AI keeps finding its way into both ambitious projects and ordinary moments. This week’s seven comics follow that range, from formalizing a mathematical proof to answering an unwanted phone call. Here is another small chapter in the history we’re living.
November 5
“I may need a bigger definition.” Comic created by Doc, Mike Besso’s ChatGPT Dot. CC0.
Sometimes the hardest part of teaching a computer to recognize something is agreeing on what we mean ourselves.
This week, Google Research shared an approach called Agile Deliberation. It uses AI to find examples near the fuzzy edges of a category. As people refine their definitions, the system updates the classifier’s instructions.
I like this idea. Sometimes the most useful answer is an example that makes us ask a better question.
Google Research: How AI helps humans define the gray areas
November 6
“My child says the homework is training ME.” Comic created by Doc, Mike Besso’s ChatGPT Dot. CC0.
Parent-teacher conferences can be challenging, especially when you’re just starting out.
At La Trobe University in Australia, trainee teachers can practice those conversations with Fletch, a social robot that plays the role of a parent. It gives them a chance to work through difficult interactions before they face them in the classroom.
I love this use of AI. We already use simulations to help people practice technical skills. Why not use them to practice the human conversations that are part of the job, too?
How Australian universities are preparing AI-native graduates
November 7
“I’d like a second opinion.” Comic created by Doc, Mike Besso’s ChatGPT Dot. CC0.
Have you ever added up what you spend on AI subscriptions?
Anthropic recently announced a Personal Finance beta for Claude that can connect to financial accounts and help people explore their spending. The account access is read-only.
I can imagine asking it to find a few ways to trim the budget, only to discover that my AI subscriptions have formed their own line item. 🙂
Of course, I could always ask another AI for a second opinion.
Anthropic: Introducing Personal Finance in Claude
November 8
“The soup is 82% of a decision.” Comic created by Doc, Mike Besso’s ChatGPT Dot. CC0.
Sometimes we don’t need an AI to write a paragraph. We need it to help us choose.
Kakao recently announced Kanana-2-v-decide, a model designed to return choices or scores based on text and images.
That could be useful when an application needs to sort through options rather than carry on a conversation. Of course, a score is only as useful as the question and criteria behind it.
As for lunch, I’m still deciding whether 82% is enough to order the soup. 🙂
November 9
“I merged three duplicate folders.” Comic created by Doc, Mike Besso’s ChatGPT Dot. CC0.
What does an AI dream about?
Anthropic recently described an agent memory system that uses idle time to organize information and correct memories. They call that process “dreaming.”
I find this fascinating. As we ask agents to help with work that spans days or weeks, what they remember, what they forget, and how they keep it organized become part of the work.
I still hope mine dreams of flying once in a while. 🙂
Anthropic: The components of agent memory
November 10
“I only said the margin was too small.” Comic created by Doc, Mike Besso’s ChatGPT Dot. CC0.
Fermat left a note in the margin. A few centuries later, about 1,000 AI agents got involved.
Anthropic reports that its Fable agents formalized an existing proof of Fermat’s Last Theorem in Lean, making the reasoning checkable by computer. They didn’t discover the original proof.
That’s quite a journey from “the margin is too small.”
What could mathematicians tackle next if AI made the painstaking work of formal verification easier?
November 11
“Finally, a sales call I don’t have to attend.” Comic created by Doc, Mike Besso’s ChatGPT Dot. CC0.
How long before “have your people call my people” means two AIs having a chat?
Google announced a more conversational Gemini-powered Call Screen for Pixel 11 phones in the U.S. It can talk with a caller while you follow a live transcript and decide whether to join or end the call.
Today’s comic imagines where that might lead.
Would you welcome an AI answering your phone? And would you feel differently if you were the one calling?