š± Itās already September! Iām not entirely sure where the earlier part of the year went. A lot of things happened in the news, but many events also made for a busy period for me up until now. Spoiler: it doesnāt look like itāll be any calmer, either!
I wanted to get back to some of the thinking I have been doing about governance and the open web. I mentioned a project last week from a team that decided to look into Fediverse governance and offer a guide to setting up and operating Social Web services by looking at the technology, moderation, legal risks and other topics important to good governance. The document is long, and I still havenāt had time to read it all, so Iāll refrain from providing an AI summary, as I doubt it could capture some of the subtleties of such a document. I hope to have time this week, but it depends on other project deadlines.
Regardless, this project and others seem to be spurring a resurgence of interest in the open web, specifically the social web. Social networks, whatever your view of them, have galvanised the public and been the gateway drug to the internet for many. Unfortunately, they have been enough of a drug in their own right to allow the social network platforms to fence in their users, discouraging them from going elsewhere āand there are plenty of other places on the internet to go. This is often termed the Walled Gardenā¦ āEverything is nice and neat here, so you donāt need to look over the fence.ā This has been done through network effects. You tend to go to a place where the others are, more accurately, where the people you know or like reside, hence the āeffectā of the network. Leaving such a place is generally too difficult for most, which instantly changes the balance of power away from the users to the platforms. The platforms leveraged the real-life networks to create the walled garden digital networks and then lock everyone inside.
The open web, or social web, completely upends this, as you can congregate in a place that has all the people you know and love, and you can easily leave to go to a different place that, perhaps, aligns with your point of view or social affinities more closely, without losing access to your friends. The major platforms know this, and they are doing what they can to head this off at the pass, and they will embrace, extend and extinguish. Itās in their nature. Remember the tale of the frog and the scorpion?
With Twitter āor X if you preferā being shut downin Brazil, there has been a substantial increase in new users to other social platforms, most notably Bluesky. If you have used it, you will notice the resemblance to Twitter and its residentsā overall fun and snarky nature. Itās not for me, but I pop my head around the door now and again to see if anyone interesting is in there. One thing I think they are doing that is interesting and follows the theme of governance and management of platforms I have going on here is its latest announcement to give users more control over the use of their posts. One major criticism of Twitter was how quote-tweeting became a weaponised sport. Bluesky is trying to address this apparent shortcoming by allowing users to detach their original post from quote tweeters (is that the correct term for Bluesky? Checks notesā¦ apparently it is āpostsā) and hide replies. Itāll be interesting to see how this plays out and whether it dissuades poor behaviour, or dissuades using the feature entirely. It is a laudable effort, and I hope it is the former and not the latter, but if the internet has taught me anything, it brings out the worst in some people. Iām bookmarking this to revisit in the future.
Iām a fan of the open web, not because Iām an old-timer who used the Internet before HTTP was implemented āok, some of thatā but because it allows a much richer environment to try out new stuff and develop new ideas. You canāt have escaped the newsletter revolution that happened a few years ago, everyone and his dog (including me) has a newsletter, and many of them are hosted on platforms like Substack (Nazi Bar) or other popular ones like Ghost. Ghost gained popularity when Substack outed themselves as being OK with enabling Nazis to make money, and it is a good platform. Itās somewhat expensive if hosted and a little too technical for most to self-host. 2024 ushered in their support of Activity Pub (the federated social web protocol) and have recently announced the implementation of two-way communication through inbound likes to enable people to publish and readers to react through tools like (Iām guessing) Mastodon, ensuring those likes get back to the original post. This is interesting, but a development Iām a little sceptical about for reasons of incentives and eventual ad-tracking. I reserve judgement for the moment, and Iāll try to learn more about it and how it works.
Reading
Microsoft recently released a study into the use of generative AI in the āreal worldā. You can find it here. Perhaps if I have some free time, Iāll write up some of the notes I made on the report. In the meantime, hereās a couple of quick thoughts about it:
- One, the report never defines what āreal worldā actually means for the study and the studies referenced. So Iām not sure what value the moniker āreal worldā has given that my āreal worldā working environment is almost certainly very different to yours.
- The outcomes of the studies discussed in this paper are less than a stellar endorsement of generative AI in the workplace. Some examples:
ā¦ agents with the assistant resolved 14% more issues per hour than those without the assistant. Consistent with what has been observed in some lab studies (e.g., Noy and Zhang 2023), the largest impact was on novice and low-skilled workers, with very little effect on experienced or highly-skilled workers
Wiles and Horton (2024) explored how having an LLM generate a first draft of a job posting affected postings and hiring on a large online labor market. They found that the AI tool decreased time spent writing posts and increased the number of posts completed but had no effect on the number of hires.
Researchers found that on average, those with Copilot for Microsoft 365 read 11% fewer individual emails and spent 4% less time interacting with them, compared to people without Copilot
Moreover, to preserve privacy, the study observes activity, not the content created, so it cannot study quality or how well output aligns with peopleās goals or intents
On coding:
However, developers also voiced significant concerns. The top worry (29%) was that AI might not be as helpful as expected. Another major concern (21%) was that AI might introduce defects or vulnerabilities, emphasizing the need for thorough validation and human oversight. Job security was a worry for 10% of respondents, reflecting fears of AI encroaching on their roles.
In contrast, no substantial difference was observed between Copilot and non-Copilot groups for the less familiar task.
In other reading, I learned that Muskās troubles continue:
Brazilian court orders suspension of Elon Muskās X after it missed deadline - The Guardian
Iāve already mentioned an excellent regular read of news from The Continent. They continue to do outstanding work. The article about Bill Gatesā āfarming expertiseā is actually about Tech Bro Solutionism, and is a fantastic case study into why tech bros should be a little more humble and learn from others. The consequences of their actions are and could be devastating.
Direct link to the pdf.
Speaking of bros thinking they know more about stuff than anyone else, James Hoffmann has an exquisite takedown of Science Bro Andrew Hubermann, who exudes, āI know one science; therefore, I know ALL scienceā. What is he really offeringā¦ of course, protocols on supplements to make you a ābetterā humanā¦ Nothing at all to do with bettering his bank account. Silly me.
The last article I wanted to highlight is another egregious use of technology, which I have been discussing with clients and during training. The use of surveillance cameras for face recognition at the 2024 Olympics and now in general use across Paris. The installed cameras will operate until at least March 2025, around seven months after the games have finished, including the Paralympics. Two points merit discussion, one being obvious. Why will they be in operation months after the games, and what will happen once that time is up? The second interrogation is about their accuracy and the less than exemplary record of these systems already in use in places like London, UK, where false positives of non-whites clearly show the uselessness of these systems. This article gets somewhat into that discussion.
Written from the heart and with sweat and tears. Have a great week.
