The Counterfactual

Home
Archive
About
When words sound (or look) like what they mean
Not all form-meaning mappings are arbitrary.
 • 
Sean Trott
2
Informed consent is central to research ethics
On the unauthorized experiment conducted on a subreddit community.
 • 
Sean Trott
Identifying signatures of LLM-generated text
What metrics from open-source LLMs can tell us about the differences between human essays and those written by ChatGPT-3 and ChatGPT-4.
 • 
Sean Trott
LLM-ology and the "moving target" problem
How can a science of LLMs keep up with technological development?
 • 
Sean Trott
An overlooked problem with LLM benchmarks
Proxies for a capability may not map linearly onto the capability itself.
 • 
Sean Trott
2
Building inductive biases into LLMs
Language models are hungry for data—can we initialize their weights in ways that help reduce how much data they need?
 • 
Sean Trott
2
Learning across timescales
Knowledge distillation happens in many forms.
 • 
Sean Trott
2
The Counterfactual
Cognitive Science, statistics, AI.
Recommendations
Sustainability by numbers
Sustainability by numbers
Hannah Ritchie
Understanding AI
Understanding AI
Timothy B. Lee
User's avatar
The Discourse Lounge
Darrell Owens
Align
Align
Pam

The Counterfactual

AboutArchiveRecommendationsSitemap
© 2025 Sean Trott
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture