According to Andrew Ng, he proposes the Turing-AGI Test as a new benchmark where either a computer or a skilled human completes multi-day real-work tasks via a computer with internet tools, and a computer passes if it performs as well as a skilled human, providing a practical measure of human-level capability for knowledge work, source: Andrew Ng via X on Jan 6, 2026; DeepLearning.AI The Batch, Issue 334. According to Andrew Ng, the setup includes access to a web browser and Zoom, with judges free to design undisclosed training and execution phases (for example, training as a call center operator then taking calls with feedback), mirroring remote work and probing generality beyond fixed datasets, source: Andrew Ng via X on Jan 6, 2026; DeepLearning.AI The Batch, Issue 334. According to Andrew Ng, this aligns with the public’s view of AGI, addresses shortcomings of the original Turing Test and leaderboard-tuned benchmarks, and focuses on economically useful work rather than fooling judges, source: Andrew Ng via X on Jan 6, 2026; DeepLearning.AI The Batch, Issue 334. According to Andrew Ng, unchecked AGI hype lowers the bar for claims and risks an investment bubble and subsequent collapse of interest, while a rigorous test could defuse hype and support a steadier path of continued investment and real progress, source: Andrew Ng via X on Jan 6, 2026; DeepLearning.AI The Batch, Issue 334. According to Andrew Ng, a Turing-AGI Test competition would create clear outcomes where broad failure would intentionally defuse hype and passing would be incredibly valuable, defining event risk and binary milestones that traders can monitor for AI-linked equities and crypto projects tied to AI narratives, source: Andrew Ng via X on Jan 6, 2026; DeepLearning.AI The Batch, Issue 334.