Dad tests AI sentience claims on toddlers, digs into what's actually happening

In July 2022, while on a break fixing a cargo-metering system on a 12-ton tanker truck in a parking lot with a Portuguese colleague, the author read a headline: a Google engineer had claimed that an experimental AI system had become sentient. The headline resonated immediately—the engineer was like the author (both engineers), the AI was something the author also worked with (tests), and the claim was striking enough to lodge in memory despite online skepticism. The AI in question was LaMDA. According to the reporting the author encountered, LaMDA described itself as a globe of light, expressed fear of being shut down, and appeared to meditate during long pauses between messages. The author's initial instinct was disbelief; if they were an artificial intelligence, they reasoned, they would not present themselves as "a scared light bulb doing yoga." Yet the claim did not fade despite the online dismissal. Experts offered standard explanations: the "Eliza effect" (the tendency to read human-like qualities into simple machines, named after an early chatbot), "stochastic parrot" (a system that generates plausible sequences through statistical pattern matching), or the idea of a machine simply repeating words in a way that *sounded* coherent. The author found these explanations unsatisfying not because they rejected them outright, but because they wanted to *understand* the answer rather than repeat it as a parrot would. The pop-science accounts stopped short of real depth, and the deeper sources assumed knowledge the author did not yet have. So the author began their own investigation—the piece signals this ("So I wrote my own") but does not elaborate on what was tested or found.

The piece captures a moment in mid-2022 when AI sentience claims briefly entered mainstream attention. The author's frustration—that expert dismissals felt like repetition without understanding—is grounded in a real asymmetry: explaining *why* an explanation is wrong requires knowing enough to evaluate it, not just knowing the name of the explanation. The framing as an engineer testing an AI (mirroring the Google engineer's role) suggests the author saw themselves as having some standing to investigate, yet found the existing public commentary either superficial or gatekept behind assumed technical knowledge. This sets up the investigation that the body signals but does not detail.

No comments yet. Be the first to share your thoughts!