Starting a traditional blog in 2023 might seem a little quaint. Modern alternatives, from social media to Substack, offer attractive networks of like-minded peers. Yet they also have significant drawbacks highlighted by the recent turmoil with Twitter. For one, there is no guarantee that anything shared on these third-party platforms will remain publicly available, searchable, or even properly archived in the long term. Twitter-style microblogging is also fundamentally ill-suited for careful writing and detailed discussion. Don’t get me wrong – I’ve had plenty of genuinely thought-provoking exchanges in my academic corner of social media, and I’ve discovered a lot of interesting research this way. But Twitter threads don’t really scratch the itch for long-form communication and can invite less-than-ideal engagement dynamics. A (macro) blog like this one is a nice halfway point between pithy 140-character pronouncements and dense academic papers. Public essays on large media platforms can sometimes fulfill that role too, but (thankfully) not everything needs to be an op-ed. Good old-fashioned blogging strikes the right balance between seriousness and frivolity, public and personal writing. There’s something nice about sharing one’s thoughts away from the sound and fury of crowded social media platforms.
Once I decided to start a blog, I considered various options. I’m a big fan of open-source software and cautiously optimistic about the Fediverse. While still in the early stages, it’s shaping up to become a healthy ecosystem that could offer a genuine alternative to traditional social media. Mastodon has already welcomed many Twitter refugees, and Lemmy is picking up steam in the wake of Reddit’s corporate blunders. Did you know that the Fediverse also includes blogging platforms? One notable benefit of federated blogging is that anyone can easily follow, share, and respond to your posts on any other platform in the Fediverse (imagine subscribing to a Substack feed directly on Twitter, for example). I briefly considered hosting my own instance of WriteFreely, one of the more attractive federated blogging platforms, but eventually decided against it. Self-hosting is messy, and I didn’t want my blog to suddenly vanish because something had gone terribly wrong with my server. I also wanted to have a simple system that just lets me write and forget about the rest. In the end, I settled for a no-frills approach that integrates version control. If you’re interested in the gory details: I write posts in markdown, push them to a GitHub repository, from where they get automatically published on my own domain with a static site generator (Hugo in my case) and free hosting service (Netlify). So far so good.
I’ve committed to a minimalist design without the bells and whistles of federation, on-page comments, and other social network integrations. But a blog needs an audience, and a standalone website like this one is less likely to reach it. Enters the POSSE system: Publish (on your) Own Site, Syndicate Elsewhere. The basic idea is very simple: publish your content on your own website first, then share or distribute it on third-party platforms, allowing viewers to interact directly with the original content. The goal is to maintain control and ownership of one’s content while encouraging reach and engagement through syndication and linkages to the original post. One major benefit of the POSSE system is that it doesn’t require committing to a single platform. I can share posts on my Mastodon account without having a federated blog, but I can do that on Twitter and Bluesky too. I’ve also created a Substack account that will mirror this blog 1:1. If people prefer to read it over there for convenience, so be it; but at least I know it’s all on here regardless of what happens to these other platforms. The POSSE system gives writers the best of both worlds: the independence of a personal website and the network of third-party platforms.
Enough with the dull technical details. What is this blog about? My unoriginal goal is to share thoughts about my research interests with an academic and curious general audience. As such, I expect it to focus mostly on the philosophy of artificial intelligence and cognitive science, perhaps with other topics sprinkled here and there. I’m interested in treating this both as a platform for public engagement and as a test bed for new ideas. The very process of writing a blog post about a research question can be useful to think about it more deeply without committing to a paper-length treatment. Regardless of who reads this blog, I suspect I will gain from this exercise. Of course, I hope some posts do spur productive engagement with potential readers, but my priority is to write without worrying about that too much. I’m not planning to keep a strict writing schedule, but I’d also like to hold myself to a certain degree of nonchalance in the process: this is not the place for obsessive revisions and endless polishing. I’d rather write quickly to get my thoughts out there and amend them in discussion or future posts if need be. That certainly doesn’t mean that I’m willing to share random half-baked ideas, but I will strive to find a sweet spot between careful editing and spontaneity.
What about the name? I simply thought I’d come up with something less boring than “Raphaël Millière’s Blog.” Given the focus of my research, I jumped on the opportunity to publicize a phrase I’ve grown to rather like – artificial cognition. Artificial intelligence is on everyone’s lips these days, but the term has always been a little controversial. There is no consensus on what intelligence is, let alone whether it’s a coherent construct. Due to the AI effect, people endlessly debate whether any given breakthrough in computer science is worthy of being called “AI.” Artificial cognition has a slightly different flavor that sidesteps some of these controversies. To be sure, people may debate whether any algorithm that exists today is really capable of cognition in a strong sense. But there is a looser and more deflationary sense in which many people are quite happy to ascribe broadly cognitive capacities, such as certain forms of reasoning, to large language models like GPT-4. Importantly, this does not require endorsing the claim that such models have human-level intelligence or various other psychological capacities, including intentions or consciousness. Artificial cognition is also an interesting phrase if we take things from the other end: as a research field, artificial intelligence used to be part of the cognitive sciences. From the very beginning, AI researchers sought not only to build smart machines but also to develop computational models of human cognition. As a researcher, I’m interested in the relationship between AI and human cognition in both directions: not only whether AI models can be attributed human-like cognitive capacities, but also whether these models can be used to gain a better understanding of human cognition itself. The name artificial cognition seems fitting to reflect both research programs. Finally, there’s something a bit artificial about sharing one’s thoughts publicly on a blog; I suppose the name works on that level too.