Sign in Subscribe

Weeknotes 263: flocks of packet-switching intelligence

By Iskander Smit —

Follow-up of GPT store, surface Humane AI, and more notions from last week’s news in human-tech relations, with events and paper for the week.

Weeknotes 263: flocks of packet-switching intelligence
an army of small autonomous moving transparent flying boxes filled with wires, populating the streets, according Midjourney

Hi, y’all!

Every week there is a (tech) news event that defines the debates. So was Humane last week. The AI in your pocket… Interesting and cringe at the same time. Some thoughts I shared earlier via Mastodon/BlueSky

Nice to read the impressions from the Humane demo by @caseynewton

How he is appreciating (again) the clear form factor of the phone. The quest for the position of wearable AI.

Will it be just another touchpoint to the intelligent cloud like the watch? Or is there a new space?

I am thinking the voice interaction and hand palm tricks are blocking the view of the real new (potential) interaction principle: creating an AI touchpoint of every surface around you.

Platformer

But the real AI debate was driven by the aftermath of OpenAI dev event, I mentioned shortly last week, as it was held just the evening before the newsletter dropped. The analysis about the event all agreed on the impact and also the positive format of the event, which brought back some of the feelings of the early Steve Jobs. Sam Altman has a charm, a very different one from Jobs, but especially the move away from a heavenly-produced product introduction was refreshing. More important was what was introduced.

There is an interesting connection, as mentioned in The Verge Podcast: it was strange there was not the slightest reference to Humane AI, while Sam Altman is one of the investors. Of course, he would not distract from the main messages, but on a side note or at least mention the physical AI as a development.

Triggered thoughts

Listening to Pivot speaking at the OpenAI keynote, not sure what triggered it, but the first wave of AI apps is the new search of chunks of data produced by people, creating knowledge, new versions, etc. All podcasts as transcripts, writers, and researchers.

Doing synthesis based on content databases instead of ‘just’ search. I can remember that there was an attempt years ago by Google to mix the search with a knowledge graph of trusted sources (a quick look in old articles); it never flew. Around that same time, we explored a platform for knowledge workers.

Will this happen now in a different format? A GPT for all my collected links and documents will not only deliver a richer knowledge base if I can have a chat with the knowledge LLM (PLM), I am also curious about what new learnings will emerge and even what it will tell me about my interests… As a start I begun uploading my docs in Notions; I cannot imagine they will not create that function seamlessly. And in the meantime, I will dive into the GPT builder.

Events

(And as update of the event I organise myself)

Notions from the news

As mentioned above, the announcements of OpenAI got a lot of attention. How to…

Opinions from Ethan Mollick, Kevin Roose, Dan Shipper, Ben Thompson

In the meantime, others are announcing too, like Samsung planning an alternative ChatGPT named Gauss. Elon Musk already did last week Groks

Co-pilots have a business case, apparently. For GitHub it is a moneymaker.

The developments finding faster or smarter chips continue. Next-Gen AI chips will mimic human brain. I am not convinced that is the way to go; as I noticed while watching one of the latest “Two Minute Paper” videos, ChatGPT is making the same mistakes as humans, which should be unnecessary having a different concept of perception. Mimicking humans is apparently more important than being a complementary power. What a time to be alive…

In the meantime, we got confused more and more

White faces generated by AI are more convincing than photos, finds survey
Photographs were seen as less realistic than computer images but there was no difference with pictures of people of colour
The GuardianNicola Davis

The chip is a crucial part in the AI competition… And a new arms race :-(

Google and Anthropic Are Gunning for AI Dominance with New Chips - Decrypt
Anthropic’s Claude LLM gains a competitive edge with Google’s TPU v5e chips as part of a collaboration that not only injects capital but also promises to push the boundaries of AI’s computational frontiers.
DecryptJose Antonio Lanz
The US Wants China to Start Talking About AI Weapons
As the US and China meet for the APEC summit in San Francisco this week, American officials are pushing for talks on the risks posed by military use of AI.
WIREDCondé Nast

The How-To’s get popular, it proves the popularity

How junior designers can use ChatGPT without risking everyone’s trust
How to make use of the productivity boost AI offers without running into issues
UX CollectiveKai Wong

Will DX be more important that UX?

DX is the new UX
The massive shift from developer-extended to developer-led platforms
Platforms, AI, and the Economics of BigTechSangeet Paul Choudary

Speaking of changing trends, IDEO is cutting staff which can be linked to the position of Design Thinking: we could be very well **beyond peak Design Thinking** acc Marco van Hout**.** And another reflection by Tobias Revell, and by Louisa Heinrich.

AI might not take over the world, but it might cause a financial crisis, Harari thinks. And Noah Smith on how liberalism is losing the information war.

Did I mention the potential emergence of sweatshops for knowledge to feed the AI? Or do we like to call it partnerships? And we need new evaluation methods

AI Companies Are Running Out of Training Data
Data is the vital force of large AI models, and thus of the industry itself. But it’s also a finite resource — and companies could run out.
FuturismMaggie Harrison
OpenAI Data Partnerships
Working together to create open-source and private datasets for AI training.
The Illusion of Understanding: MIT Unmasks the Myth of AI’s Formal Specifications
Some researchers see formal specifications as a way for autonomous systems to “explain themselves” to humans. But a new study finds that we aren’t understanding. As autonomous systems and artificial intelligence become increasingly common in daily life, new methods are emerging to help humans che
SciTechDailyMike O’Neill

New AI copilots from Instagram, Figjam, GitHub, Waze

Personal Digital Twins were a thing already, but are now more and more the intelligence reference, especially for personal health treatment.

How digital twins may enable personalised health treatment
Research is growing into computational models that will move medicine beyond what works on the average patient
The GuardianLinda Geddes

Robots are getting more attention and benefiting from the Gen AI wave (”the GPT moment is near”), which is clear every week again. How robots are expanding their fields of work by attaining emotional designs.

How Robots Are Expanding Their Fields of Work By Attaining Emotional Designs - Yanko Design
Robots have long captured our imaginations, from the helpful household robots of science fiction to the industrial machines that streamline manufacturing processes. However, the integration of robots into our daily lives has often been hampered by the challenges they face when navigating complex hum…
Yanko DesignTanvi Joshi

And the dangers of living with robots are clear, too.

Robot Kills Man After Mistaking Him for Box of Vegetables
The man was forced onto a conveyor belt at a pepper sorting plant in South Korea, according to reports.
The Daily BeastAllison Quinn

Self-driving cars are the perpetual promise… Also, here, we need humans to help the machines.

Self-driving cars still aren’t the future
A Cruise disaster deflates the robotaxi hype once again
DisconnectParis Marx

Making spatial video with the new iPhone is working well, John Gruber was not disappointed at all, and also in the impact of the experience in a second use, he mentioned in the Dithering podcast. Interesting he described how traditional panorama pictures become 3D experiences. Good to know, as I like to capture moments in panos.

First Impressions: iPhone 15 Pro Spatial Videos on Vision Pro
Starting with iOS 17.2, which is currently in beta and expected to be released in December, the iPhone 15 Pro and iPhone 15 Pro Max can record…
MacRumorsJoe Rossignol

For a reflective moment:

A 4.5 Billion Year Video Timeline of Earth in 60 Minutes
To mark the 10th anniversary of their YouTube channel, Kurzgesagt has released a video timeline of the Earth’s evolution,
kottke.orgJason Kottke

Is Singapore the poster child for the solarpunk city? Something else than a sponge city.

Singapore urbanism
Thoughts from a visit to the solarpunk city.
NoahpinionNoah Smith

Paper for the week

In case I dive further in LLMs, this is an insightful paper I expect: A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions

In this survey, we aim to provide a thorough and in-depth overview of recent advances in the field of LLM hallucinations. We begin with an innovative taxonomy of LLM hallucinations, then delve into the factors contributing to hallucinations. Subsequently, we present a comprehensive overview of hallucination detection methods and benchmarks. Additionally, representative approaches designed to mitigate hallucinations are introduced accordingly. Finally, we analyze the challenges that highlight the current limitations and formulate open questions, aiming to delineate pathways for future research on hallucinations in LLMs.

Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., ... & Liu, T. (2023). A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions. arXiv preprint arXiv:2311.05232.

Link

See you next week!

I will be up to Wijkbot meetings and student coaching sessions, ThingsCon preparations and most of the time writing on a new policy agenda for creative industries… And depending time- and weather-conditions, we might check out Glow, a yearly tradition…

Buy Me a Coffee at ko-fi.com