Flooding and astroturfing abound at the moment. Both of those things were happening before the advent of large language models (LLMs), but they have increased in frequency now that bad actors are able to just open the floodgates for content. Making large swaths of the internet that is just designed for search engine placement and self-referencial boosting has become so much easier recently. Sure all that bad data was abounding before this shift in what is now happening with OpenAI, Google, Microsoft, and Facebook recently sharing out chat services.
It’s one of those things where it is hard to put words on a page about it. Working with one of the chat systems to make content seems to trivialize the writing process. My day starts with an hour of focused academic work. That time is the fulfilled promise of decades of training that included a lot of hard work to get to this point. I can focus on a topic and work toward understanding it. All of that requires my focus and attention on something for that hour. Sometimes on the weekends I spend a couple of hours doing the same thing on a very focused topic. Those chat models with their large language model backends (LLM) produce content within seconds. It’s literally like a 1:60 ratio for output. It takes me an hour to produce what it creates within that minute including the time for the user to enter the prompt.
Maybe I did not expect this type of interaction to really affect me in this way. Everything has been questioned in terms of my writing output and what exactly is going to happen now. The door has been flung open to the creation of content. Central to that problem is the reality that the careful curation of content within academics and the publish first curation of the media are going to get flooded. Both systems are going to get absolutely overloaded with submissions. Something has to give based on the amount of attention that exists. They are not minting any new capacity for attention and the channels for grabbing that attention are relatively limited. The next couple of years are going to be a mad scrabble toward some sort of equilibrium between the competing forces of content curation and flooding.
This really is something that I’m concerned about on an onboarding basis. Do all the books, photos, articles, and paintings in the before times just end up with a higher value weighting going forward? Will this AI revolution have cheapened the next generation of information delivery in ways we will not fully get to appreciate until the wave has passed us and we can see the aftermath of that scenario? Those questions are at the heart of what I’m concerned about. Selfishly they are questions about the value and purpose of my own current writing efforts. More broadly they are questions about the value of writing within our civil society as we work toward the curation of sharable knowledge. We all work toward that perfect possible future either with purpose or without it. Knowledge is built on the shoulders of the giants that came before us adding to collective understanding of the world around us. Anyone with access and an adventurous spirit can pick up the advancement of some very complex efforts to enhance the academy's knowledge on a topic.
Maybe I’m worried that the degree of flooding with flatten information so much that the ability to move things forward will diminish. Sorting, seeking, and trying to distill value from an oversupply of newly minted information may well create that diminishing effect. We will move from intellectual overcrowding in the academy to just an overwhelming sea of derivative content marching along beyond any ability to constrain or consume. I’m going to stop with that last argument as it may be the best way to sum this up.
Links and thoughts:
Top 5 Tweets of the week:
https://twitter.com/CaseyNewton/status/1651967468322320387
https://twitter.com/Benioff/status/1651605849310130178
https://twitter.com/AndrewYNg/status/1651605660382134274
https://twitter.com/verge/status/1651355922751422464
https://twitter.com/NateSilver538/status/1650899579234140168
What’s next for The Lindahl Letter?
Week 120: That one with an obligatory AI trend’s post
Week 121: Considering an independent study applied AI syllabus
Week 122: Will AI be a platform or a service?
Week 123: Considering open source AI
Week 124: Profiling OpenAI
If you enjoyed this content, then please take a moment and share it with a friend. If you are new to The Lindahl Letter, then please consider subscribing. New editions arrive every Friday. Thank you and enjoy the week ahead.
All that bad data abounds