Posts – Ken Arnold (Calvin U)

Why the University?

When content is free, what’s the value of educational institutions?

Reflective Paper Reviewing

llm

interaction

reflection

I was tasked with reviewing a paper for a workshop, but the paper was so unclear that I had trouble seeing past the clarity issues to understand the contributions in order…

Careful Translation Workflow

llm

interaction

hospitality

Using generative AI in a transparent workflow for high-quality translations.

Screen-Free Reflective Practice using GenAI

llm

interaction

reflection

Voice notes make space for deep reflection on professional practice without screen distractions; LLMs help synthesize these reflections

GenAI in Reflective Mode

llm

interaction

reflection

GenAI can help you reflect on your work, not just get it done faster.

Reflective Communication

llm

interaction

reflection

GenAI can help you reflect on how your communication serves others.

Interacting with AI: X instead of Y

ux

ai

AI is typically designed in ways that push people aside, but what if we could reimagine the relationship?

ML Interactives

coding

ai

ml

As I’ve been teaching AI / Machine Learning, I’ve made a few interactive demos to help students understand some concepts. Here’s a few:

Debugging Hints

coding

Here are a few debugging hints that I often give students.

Mapping to Mimicry

ai

ml

math

Loss minimization is a flexible tool.

Simplifying Instructions

LMs can simplify LM system messages.

Intelligence via Mimicry

AI systems aren’t magic or inherently objective. For all their strengths, most AI systems mimic human behavior.

Flourishing with Intelligence Augmentation Systems

I’ve noticed some dominant narratives in public discourse around doing-what’s-right-with-AI. I’ll give them some oversimplified names; this is a complex topic.

Information and Decision

hci

ml

ia

A design space of interactions with AI systems

Transparency for Generated Text

ml

llm

nlp

Lightweight marks for AI-generated text

Two Ways that LLMs Hallucinate

ai

ml

lm

Do you ever make your decision first and think about your reasons later?

Does increasing ‘alignment’ increase risk?

ai

OpenAI’s stated purpose for working on the sort of adjustments to language models that led to ChatGPT was to “align” their behavior with human goals. Their motivation for…

Language Models have Personas?

ai

ml

lm

RLHF is fascinating.

Humanity of the Gaps

ai

perspectival

Reality is spiky.

ChatGPT Doesn’t Have a Database of Facts

ai

ml

It was trained to be believable, not to be right.

Self-Study Advice

coding

Practice solving problems.

Multidimensional Array Terminology

ml

coding

A surprisingly large amount of the thinking that goes into implementing neural net code is getting the shapes of multidimensional structures right. I’d heard that from…

Avoid Magical Thinking about AI

ai

ml

Even if you’re not a computer scientist, you should learn enough about AI that you can steer its progress.

Stop trying to detect computer-generated writing.

ai

ml

lm

It’s a risky distraction that just encourages an arms race.

Solomon’s proverbs about AI

ai

ml

lm

A reflection without words on the difference between wisdom and its imitation.

Without Acknowledgment: Cultivating Gratitude in an ML age

ai

ml

lm

ia

Creativity support tools based on machine learning hide the human authors that contributed to the work. Does it have to be that way?

Language Models for Species-Scale Collaboration

ai

ml

lm

ia

How knowledge fusion will continue the trend of scaling humam collaboration