The Best Models Still Reason Like Toddlers - Andrew Dai, Elorian

AI Engineer18m 16s
Andrew Dai beside the headline AI STILL CAN'T SEE on a black background.
0 comments · 0 votesOpen discussionClose discussion
Sign in to join the discussion

    Andrew Dai argues that multimodal models can recognize visual patternsA vision-language model jointly processes images and language so it can describe, answer questions about or act on visual information. while failing to reason reliably about counts, spatial relationships and changes over time. Andrew Dai uses chess, board-game and object-trackingComputer vision enables software to analyze visual inputs such as images or video and extract information about their contents or structure. examples to distinguish recognition from deliberate visual reasoning.

    Andrew Dai describes Elorian's approach to multimodal data, synthetic trainingSynthetic data is information generated to represent examples or situations rather than collected directly from the events it describes., architecture and visual chains of thought. RoboticsEmbodied AI perceives and acts through a physical or simulated body while interacting with an environment., construction monitoring and CAD illustrate intended applications rather than independently verified production capability, and no safety guarantees are established.

    Original YouTube thumbnailWatch on YouTube

    Video topic coverage

    • Agents & workflows0
    • Coding & tools0
    • Models & learning100
    • Infrastructure & compute25
    • Research & evaluation100
    • Safety & security0
    • Business & productivity25
    • Creative & media0
    • Robotics & perception100
    • Society & policy0

    Share this video

    Andrew Dai beside the headline AI STILL CAN'T SEE on a black background. With WWW.ARTIFICIAL-INTELLIGENCE.VIDEO, 22 September 2026 and duration 18m 16s.

    Andrew Dai distinguishes visual recognition from reasoning and presents Elorian's approach to more grounded multimodal models.