Large Action Models, Data Voids, and thinking differently about social media [newsletter]
Hey hey,
Lots of interesting stuff that's been collecting in my tabs and notes. From Large Action Models to Data Voids to some ideas on how to think differently about social media. So:
Onwards!
Large Action Models
A fun/interesting thing that came out of CES is this collab between Rabbit and Teenage Engineering. It's a handheld AI-powerered device to "use apps for you". It sure is a new paradigm that feels fresh, kinda? That said, I'm not sure it has any legs to stand on. After all, it seems like you might as well replicate that in an app rather than separate hardware?
First thoughts:
- Does this make sense? (Maybe?)
- Is this something that would work well in China because smartphone screens tend do be larger and so it has the USP of being smaller? Somehow I can't see many people in Europe or the US use something this specific on a regular basis. (Also, where would you even put it? Even as a guy - who wears pants that by default have pockets that can actually hold stuff! - wouldn't know where to put this. I wouldn't know where to put another device.)
- Is its biggest value-add that it looks cool? I'm with Dan Hon on this one: I think that the Teenage Engineering case is a huge part of the appeal. Give me some Teenage Engineering design any day, it's always a joy.
- The Large Action Model part, though, essentially emulating a human app user, is quite an interesting approach, even if it maybe feels a little hacky as a way to get around API limitations. I think of it as a kind of finger-simulator for apps, is this really what it is?
Data Voids
So here’s a fun new thing to consider when thinking about election integrity and online discourse: data voids, “information spaces that lack evidence, into which people searching to check the accuracy of controversial topics can easily fall.” If disinformation claims aren’t widely covered, folks will be served results that support their wrong views. According to this study, online search might serve as a reinforcer of disinformation.
Which brings me back to an old pet peeve of mine: I truly believe we still under-appreciate the problematic role of paywalls. Disinformation and conspiracy theories are free, real news and research is often invisible behind paywalls. Disinformation has a big structural advantage because we don’t properly fund journalism.
Democracy under pressure
Speaking of journalism, there is an uproad in Germany after some dope (if very, very worrisome) investigative reporting by the public interest investigative journalism org Correctiv about a secret / high profile neo-nazi meeting. The whole thing is more than a little worrisome. This is like straight up out of the fascist playbook from 100 years ago.
A kind of footnote that caught my eye from a professional point of view, though, was the mention of this group of extreme right-winger planning to start an agency for right wing social media influencers. Which kinda made me wonder, is this not yet a thing? The radical parties have been doing better on social than established once for quite some time. I mean, that's structurally not a surprise: More radical positions are favored by engagement-optimized algorithms, and the established parties are fairly snorifying online.
Some ways to think about alternative approaches to social media
We know a bunch of the issues with the current crop of big social media platforms. But how else to approach the space is much less clear. Personally I don't believe in government-run social media platforms, but other than that the debate mostly focuses on either centralized vs decentralized or closed vs open. All of which are legitimate aspects, but I think we would benefit from a richer vocabulary to explore more ideas, and more-dimensional ideas. I'd like to suggest a few to try on:
- Public Ownership (gov owner, taxpayer paid)
- Public Governance (controlled by public through governing body)
- Public Interest (public interest mission, ownership secondary)
I think it can be tremendously helpful to expand our vocabulary and thinking about these things to explore new avenues. This classification might be useful to try? Let me know what you think.
Are social platforms a bit like fossil energy?
I wrote this up in a a bit longer over on the blog, but here's the TL&DR;
- I never subscribed to the idea that data is the new oil (i.e. it's valuable and a commodity) but maybe it's similar to fossil energy in that data driven systems (incl social media) have both amazing short term effects and less-understood (and possibly also just bad) long-term effects? Either way, both putting our money on fossil energy and on big data driven platforms marked a phase-change, a transition that left society in a new and different state.
- As an example, I think the heyday of social/local/mobile services in the early to mid-2010s like Foursquare enabled amazing kinds of discovery. But once the systems had ingested all the data into their systems, once they had mapped out the world sufficiently, we ended up with unintended consequences that we now have to deal with, including crappy recommendations, a flattened culture and also just much worse discovery than before.
Misc
- A podcast interview on I Never Metadata with Lisa Grimm on taxonomy emergencies (Spotify) caught my eye, uhm, ear. Mostly because frankly, taxonomy emergency sounds so nicely dramatic for a recommender algorithm gone wrong.
- The Chamber of Progress's Tech Policy outlook for 2024 (PDF) is an interesting read. It's a progressive tech industry coalition, so be aware of their very clear political bias.
Behind the scenes
I've been looking into turning Obsidian into my one-and-all organizational system. I'm not sure yet if I want Yet Another System but at least it's text-based and it would allow me to combine my note-taking and my task management more effectively in one place, which should help reduce friction. My biggest obstacle is that I find Obsidian's interface pretty visually busy/bloated to begin with, so that won't help... But my friend Alper has a write-up that might help. Again, curious what works for you!