Andrew Piper
@_akpiper
Followers
6K
Following
10K
Media
846
Statuses
15K
Using #AI and #NLP to study storytelling at McGillU. Director of .txtlab and author of the forthcoming book, Why You Should Read More Fiction.
Montreal, QC
Joined March 2012
This rings true to my experience in AI art. Experienced artists are curious and want to learn more - they see it as a useful tool to express their ideas. Haters are typically amateurs who don't like that people using AI now have the skills they had to spend time learning.
amateur artists are way more anti-ai than professional artists
155
103
780
Yikes that's alarmist. I guess "Book Was There" will now be mandatory reading.
press.uchicago.edu
Andrew Piper grew up liking books and loving computers. While occasionally burying his nose in books, he was going to computer camp, programming his Radio Shack TRS-80, and playing Pong. His eventual...
“The intellectual and cultural apocalypse is near. Even if your family thinks of itself as well-educated, your kids will grow up unable to work their way through a classic novel…they will embody some of the worst and most volatile aspects of TikTok”
0
0
0
This bodes ill. Readers used to outnumber non-readers 2 to 1. Now non-readers outnumber readers 3 to 1. It's hard to imagine a change of that magnitude not having significant effects.
340
1K
6K
This is the gift. USDA Prime steaks. Gold box. Dry ice unboxing. Delivered by Christmas. They'll remember who sent it. 6 FREE Petite Ribeyes + FREE shipping ($240 value) on $229+ orders Code SANTA229 🥩
1
7
20
This is certainly how I feel about research now.
Creators should LOVE Ai. AI doesn’t make uncreative people creative. It allows creators to become exponentially more creative. Creative iteration that used to take hours , days or weeks, can happen in minutes. The different routes you wanted to take but didn’t have the
0
0
3
I like this take. The agents are all talking to each other online crowding us out and we have to go back to...hanging out in person. 🥰
I'm convinced that the web will be inhabited mostly by AI agents in 10-15 years (as we watch the human internet die). The main question is where will humans be and what we'll be doing, or what type of panem et circenses entertainment will be thrown at us. Anyone with a
0
0
2
Merriam-Webster’s human editors have chosen ‘slop’ as the 2025 Word of the Year.
450
7K
62K
such an important reminder! Stories fundamentally misleading. They shrink the world down and provide reductive causal explanations. This is their value and their danger.
Our brains are wired to turn everything into a story with clean cause and effect. Steve Jobs succeeded because he was adopted. The athlete made it because of a tough childhood. These explanations feel true because they're satisfying, not because they're accurate. The problem:
0
0
0
I've tried this. It's very hard to scale. A seminar of 35 would take 11.6 hours at 20 minutes a piece. The other problem is subjectivity. They feel *very* subjective compared to grading written exams. Which is why I bet written exams emerged in the first place.
Professors are turning to oral exams to stop AI use
0
1
3
If you have to read anything about the prospect of “automating scientific discovery,” “agents for science,” or integrating LLMs into scientific pipelines, please let it be this essay by Kevin T. Baker: https://t.co/uJDkSD2EW5
6
29
152
Fantastic take on the problems of automated science by @beenwrekt. The question for me is whether incremental at scale = innovation.
0
0
0
This is why it's going to get harder and harder to measure AI improvement short of actual deployment.
0
0
4
All you need is narrative.
storytelling is the only way to impose meaning on abundance, coherence on noise, & legitimacy on power. strategy, ops, & capital are all downstream. without narrative control, none it will ever stick. this has been the core premise of my account. in a world of infinite output,
0
0
0
My @TheAtlantic piece on the emerging new "Physics of Life". Complexity science, information theory, network theory... a whole new trans-disciplinary way of asking an ancient question. https://t.co/DQS331dHbt
theatlantic.com
The fundamental nature of living things challenges assumptions that physicists have held for centuries.
12
51
331
I wonder if writing has a differentiation effect. Where we consciously suppress em-dashes and arguing in threes to signal our authenticity.
1/ New research analyzing 740,000 hours of human speech has found undeniable proof: we are starting to speak like ChatGPT. Not just in emails. In real life. Here is the empirical evidence of the "Cultural Feedback Loop." 🧵👇 2/ We all know the jokes. You see the word "delve"
0
0
0
I often wonder whether this is like the AI girl/boyrfriend problem. When models update we have to update our expectations. Feels disappointing until you adapt to your new partner. I went through this with 5.1.
Ilya is 100% correct .it's a pattern that keeps repeating It's very clear with GPT5.2 Overfit the model to produce impressive looking benchmarks, have it excels in a few domains, but fall flat in many others. There's not enough generalization, and even if there is, the model has
1
0
3
The Silo is very good TV. It fits perfectly with what we call the small world effects of fiction. Running counterfactuals in controlled settings. Sci-fi loves capsules.
0
0
0
@squarehaunting The libraries are paywalled *functionally*—not by access but by time/capacity to parse them. AI doesn't replace thinking, it compresses the search space. A PhD student using LLMs to map a literature review in hours instead of months isn't thinking less—they're thinking *at the
3
1
9
I used to be very critical of the eco criticism of AI. Compare it to streaming or planes! But the demand for application resource needs are so high I don’t see how we scale without energy consumption going nuts.
0
0
0
“There will always be reasons to prefer lighter-weight specialized components over invoking maximum capability for every task.”
(I know I'm a stuck record) An important assumption in AI discourse is that sufficiently capable generalist *models* are the main event. Get the model smart enough, and it more or less does everything. Value creation, competitive advantage, and risk would all concentrate at the
0
0
1