problem and is a major research endeavor as well as a very practical one. In a book with that same title, author Brian Christian takes us on a mind-blowing journey through the most complex problems in AI ethics, from reducing inherent data biases to making AI that believes in what we believe. It’s a powerful book, full of very technical topics, but delivered in a very approachable language, suited for all levels of expertise.
🎤 Word of mouth
Interesting conversations about AI happening all around social media, where you can go listen to others and share your thoughts.
Two interesting conversations I found on Reddit this week, for those interested in pursuing a research path in machine learning are what to do with a paper rejection, and how much does the prestige of your institution affect your future career prospects. As usual, Reddit is full of witty and humourous replies, but also very deep insights from a community of like-minded people.
My own weekly AMA has some interesting conversations as well, both on technical topics, such as how to handle imbalanced datasets, and more philosophical topics about the future of AI and education. Check it out!
Alejandro Piad Morffis
Hey folks 🖖!

🎙️You know the day: Saturday, and you know the drill: let's have yet another AMA about artificial intelligence, machine learning, computer science, or anything else you'd like to get my views on!

I can't promise I'll be right, but I'll try to be useful.

Shoot 👇
👥 Community
Interesting people from the AI community that you can follow, from big influencers to tiny accounts, but always people that are worth listening to.
This week I want you to give your Twitter timeline an upgrade, and follow Apoorv. He’s everywhere, from software engineering to cryptography and blockchain, to psychological biases, my timeline always feels fresh when Apoorv shares his latest insights. He just started a newsletter as well and writes on his blog at Hashnode.
And if we’re talking upgrades, check out Tivadar’s threads on fundamental math concepts, explained in beautiful simplicity. He’s the co-founder of Telesto.ai, a growing competition playground for AI, and the creator of the modAL, the most used active learning framework in Python. Tivadar is very knowledgeable in machine learning (he’s got a Ph.D., you know) but he is also super approachable and always eager to help others.
☕ Homebrew
The latest bits of my own harvest: Twitter threads, blog posts, projects, videos, and any other piece of content I’m producing.
For a couple of weeks, I’ve been working non-stop on a new idea: starting a podcast! I’m planning on publishing the first episode sometime next week 🤞. If you’re interested in checking it out, you can subscribe anytime.
The Mostly Harmless AI podcast
While you’re here, let me plug in some of my older stuff. Here you can find a few of the things I’ve written in the past. These are short essays on anything from the nature of education to how to fix peer review. Full of my own biases and opinionated nonsense, so beware!
And here you can find my most carefully crafted Twitter threads in plain format. They range from practical stuff, tools, and libraries, to theoretical and philosophical issues.
👋 That’s it for now. Please let me know what do you think of this issue, what would you like to see more or less of, and any feedback you want to share. If you liked this newsletter, consider subscribing (in case you’re not) and forwarding it to those you love. It’s 💯 free!
Did you enjoy this issue?
Alejandro Piad Morffis

A weekly newsletter on all things AI, including recent news, hot resources, and interesting conversations happening all around the Internet.

If you don't want these updates anymore, please unsubscribe here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue