Bryan, this is a sharp and timely analysis. Your point that AI search is reshaping rather than revolutionizing our relationship to information is particularly thought-provoking. The movement from pluralistic exploration to synthesized authority raises serious concerns for both education and democracy.
As an educator, I share your concern about what this means for critical thinking and source literacy. If AI becomes the default interface for inquiry, how will we teach students to question, compare, and discern? Can institutions procure instruments with these safeguards embedded in the interface?
Your forecast of competing authority bots is chilling. Perhaps the antidote lies in building human-centered platforms, newsletters, forums, and classrooms that cultivate dialogue and nuance. As I reminded my housemate just now, the best technology is still the one between our ears.
Another thought, maybe a sci fi book about competing authority bots?
Certainly the way we, Deep Wisdom House, are traversing, and happily so: “Publishers are playing defense, building their own destinations and weaning themselves off platforms that use their content as training data. The goal is to own the audience connection and no longer be vulnerable to algorithmic and other platform shifts.”
One of your more crucial observations: "As I’ve said for years, and has started to happen, different entities will want to offer their own takes on the world: nations, religions, companies, etc."
Potential long comment warning. There are some hairs to split when talking about "search" is one thing. Searching for specific facts or answers differs from searching when researching differers from what I most enjoy, just looking to explore. For the latter, I get more out of results beyond the best, things I would not even consider or just what pops up as curious. GenAI removes serendipty.
Also the concerns seem focused on impact of what people now rely on from search, to get more clicks on their stuff. You did touch on this in literacy but I see a dramatic decrease in understanding fundamental basics of what URLs do. I remember when Jon Udell spoke about the literacy of reading and fiddling with parameters in search results. Since modern web tools with visual editors remove us from the HTML soul of the web, I see people now knowing at all the link they are sharing is encrusted with UTM tracking tags.
Or today, on a web form, I see people submitting 50 mile long URLs generate by Outlook the stuff that starts with namo2.safelinks.protection.outlook.com. We have lost the wide understanding of how the web works.
And looking at the links most shared in socials as a reference are often reports or summaries from things written in online news sites that exist to attract clicks and spit ads, but also finding links share to research going to a university news post. We have lost the Mike Caulfield SIFT sensibility of going upstream to the sources. Each time I go upstream to find the best link, I invariably discover more of interest tangentially.
I relish doing my own analysis what to read, and in general feel like we are dumbing down society by replacing that mental exercise with the gray goo summarizing of AI. I glance at them, and I might grab and go if its something as mundane as a simple fact, but I'm going to keep John Henrying my own searching and not takling what the vending machine spits back.
While the bemoaning of the web can be laid at the enhittificating feet of the usual big entities, much also has been lost because we have reached for convenience of media streams to inform and packaged systems to make our web stuff. Web weaving as a craft is passé. To me searching for just pat answers is process if diminishing returns over the long run.
But hey, I'm an outlier. I'm outside the statistical mean of GenAI-ville. It's so much more interesting here. Thanks for letting me vent in your post ;-)
Thank you, this was helpful. One minor note on "Podcasting, for example, could grow even further as users tend to look for programs through podcatchers.". I am already using tools that summarize video and audio. So I don't see podcasts as an example of content that will not be gobbled up and reduced by AI. It is/will be (ab)used by AI, similarly to how it does with text content (that you described so well).
Many of the most popular podcasts and video channels are more personality-driven than information-driven. People find Joe Rogan, Amy Poehler, or Mr. Beast entertaining and want to spend hours "with" them; they wouldn't want a summary. Even with more topical podcasts, why some succeed better than others about the same/similar topic can have a lot to do with the personalities/presentational skills of the hosts.
(Would fans happily listen to hours of a convincing AI simulation of Joe Rogan or Amy Poehler interviewing convincing AI simulations of celebrity guests, or would they feel simulations lack (the illusion of) authenticity and connection? We'll most likely find out before too long...)
That's a powerful point, Amber. We want the human feeling. It's a reason why lonely people watch so much tv. And it does cut against search.
I suspect some people will accept AI personae. We're already seeing that with characterbots like Replika, but also with folks relying on chatbots for companionship.
Nate B. Jones recently posted a video about this issue and offered strategies for reconstructing websites so that one' content can be optimally integrated into the training data used as a basis for user responses in AI queries. It's a bit on the Web developer / techie side of the issue, but he has an accessible way of expressing how the systems work and how the Internet will adapt to this new paradigm.
I've written a bit about this early when Google launched / announced AI Search mode back in June. I don't see AI search collapsing - have you used Dia or Comet, the new AI browsers? That's the next evolution. It's another reason when professors try to "ban" AI, they are now basically telling them the can't use Google.
Bryan, this is a sharp and timely analysis. Your point that AI search is reshaping rather than revolutionizing our relationship to information is particularly thought-provoking. The movement from pluralistic exploration to synthesized authority raises serious concerns for both education and democracy.
As an educator, I share your concern about what this means for critical thinking and source literacy. If AI becomes the default interface for inquiry, how will we teach students to question, compare, and discern? Can institutions procure instruments with these safeguards embedded in the interface?
Your forecast of competing authority bots is chilling. Perhaps the antidote lies in building human-centered platforms, newsletters, forums, and classrooms that cultivate dialogue and nuance. As I reminded my housemate just now, the best technology is still the one between our ears.
Another thought, maybe a sci fi book about competing authority bots?
Certainly the way we, Deep Wisdom House, are traversing, and happily so: “Publishers are playing defense, building their own destinations and weaning themselves off platforms that use their content as training data. The goal is to own the audience connection and no longer be vulnerable to algorithmic and other platform shifts.”
One of your more crucial observations: "As I’ve said for years, and has started to happen, different entities will want to offer their own takes on the world: nations, religions, companies, etc."
Sifu, may I ask how Deep Wisdom is doing this? Are you hosting in-person events, virtual sessions, chats with readers, or...?
Open source agentic in-house network cores. And most certainly localised F2F game world community.
F2F gaming: is that sports, tabletop games, role playing...?
Call it role-playing intel -- it's Oracular Gaming contiguous to realtime unfolding events.
I love this so much.
I've come close, working on tabletop exercises for near-future scenarios.
Dear esteemed Georgetown colleague, our worlds apart are one. With respect from the Antipodes...
great post! thanks!
Potential long comment warning. There are some hairs to split when talking about "search" is one thing. Searching for specific facts or answers differs from searching when researching differers from what I most enjoy, just looking to explore. For the latter, I get more out of results beyond the best, things I would not even consider or just what pops up as curious. GenAI removes serendipty.
Also the concerns seem focused on impact of what people now rely on from search, to get more clicks on their stuff. You did touch on this in literacy but I see a dramatic decrease in understanding fundamental basics of what URLs do. I remember when Jon Udell spoke about the literacy of reading and fiddling with parameters in search results. Since modern web tools with visual editors remove us from the HTML soul of the web, I see people now knowing at all the link they are sharing is encrusted with UTM tracking tags.
Or today, on a web form, I see people submitting 50 mile long URLs generate by Outlook the stuff that starts with namo2.safelinks.protection.outlook.com. We have lost the wide understanding of how the web works.
And looking at the links most shared in socials as a reference are often reports or summaries from things written in online news sites that exist to attract clicks and spit ads, but also finding links share to research going to a university news post. We have lost the Mike Caulfield SIFT sensibility of going upstream to the sources. Each time I go upstream to find the best link, I invariably discover more of interest tangentially.
I relish doing my own analysis what to read, and in general feel like we are dumbing down society by replacing that mental exercise with the gray goo summarizing of AI. I glance at them, and I might grab and go if its something as mundane as a simple fact, but I'm going to keep John Henrying my own searching and not takling what the vending machine spits back.
While the bemoaning of the web can be laid at the enhittificating feet of the usual big entities, much also has been lost because we have reached for convenience of media streams to inform and packaged systems to make our web stuff. Web weaving as a craft is passé. To me searching for just pat answers is process if diminishing returns over the long run.
But hey, I'm an outlier. I'm outside the statistical mean of GenAI-ville. It's so much more interesting here. Thanks for letting me vent in your post ;-)
A fine comment, dog.
Great point about losing a sense of what URLs do, following on a loss of HTML. Ditto SIFT.
I agree on the change of behaviors. That's something I was getting at in the behavior and culture sections here. But you take it further.
You've given me an idea for my upcoming technology seminar. I should have them build HTML pages from scratch, building up web weaving from the basics.
Thank you, this was helpful. One minor note on "Podcasting, for example, could grow even further as users tend to look for programs through podcatchers.". I am already using tools that summarize video and audio. So I don't see podcasts as an example of content that will not be gobbled up and reduced by AI. It is/will be (ab)used by AI, similarly to how it does with text content (that you described so well).
Many of the most popular podcasts and video channels are more personality-driven than information-driven. People find Joe Rogan, Amy Poehler, or Mr. Beast entertaining and want to spend hours "with" them; they wouldn't want a summary. Even with more topical podcasts, why some succeed better than others about the same/similar topic can have a lot to do with the personalities/presentational skills of the hosts.
(Would fans happily listen to hours of a convincing AI simulation of Joe Rogan or Amy Poehler interviewing convincing AI simulations of celebrity guests, or would they feel simulations lack (the illusion of) authenticity and connection? We'll most likely find out before too long...)
That's a powerful point, Amber. We want the human feeling. It's a reason why lonely people watch so much tv. And it does cut against search.
I suspect some people will accept AI personae. We're already seeing that with characterbots like Replika, but also with folks relying on chatbots for companionship.
Good point. Any tools you'd recommend?
(That question was for Gabor)
Nate B. Jones recently posted a video about this issue and offered strategies for reconstructing websites so that one' content can be optimally integrated into the training data used as a basis for user responses in AI queries. It's a bit on the Web developer / techie side of the issue, but he has an accessible way of expressing how the systems work and how the Internet will adapt to this new paradigm.
https://youtu.be/hW5ne_14OQg?si=iXXnv8Q-rfJnTu3b
I've written a bit about this early when Google launched / announced AI Search mode back in June. I don't see AI search collapsing - have you used Dia or Comet, the new AI browsers? That's the next evolution. It's another reason when professors try to "ban" AI, they are now basically telling them the can't use Google.
I started using Comet. I'm not sure it's better than having the ChatGPT plugin in Chrome.
But what would keep the web going?