Weeknotes 297 - AI emotional lock-in strategies

Hi, y’all! The EU championship is over. To be honest, I am not a big follower, but I know who has won. I noticed the turmoil of discussing the role of VAR in negatively influencing the game. I was thinking: Is this also part of our different relationship with reality? Do we trust analog thinking when we can access exact digital referees? Or are these kinds of applications the fuel for a backlash?

In the meantime, I received the latest gadget I ordered–a rare occasion in the last few years. Maybe I got tired of all the gadgets I got in the last decade, or the newness is not so distinctive anymore. Anyhow, Rabbit R1 has arrived, and I was able to test some of its features in between running activities…

Two short observations -apart from all earlier reviews on the complex interaction especially navigating through the menus and the underwhelming applications that are available. Not so much even apps, because they promised a different model of use: the LAM, action-based modeling.

On the positive side, I can imagine that the translator function can work in a social context with multiple people around the table. The push-to-talk drives a more structured conversation and less confusion for the AI than if it was just there lying in between a conversation.

Rather disappointing was the lack of contextual intelligence, especially in the spatial sense. I took a picture of a building in Amsterdam, and it referenced it to Rotterdam as it looks more like a typical high rise. That should be possible to do better. Even when I provided Rabbit with the situational context it still was not able to try to link the image to existing images of buildings, make that comparison, conclude what building and provide me with extra information.

I will keep trying; Maybe my prompts should be more specific. But with the right consciousness on anonymity.

Also, I would opt for a text-based interface, but hey, that is what the phone-based AIs promise, of course. Apple Intelligence is delayed, but I cannot imagine a better fulfillment of the functions than a phone for most of the use cases now.

Triggered thought

Listening to Poki podcast on AI (Dutch) where the Character.ai service was discussed. It has existed for a longer time already. Still, with the acceptance or adaptation of AI as a logical partner, we see that these new forms of artificial realities become real realities.

We not only use this kind of AI personality for reflection or fun but also start living together. And that living together is more than caring for a Tamagotchi. However, some of the same principles are at work; we like to dedicate human characteristics to things that feel and behave like humans. I am wondering if we would have relations with non-human-like but conversational-behaving things.

Before this happens, we will get used to more psychological conversational pals in our chats. This is of course already more than a decade-old idea, the service agents as conversational agents are as concept as old as the internet and even older. The execution and specifically the intuitive character and generative behavior, lowers the barrier and makes it more valuable. And we got more dependent on it, which was also addressed in the podcast in relation to the changes to the service of character.ai, when they toned down the intimate interactions. People find it much harder to let these types of tools go than if you lose a book in your digital library. So that gives a bigger lock-in effect and vulnerability for influencing.

It could become tricky when intelligent social companions feed conspiracy theories and even link people to opinions. Naming and shaming in the age of character bots is super dangerous for a polarized world. There might be a positive strategy, too; see the paper for this week below. Character.ai might need an “only fans” version…

For the subscribers or first-time readers (welcome!), thanks for joining! A short general intro: I am Iskander Smit, educated as an industrial design engineer, and have worked in digital technology all my life, with a particular interest in digital-physical interactions and a focus on human-tech intelligence co-performance. I like to (critically) explore the near future in the context of cities of things. And organising ThingsCon. I call Target_is_New my practice for making sense of unpredictable futures in human-AI partnerships. That is the lens I use to capture interesting news and share a paper every week.

Notions from the news

The New quantum record, the release of Apple’s public betas (without Apple Intelligence), these do not directly fit in the categories. KPIs for smart mobility policies? The rise of the AI ‘villains’. Or dominating intermediairs. Read below the usual new human-ai happenings, robotic performances and immersive connectness, and some more societal tech impact. Moments…

Human-AI partnerships

Humorous AI. Is the AI knowing how to time a joke, or is it in the eye of the beholder?

AI generates humorous installations from geo-localized keywords at london exhibition

fabio lattanzi antinori’s office party exhibition in london uses AI to question the production and consumption of data.

designboom | architecture & design magazineFabio Antinori

OpenAI is developing a scale to track progress towards human-like AI. The scale ranges from basic problem-solving to surpassing human abilities (as summarized by OpenAI).

Here’s how OpenAI will determine how powerful its AI systems are

OpenAI is grading its large language models on a new internal scale to determine when it reaches AGI.

The Verge

Not really AI but in order to build the right partnerships with ultrafast computing we need slow computing. Apparently

Why every quantum computer will need a powerful classical computer

Error-correcting a quantum computer can mean processing 100TB every second.

Ars TechnicaJohn Timmer

You would expect these tailored playlists were already common business…

YouTube Music tests AI playlists tailored to what you want to hear

In need of a new playlist?

The Verge

Reputation is becoming such an important asset for this company…

OpenAI is plagued by safety concerns

OpenAI claims to be developing artificial general intelligence, but growing safety concerns plague its progress.

The Verge

Imitation intelligence; on my watch list…

Imitation Intelligence, my keynote for PyCon US 2024

I gave an invited keynote at PyCon US 2024 in Pittsburgh this year. My goal was to say some interesting things about AI—specifically about Large Language Models—both to help catch …

Robotic performances

It would be a great disappointment if this was not the case.

Google says Gemini AI is making its robots smarter

And maybe identify and fetch your favorite soda.

The Verge

Not surprising maybe, the influence of partly supported is not making driving more dangerous. It might be perception too: is are you delegating control or adding a supportsystem?

Partial automated driving systems don’t make driving safer, study finds

Many driver assists do increase safety, but little evidence lane keeping is one.

Ars TechnicaJonathan M. Gitlin

Is this the case? “Robotics is having its ChatGPT moment. New advances in AI allow robots to ‘understand’ and react to actions in real-time, which means they can take on more complex jobs”

Standard Bots raises $63M to bring cobot arms to market

Standard Bots says it will use the funding to expand its engineering team and accelerate the development of its robots in new sizes.

The Robot ReportThe Robot Report Staff

Immersive connectedness

New technologies trigger new supporting services…

How to move a smart home

We offer some tips on moving from one smart home to another.

The Verge

Arduino still going strong with new approaches. Somewhere in between Grove, Arduino One, and Little things.

Arduino’s Plug and Make Kit lets your hacking imagination run wild, sans solder

Daisy-chain tiny boards into weather stations, game controllers, and way more.

Ars TechnicaKevin Purdy

Tech societies

The future of recruitment; replacing a workforce.

Intuit’s AI gamble: Mass layoff of 1,800 paired with hiring spree

Intuit CEO: “Companies that aren’t prepared to take advantage of [AI] will fall behind.”

Ars TechnicaBenj Edwards

The battle of AI search engines for supremacy.

Perplexity, Google, and the battle for AI search supremacy

AIs that generate answers to user queries could transform search, but only if someone can get the tech and the business model right.

Freethink MediaKristin Houser

Training of AI becomes the new “You are the product” for Meta and probably also others. What do you think; is public content open for all uses? Or should it be part of the terms first?

Meta is training its AI with public Instagram posts. Artists in Latin America can’t opt out

Latin America lacks robust data protection laws that would allow Meta users in the region to prohibit the company from using their content.

Rest of WorldMunira Mutaher

Is slop the right term for AI content that is result of overproduction? Is it the same as Ewaste, or is that more the offset of the production of AI content. Maybe something for another dive sometime. What is said about this in the just published EU AI Act?

Floating down the Love Canal

I’m not sure that there is much of a need for “slop” as a special term for AI-generated content. We already have “content.” Is there material we are content to call “content” that isn’t “slop” made to fill bottomless feeds? It’s not like the problem with AI content is just that it’s not “good” and it would right to embrace it if it begins to make content that people really like.

Internal exileRob Horning

Are we doomed? An Oppenheimer moment is something different than a ChatGPT moment…

AI’s ‘Oppenheimer moment’: autonomous weapons enter the battlefield

The military use of AI-enabled weapons is growing, and the industry that provides them is booming

The GuardianNick Robins-Early

Paper for the week

A very relevant and timely paper on the impact of conspiracy theories and how to counter these.

Durably reducing conspiracy beliefs through dialogues with AI

Here, we raise the possibility that previous attempts to correct conspiracy beliefs have been unsuccessful merely because they failed to deliver counterevidence that was sufficiently compelling and tailored to each believer’s specific conspiracy theory (which vary dramatically from believer to believer). To evaluate this possibility, we leverage recent developments in generative artificial intelligence (AI) to deliver well-argued, person-specific debunks to a total of N = 2,190 conspiracy theory believers.

Costello, T. H., Pennycook, G., & Rand, D. G. (2024, April 3). Durably reducing conspiracy beliefs through dialogues with AI. https://doi.org/10.31234/osf.io/xcwdn

Looking forward

Last week I had nice conversations with old and new contacts, readers of this newsletter :) And we discussed the coming ThingsCon events with different partners, both the Salons in September and the celebration edition in December. The website is updated, and will have some more updates this week.

Also curious to experience a ‘hard-core’ academic conference on SST – EASST-4s – to be precise; and be part of a panel on the kit-economy with our experiences with the WijkbotKit as co-design tool for future cities. And I will try to attend a couple of other sessions, specifically the one on agent agency.

And furthermore, there is a UX crawl at IKEA that is, as to be expected very popular. I am still on the waiting list 🙂

Enjoy your week!