Street Smart
Polymathematics is a blog and newsletter about trying to do many things well. The three things I am trying to do well are writing, music, and invention. Creative people in tech or practical people in creative fields tend to enjoy it most.
If you enjoy this post, I’d love if you subscribed or forwarded it to a friend you think might like it. Thanks!
Communal Computing
What cool things could happen if every city had an app? Every block an API?
We have been overwhelmed with tools for the Smart Home, why are there so few tools for the Smart Street?
Don’t think of the banal government city websites.
Think of open source, think of Myspace, geocaching, little free libraries, adopt a Highway, concierges and tour guides, walking tours, graffiti, listening to the history of art pieces with a walkman in a museum, classified ads in local papers, wedding announcements in there too, telephone booths and telephone poles, stapled band posters and “take one” flyers advertising dog walkers or babysitters, think of Michelin guides, Pokémon go, and only playing music by local artists in an airport…
Why is the magic of AI currently only accessible at a desk?
There are people out there experimenting, trying to get AI out of the office and into the sunlight. Many of these visions involve a companion, a friend for you the individual. The AI is serving you and sort of belongs to you. Think of Humane, Rabbit, Friend (previously tab), and others.
But I’ve been wondering what it means to use AI as a companion, friend, or honest broker for an entire community. For a city, a school, a company, or a neighborhood.
The closest glimpse to this I’ve seen in mainstream tech circles has been to train a large language model on an organization’s IP, their documentation and codebase for example, and then let anyone in the company chat with the AI for company-specific help. Notion has built this directly into their software and many others have trained RAG based AI models on their docs in a more manual way.
That’s cool! But, what would this look like for different groups outside of corporations? Has anyone done the same for an open source project? As a way to supplement the documentation and the read me-s, this seems very valuable (I vaguely recall someone writing a chrome extension that did this for github repos). Imagine a future where every college class has its own RAG trained AI, like an artificial TA… a TAI. Imagine if every student granted it access to their personal notes. Imagine, after a decade of students passing through it, how powerful the AI could be as an aid for learning. Everyone would get better, if they wanted to.
Has anyone made a LLM for their family? Trained on their unique ancestry, family group chats, medical history, schedules and hobbies, day to day routines and collective interests? Imagine passing a gift like that onto your children or grandchildren, rich family history preserved by the family “historian”.
In Science Fiction, AI is often a universal asset, accessible to anyone. Think of the AI in Asimov’s famous short story The Last Question, or for a more recent example, the robot Clovis from Robin Sloan’s latest novel Moonbound. Clovis is cool because there are countless physical versions of Clovis all over, but talking to any one Clovis means accessing the knowledge of any other. Very cool stuff right?
These ideas of computers benefitting an entire community is related to what Spencer Chang and others have called Communal Computing, as opposed to Personal Computing. Isn’t it strange that you’ve rarely interacted with your computer with someone else? Sure, there are virtual multiplayer use cases for the internet (in fact that is sort of its whole thing), but physically collaborating on a computer like one might share a kitchen while making a meal with someone is nearly unnatural. The work Dynamicland is pursuing is a cool example of what this can look like1.
Beyond just AI, how can we use computers (and technology more broadly) to serve more than just individuals in their silos? Personal Computers are beautiful and will never go away, but isn’t it odd we haven’t experimented more with computing as groups? There is a fascinating history here about Englebart, Vennevar Bush, and the choice to pursue the personal path in computing. Matt Webb (more on Matt soon) had a fantastic post that touched on this, but alas I cannot dig it up.
Leading By Example (How To Make Your AI Cool)
Below are some examples of very cool stuff I am excited by right now, which is rethinking how to use AI outside of the “intelligent companion for individuals” mindset. These aren’t all necessarily communal (though most have a communal dimension to them), but they are a bit more clever than the median endeavor in the AI Gold Rush.
The Poetry Camera looks like a polaroid camera. You snap your photo of anything, but instead of an image, a poem composed by a large language model prints out on receipt paper.
The Poem/1 is a simple clock with an eInk display, which uses Large Language Models to compose poems that tell you the time. One new poem every minute. I am a backer of this one and interviewed Matt who created it for my upcoming podcast (stay tuned)!
The Grok AI Toy from Grimes et al is an especially cute example. It is a stuffed animal for kids which can never really run out of things to talk about. How refreshing to use Large Language Models for such a differentiated concept!
Flower Computer Company is another project I recently discovered (hat tip to Jack), which is aiming to turn physical objects into computers with their project ‘Dia’. They are a bit mysterious about their plans, but I think of it like this:
You open the camera on your iPhone and point it at your coffee cup. You add a website with the best coffee recs in every US city. When someone else points their camera at their coffee up, the website you added pulls up for them.
Pedestrian Computers
These sorts of ideas have been swirling in my head for a while now. I’ve had various ideas of how to experiment with communal experiences and AI too. One cool one still kicking around was to make a website which uses AI to “cross-pollinate” all the ideas uploaded by individual users. But I haven’t gotten viscerally excited about any of the early ideas yet. Until this past weekend.
I was talking with Tessa about something routine like which movie we planned to see later that night, and the idea suddenly landed fully formed in my mind. I texted it to myself and we went to see our movie (DiDi, which is worth seeing).
I envisioned beautiful ceramic tiles on the sides of buildings which when tapped with a phone would let you interact with the block or neighborhood. You could learn about the history of the area, get local personalized recommendations (“what’s the best vegan spot around here?”) and use services like Uber or get the ETA for the next closest bus.
I couldn’t stop wondering about what you could do if you explored this concept further. Every neighborhood would essentially have a website, with info like the current representatives for the district, the demographics of the residents, and local community organizations active there. You could see the closest Public Library or park, view the local job postings and events happening nearby, stumble on a coupon for the cool cocktail bar right down the street.
I thought about the flaws and potential criticism too. After all, almost all of this could be done on an iPhone or with an app. But I am a firm believe in being an idea expander and protecting your early fragile ideas and have learned to trust the raw excitement an idea brings out in me. The more I thought about it, the more paths for exploration I could foresee. The more I felt intuitively like the physical component, the tactile interaction, mattered. I tweeted about all of this and other people seemed excited too.
The very next morning, when grabbing a coffee, we stumbled upon this stray ceramic tile. It was exciting to envision what might happen if I tapped my phone to it. It helped me imagine what that experience would feel like.
I began programming some early ideas.
I hooked the prototype up with the Google Places API to get local recommendations that are open and near the tile. Local history is given with a combination of LLMs and curated json files for keeping the AI from hallucinating. I also hooked it up to Uber to automatically pull up the Uber app with the tile’s cross streets as the pickup spot.
Then I started building the bulletin board.
The bulletin board is basically a website for the neighborhood where a given tile is located. Anyone can tap the tile and view the local bulletin board, where people can leave job postings, photos, URLs, or even a simple note. Anyone can contribute to the bulletin board. I imagine much cooler ideas will begin to sprout up around this feature too.
The idea of a communal blog or website which grants edit rights to anyone who physically interacts with it seems limitless, experimental, and worth tinkering with.
I plan to put the first physical tile out in Austin sometime within the next week once the remaining bugs have been ironed out of the proof of concept. It will be interesting to see if anyone approaches the tile and interacts with it, contributes to it even.
This is an experiment not an endeavor, so I have no expectations for this project other than exploration. Yet, I can’t help but think about what cool things might happen when an entire neighborhood has its own digital twin.
If you have an idea for the tile, no matter how strange or “useless”, please leave a comment! And if you’re interesting in helping in a larger way, you can email me or DM me. It would be particularly helpful if you are a ceramicist who can help make the first ceramic tile (otherwise I’ll need to retrofit something together) or a programmer who is excited about the ideas here and wants to contribute (I’ve been considering open sourcing this idea too). Lastly, if you’ve seen another cool example of AI being used in clever ways I’d love to hear about it.
Thanks for reading!
Jake
Side thought: why haven’t we seen more successful board games which make use of simple computers, projectors, or higher interactivity?
So interesting! I would’ve loved these when I first moved to Dallas. I can see a future where these become essential in every neighborhood, city, and college campus.