The Best Models Still Reason Like Toddlers — Andrew Dai, Elorian
Source
youtube.com
Author
AI Engineer
Date
Context
Recognizing a familiar object can interfere with answering a question about the particular image in front of a model. In the description of Andrew Dai's talk, the example is a partially visible chessboard: asked how many white squares are visible, a frontier model answers with the number on a complete board. Dai argues that reliable visual reasoning requires tracking and counting the visible evidence. His proposed approach includes intermediate visual steps, such as drawing boxes around candidate objects before narrowing to the relevant ones. That is a research direction for tasks involving spatial detail; the described examples do not establish a ranking of vision models.
Key takeaways · AI-distilled
Andrew Dai's chessboard example: shown only part of a board, a frontier model answers 32 white squares because it recognizes a chessboard and recalls the full count, rather than counting what is actually visible.
His rule of thumb: if a person can answer a visual question in about one second, the model can too; if it takes a human longer, models fall apart, as with counting Catan roads or tracking a robot arm lifting a lid across a long video.
Dai says benchmarks hide the gap: one popular reasoning suite uses 32 by 32 pixel images, and a multimodalA model that works with more than text — reading images, audio, or video, and sometimes generating them too.Full definition → science exam can mostly be answered without looking at the image.
Elorian's approach combines visual reasoning data that does not exist online, a synthetic flywheel of evals, agents, SFT and RL, architectural changes on the transformerThe neural network architecture behind modern AI models, built on attention — letting every word directly consider every other word in parallel.Full definition →, and native visual chain of thought, such as boxing every hotel before narrowing to the red ones.
Terms in this piece · Glossary
multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
hallucination — When a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.
chain-of-thought — Having a model write out intermediate reasoning steps before its answer, which markedly improves performance on hard problems.
transformer — The neural network architecture behind modern AI models, built on attention — letting every word directly consider every other word in parallel.
videoHow We Mapped a 35-Million-Line Mainframe in 4 Days - Stephen Chin, Neo4j & Neeraj Deshmukh, Blitzy
videoHow Netflix Evaluates AI When There's No Right Answer - Cameron Wolfe, Netflix
videoStop Giving Your AI Agents API Keys - Kim Maida, Keycard & Mason Egger, Temporal
videoYour AI Agent Keeps Making the Same Mistakes - Sparsh Jain, CentralAgent AI
