Breaking the blinding boxes outside in for AGI

By Iskander Smit in weeknotes — Feb 11, 2025

Weeknotes 326 - Can we trust AI to be our AGI assistent as it can fail on unknown knowns? Reflections, and reads from last week.

Unknown unknowns of deep research, interpreted by Midjourney

Weeknotes 326

Thanks for landing here and reading my weekly newsletter. If you are new here, have a more extended bio on targetisnew.com. This newsletter is my personal weekly reflection on the news of the past week, with a lens of understanding the unpredictable futures of human-ai co-performances in a context of full immersive connectedness and the impact on society, organizations, and design. Don’t hesitate to reach out if you want to know more or more specifically.

Voor diegene die liever Nederlands lezen, sinds dit jaar maak ik ook een Nederlandse versie, vooralsnog via deze Substack. Verschijnt soms wat later op de dag.,

Hi all!

Last week I was unhappy with the first version of the triggered thought part I sent out. So, I edited it a bit and updated the version on the website. So that you know.

What did happen last week?

I try to keep this short. The Wijkbot/Robo-perspective might become a tool for journalists to discuss citizen interests. We announced the plans for the ThingsCon RIOT publication 2025. If you are interested in the movies I saw at IFFR (and what I found), check my Letterboxd diary.

What did I notice last week?

Another busy week in tech & society, and human-AI partnerships. Find the complete list with links below, but a summary here:

Just in; Elon wants to buy OpenAI. Powerplay, settling old accounts, not make but buy? OpenAI is partnering with TSMC to build custom AI chip. Sam Altman does three observations on the near future. → → →
Deep Research was announced last week, and this week, we had all the tech media writing about their first experiences. I did not buy a 200 eu Pro account (yet), but the reviews are quite good. I can very well imagine it will contribute to the work process of a research-oriented role. Now, as a great assistant who helps to shift perspectives and find forgotten sources, As a mirror of your own work. That will change, of course, and the collaboration with these types of tools (the others will follow soon) will define the future of knowledge work focus. Check also Triggered Thoughts below, Ben Thompson had an inspiring review. → → →
Google was introducing similar but not the same thinking models. →
Who’s smart, who’s just guessing? Testing 1 2 →
DeepSeek’s impact is still discussed. Like, in light of the war on talent, are Chinese AI talent staying in China now? →
Research of the research; a comparison. →
Cruise is shifting focus from robo taxis to personal autonomous vehicles. I think the Tesla robotaxis are coming soon, too (in the Valley, of course); Zoox is entering the market → → →
The power of image generation, or better video generation. →
In the meantime, Google is leaving diversity commitments and opening up for weapons → →
Scott Galloway sees potential for a positive impact by AI for a society benefitting stakeholders more than shareholders, and open-source AI is breaking the dominance of major companies. Let’s hope with him. →
Also in robotics Physical Intelligence has open-sourced its Pi0 robotic foundation model. → →
Can Africa become a major player in AI giants? What are the economics of big tech in times of the AI arms race fallacy? → →
Are exoskeletons coming our way? The promise of learning devices makes it even more interesting → →

More references are below.

Triggered thought

As artificial intelligence continues to evolve, promising to unlock vast realms of knowledge, a critical limitation is emerging—one that challenges our expectations of AI's capabilities. This limitation isn't about the "unknown unknowns" that AI might eventually conquer but rather the "unknown knowns" that it might consistently miss. Ben Thompson, in a recent Dithering podcast and follow-up article, illuminates this problem:

The issue with the report I generated — and once again, I'm not going to share the results, but this time for reasons that are non-obvious — is that it completely missed a major entity in the industry in question. This particular entity is not a well-known brand, but is a major player in the supply chain. It is a significant enough entity that any report about the industry that did not include them is, if you want to be generous, incomplete.

It is, in fact, the fourth categorization that Rumsfeld didn't mention: "the unknown known." Anyone who read the report that Deep Research generated would be given the illusion of knowledge, but would not know what they think they know.

This revelation presents a paradox at the heart of AI's promise. While we often focus on AI's potential to uncover "unknown unknowns," Thompson's experience highlights a more insidious problem: AI's potential to overlook crucial "unknown knowns," creating a false sense of comprehensive knowledge.

This oversight isn't just a limitation of current AI systems—it challenges our fundamental assumptions about artificial general intelligence (AGI) and its role as a partner in expanding human knowledge.

This problem should be solved, not so much by more intelligence but by another design, by developing a new form of breaking-the-box-thinking.

What are the plans for the coming week?

This week, I am looking forward to catching up with some good people, and I will check out -if the agenda allows me- some results from innovationlabs program and visit the opening of the new Next Nature exhibition in Evoluon. The research project Human Values for Smarter Cities has a consortium update. There are also meetups I think, like one from Amsterdam UX on Wednesday. And more artistic in Rotterdam: The Holistic Technology Salon. Next week Sensemakers is doing a lecture edition.

See you next week!

References with the notions

Human-AI partnerships

Robotic performances

Tech societies