Prioritizing Creative Freedom (055)

Your weekly download of accelerated implementation.

Jun 03, 2024

Happy Monday and welcome to your weekly download of accelerated implementation.

This week I find myself in beautiful Vienna for what was supposed to be a conference trip to talk about centaurs and augmented intelligence – and which turned into an excellent opportunity to spend time with collaborators and clients.

A few days ago I polled our WhatsApp community for interest in either presenting at or simply joining a remote newsletter meetup. The turnout was reassuringly positive which means we’ve started organizing something where community members can showcase what they have been working on and interact with & receive feedback from fellow members. We have half dozen potential speakers already but I want to make sure all readers are equally invited – so if you would like to briefly present your AI-related work to fellow thinkers sometime next month, please either reply to this message and/or join us in the community. Presentations will be as short as possible, with emphasis on group interaction, so don’t hold back if you don’t feel ready to present yet.

In fact, I can’t overstate how delightful it has been to host an ongoing conversation with 150 of my closest friends on WhatsApp – all bringing their personal perspective to the ongoing evolution of AI. If you don’t mind a high-volume & high-signal group, you should absolutely join.

On a different note, last week I sat down with friend Ina Catrinescu for a long conversation about … everything that fascinates me. We obviously talked about AI and possible futures but we also went down philosophical rabbit holes and intellectual journeys, exploring the nature of consciousness, the ethics of technological advancements, and the intersection of science and spirituality and much more. Check it out!

Until next week,
MZ

AI Insights, Challenges and Collaborations

Deep and insightful interview with Jeff Hinton, one of the absolute pioneers in the field, with lucid remarks about where we are heading.

Fast weights for temporary memory... that's one of the biggest things we have to learn.

AI For Good Global Summit

I haven’t found a good edit of any of these presentations, but last week in Geneva the UN-ITU organized an AI For Good Global Summit which featured a livestream of the entire session.

The Power of Storytelling in Understanding AI

One more excellent lecture from the Sana AI Summit where author Jeanette Winterson delivers a powerful philosophical take on AI, religion and alternative intelligence.

I finally realized what I was missing: the story. You see, humans need stories to understand ourselves.

Do impossible things

Microsoft CTO Kevin Scott’s suggestion for what to focus on in the coming years.

Transcending current limitations of LLMs

If you have spent significant time prompting you might have noticed how often instructions and content from the middle of the context window gets lost, as inference tends to overemphasize the beginnings and ends. This paper explains how to solve for this and make LLMs consider more of input in its instructions. Excellent summary by Ruben Hassid.

What's interesting to me isn't so much the particular incremental innovation - but the evidence that more or less every shortcoming that we notice today in AI systems is actively being researched and worked on.

Consciousness and phenomenology

Tangentially related to AI, but this is an amazing visual essay exploring the nature of consciousness, first person experiences, phenomenology and science. Excellent use of animation to convey complex concepts.

The receiving end

Brilliant short essay on what it feels like to receive an AI generated email from a friend. The author captures what bothers so many of us when comes to “consuming” AI generated slop. The feeling that the sender hasn’t put effort into writing but expects the reader to exert energy reading. I’m sure we’ll see a lot more such experiences and reactions, even as llms get better with time.

It felt like getting a birthday card with only the prewritten message inside, and no added well-wishes from the wisher’s own pen. An item off the shelf, paid for and handed over, transaction complete.

The real threat with super intelligence is falling prey to the hype

Incredible essay by Navneet Alang in The Walrus.

Today, AI can take previously unconnected, even random things, such as the skyline of Toronto and the style of the impressionists, and join them to create what hasn’t existed before. But there is a sort of discomforting or unnerving implication here. Isn’t that also, in a way, how we think?