Chris MacLellan

155 Followers
168 Following
450 Posts

Assistant Professor at Georgia Tech's School of Interactive Computing.

I do research at the intersection of #CognitiveSystems, #AI, #HCI, #CognitiveScience, and #LearningScience. My work focuses on understanding how people teach and learn and building computational systems that can teach and learn like people do.


Outside work, I am into #dogs, #backyardchickens, #beekeeping, and #sourdough. I love #scifi and #fantasy.

Personal Websitehttps://chrismaclellan.com
Lab Websitehttps://tail.cc.gatech.edu
Google Scholarhttps://scholar.google.com/citations?hl=en&user=q_K8iFQAAAAJ
With the new video of GPT-4o tutoring a young person on a geometry problem, I thought I’d test it out by throwing it a curve ball…
Lane Lawley just did an awesome job presenting our new interactive task learning system called VAL at #chi2024. Our system can learn new tasks on the fly from interactive language instructions and can then execute these tasks at the user’s command. Although we use an LLM to support natural language interaction, VAL’s real power comes from its hierarchical task network representation/planner, which lets it generalize and reason. You can learn more about our system here: https://dl.acm.org/doi/10.1145/3613904.3641915
VAL: Interactive Task Learning with GPT Dialog Parsing | Proceedings of the CHI Conference on Human Factors in Computing Systems

ACM Conferences
Went outside this afternoon to this—my #dog Freyja loves to sun bath on her back ☀️. When I got closer she was snoring! This dog knows how to live 😂

My new favorite ChatGPT “glitch”.

I’ve found it is impossible to get ChatGPT (with DALLE) to generate a pentagon. I’ve tried several prompts, such as asking for a pentagon (it gives a hexagon), asking for a baseball home plate (which it depicts as a square), and asking for a Birds Eye view of the pentagon (which it renders with six sides instead of five).

Many AI/ML systems can achieve human-like performance (or even super human!), but come no where close when it comes to exhibiting human-like learning capabilities. E.g., humans can engage in online, incremental, continual, interactive, and data/compute efficient learning (not even mentioning social/cultural learning capabilities).

We’re organizing a AAAI Spring Symposium to explore this topic (https://humanlikelearning.com/aaai24-ss/). If this sounds interesting, consider participating!

AAAI-24 Spring Symposium on Human-Like Learning

Exploring how human-like learning can be realized in artificial systems.

Human-Like Learning
Holiday husky #dogs #dogsofmastodon #dog
Just wanted to share a #dog photo. Valkyrie is loving laying out on the grass in the sun.

Interested in #CognitiveSystems?

The Tenth Annual Conference on Advances in Cognitive Systems (http://cogsys.org/conference/2022/) will take place from 1PM on Saturday, November 19 into the late afternoon on Monday, November 21st. Additional workshops will occur on Tuesday, November 22nd.

The conference will be hybrid and free to virtual attendees. Virtual registration is here: https://docs.google.com/forms/d/e/1FAIpQLSdxlvtW95NiuozTaVz3Ie0Hbydmn-eCHnxKLGGAnPFa-alqBg/viewform

Here are the topics covered:

ACS-2022

#Introduction I do #CognitiveSystems and #HCI research to better understand how people teach and learn and to build machines that can teach and learn like people do.

What is #CognitiveSystems? It aims to better understand the nature of the mind. It adopts a broad perspective that revisits earlier classical #SymbolicAI ideas in light of recent advancements. I like to think of my work as #NeoclassicalAI, or as @shiwali nicely puts it, the OG #AI.

Here is a non-AI generated photo of my dog!