Weeknotes 255 - Acceptance through convenience. Or agency
Another week with more AI enhancements, robots, new things and more.
Hi y’all, that was a great week, with the ThingsCon Salon as the pleasant anchor point: good crowd and lovely talks and workshop. As always, I needed to share my attention between organisational focus and listening in to the talks. Douwe Schmidt shared how hands-on the municipality of Amsterdam is working on the Al leaflet, which starts with a consciousness with all involved civil servants (Raw Data is an Oxymoron). Anna Noyons took us on a personal journey towards the beliefs of her social design studio; ‘Design is not only about the direct, individual effects but also about the long-term collective effects’. Tessa and Mike showed how the Human Values research program combined theoretical frameworks and experiments. There will be a full report that I will share here, too.
This evening, Dutch local time, the next Apple event will take place to introduce the latest iPhones. Will it be underwhelming, as we have heard all the news already upfront, or can we expect a surprise? Both? As often in the last years (maybe always), the real interesting parts are between the lines. What will the changes, how small even, will say about the choices for the platform? Will we see a boost in the conversational AI will be announced?
Triggered thoughts
Starship Technologies self-driving delivery vehicles are all over the place in Tallinn, their home ground. Michiel Berger, who was visiting the town, shared the apparently common attitude with people on the streets there, hardly noticing them. The self-driving taxis in San Francisco have a different status and are more of an attraction than a common thing. In the podcast “Have a nice future” Lauren Good shares her first experiences. As an introduction, how the self-driving trucks might be there sooner. Aurora is about efficiency capitalism, as the other host, Gideon Lichfield, framed it nicely. If indeed we have an end-of-the-decade capable system to speed with 130 on the Autobahn, as was suggested at IAA in Munich, that might push back on speeders.
So, what are my triggered thoughts…? The title is telling: Learning to let go. But we need trust on multiple levels, and trust is not in the quality of driving but more in the agency we have. On the way, we are treated as we drive the car and what we share while using the services.
Events to check
- 13 September - General Seminar - online - The C-Corp is dead
- 14 September - The Future of AI in the Cloud - Mediamatic, Amsterdam
- 14-17 September - GLUE - a four-day design route in Amsterdam
- 15 September, an MIT lunch lecture (dinner for us) on computational typography, organised by Vera van de Seyp
- 16,17 September, Groot ateliers weekend, Rotterdam
- 19 September prompts, prototypes and pulp - De Zwijger Amsterdam
- 21-22 September: Conference from The Things Network (get a discount with TTC2023)
Notions from the news
Tech is geopolitical. The decision of the Chinese government to ban Apple had a big impact. And in Europe, the Digital Markets Act designates and empowers big tech as official gatekeepers.
And some weird conspiracy theories about wildfires linked to smart cities…
What about AI?
A weekly roundup of new AI assistants and enhancements of existing and new tools. Like an iPhone app for the LLM knowledge search Perplexity. IBM is rolling out generative AI features, Zoom is already rebranding its gen AI features, and the US tax authority is creating AI enforcement for wealthy tax violators, Roblox offering help for building virtual worlds can feel strange, killing the core creativity in the game, but if it is only a pal than it makes sense.
Pinterest has been a silent powerhouse in intelligence features for some time. Now it is introducing new algos that can alter clothes to your body type. Loom is adding the usual enhancements.
Ironclad is known as a party for smart contracts, but they are now introducing an open-source visual AO programming environment. https://www.producthunt.com/posts/rivet-by-ironclad
IKEA upgrades customer service capabilities into interior design advisors with - you guessed it right - AI.
And AI as part of society. Or entertainment. Tim Burton does not like the imitations of his animation style.
I think I agree here with Ethan Mollick (and could that make a Tim Burton character even more weird? “the real value of AI comes not from having it emulate old ways of solving problems, but, instead, by helping us unlock new capabilities”
Link this to this psychology perspective on how we balance our limitations with AI. Potentially.
Weather forecasting might be a compelling case here:
https://phys.org/news/2023-09-ai-weather-showcase-data-driven.html
Also, in building trust as we have a well-developed feedback loop…
Speaking of trust, what technology even Facebook and Google did not dare to release was all about breaking trust.
https://www.nytimes.com/2023/09/09/technology/google-facebook-facial-recognition.html
Typical tech bro talk? Anti-regulation, optimistic and capital-driven. That went well last time…
Who is training the AI?
https://www.wired.com/story/prisoners-training-ai-finland/
And in the meantime, I like to see these AI designs in real tangible furniture. Can it keep the dynamics of the animation intact?
Is ChatGPT in a summer slowdown, or is it a structural development?
And on the reading list. Or better catch up on what I missed at the weekend?
https://www.wired.com/story/sundar-pichai-google-ai-microsoft-openai/
https://www.noahpinion.blog/p/at-least-five-interesting-things-6bf
http://www.downes.ca/post/75564/rd
https://www.wired.com/story/what-openai-really-wants/
And how about robots?
Funghi-based skin for ‘humanoids’ sounds a bit creepy but can deliver a whole new category of cyborg… And it makes sense to have a worm-like creature that can operate in places that are hard to enter.
https://futurism.com/the-byte/living-skin-robot-fungus
https://www.popsci.com/technology/ge-aerospace-sensiworm-robot/
You might have been wondering if Clubhouse was still around. Apparently, but now I'm pivoting towards messaging instead of chat-focused.
And the same for flying packages, aka drone delivery. There is a potential breakthrough in clearing some regulations in the US, that is at least. So much is clear again about how defining the rules are.
https://www.theverge.com/2023/9/6/23861764/faa-ups-delivery-drones-amazon-prime-air
And more (or misc)
What do doomers think?
https://www.okdoomer.io/nobodys-driving/
Sweet, design your house for your cats.
https://read.readwise.io/new/read/01h9ykv1y1pd5j7mgyb9dhrr5b
Paper for this week
Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback In this paper, they (1) survey open problems and fundamental limitations of RLHF and related methods; (2) overview techniques to understand, improve, and complement RLHF in practice; and (3) propose auditing and disclosure standards to improve societal oversight of RLHF systems.
Casper, S., Davies, X., Shi, C., Gilbert, T. K., Scheurer, J., Rando, J., ... & Hadfield-Menell, D. (2023). Open problems and fundamental limitations of reinforcement learning from human feedback. arXiv preprint arXiv:2307.15217.
See you'll next week*
For me, this is a shorter week, and * please expect a different newsletter next week as I will be travelling for my holidays. And after the holidays, I will share more on the changing work perspectives I have been preparing… Enjoy the coming weeks!
