Halfway to 2025

Happy Monday and welcome to your weekly sense-making experience around AI.

Today marks the halfway point of the year and might serve as a reminder that only transformation is certain. Having spent the weekend moving house here in Amsterdam, I have very little in terms of personal commentary this week – only that 1) forcefully stepping away from screens can be a relief 2) good housing in a dense city is a tremendous privilege and 3) extending my abilities with AI has become second nature and it would feel like a cognitive downgrade losing access to it.

Trivial things like finding garbage zoning information, gardening basics or organization tips can be gleaned through LLMs, which of course was always possible with search engines, but now offers approach which adapts information to your needs. I literally cannot imagine this going away.

If this newsletter tickles your fancy, consider joining our WhatsApp group, which has by far become my favorite spot on the internet to follow & discuss AI.

Until next week,
MZ

---

Via @constantbageltherapy

---

Building with AI systems

Recommended read from Substack: another write-up about “A year of building with AI systems” that was linked to previously. While these insights are pretty technical, I believe anyone can take away applicable learnings from how this particular set of experts is using LLMs to build and test software. Last week’s 3h livestream is highly recommended as well.

[

Vanishing Gradients

42 Lessons from a Year of Building with AI Systems

Hi all! I’ve recently started a newsletter around all things data science, ML, and AI, primarily to keep track of interesting things in the space and what I’ve been up to. This is an experiment so please do let me know what you’d like to see here. There’s a lot to share this week so let’s jump right in…

Read more

2 years ago · 1 like · Hugo Bowne-Anderson

](https://hugobowne.substack.com/p/42-lessons-from-a-year-of-building?utm_source=substack&utm_campaign=post_embed&utm_medium=web)

---

Synthetic and structured real-world data

Marginally related to AI, but this is a spectacular 30-min presentation from IKEA's head of 3d visualization / spatial computing by fellow Swede, Martin Enthed. The technical details are incredible and easy to follow, and naturally there is an AI angle around labeled synthetic data.

---

On the disproportionate effects of AI.

These YC sessions are pretty high signal and don’t meander much.

---

I'm curious about Ray Kurzweil's new book, The Singularity is Nearer. The Singularity is Near kicked off Envisioning unconsciously back in 2005. I'm not necessarily a singularitarian but his predictions have been remarkable. Good summary of the book on X.

---

Toys R Us + Sora commercial

No commentary, just an interesting and often uncanny use of AI generated video.

---

Chains of Models

OpenAI announced CriticGPT, a specially trained GPT-4 model designed to catch errors in ChatGPT's code output. The remarkable bit (to me) is how they're integrating it into their RLHF (human feedback) pipeline and thus improve the output quality substantially.

Long term implications are probably chains of models, where your completion actually goes through several layers of verification before returning.

When people use CriticGPT, the AI augments their skills, resulting in more comprehensive critiques than when people work alone, and fewer hallucinated bugs than when the model works alone.

---

---

Exponentials are hard!

---

Leopold Aschenbrenner’s recent essay and interview have been featured in the group a couple of times, here is a great primer about some of the implications (X thread).

---

Nik McFly

---

---

Becoming Superhuman (Workshop in Amsterdam)

My dear friend and THNK colleague Moulsari Jain is organizing a salon and workshop around super-humanity in an age of AI. The session is designed to help you transcend identity in order to better prepare yourself for possible futures. Join her and fellow likeminded people next week. Find out more and sign up on Eventbrite.

What can humanity do better than any machine? And how do we get there?Experience an often overlooked and instantly accessible method of tapping into our (super)humanity in a way that can never be replaced by any technology.

---

Envisioning Merch

We recently made stickers and other goodies for newsletter readers. If you’d like some in your mailbox, please share your postal address using this form.

• 🏅 Forwarding this issue to colleagues and friends.

• 🎯 Commenting with your favorite talks and thinkers.

You are receiving this newsletter because you signed up on envisioning.io or Substack.

---