If the algorithm used to optimize the weights of the AGI involves backpropagation, this market will resolve YES. This includes cases where the algorithm is a modified or evolved version of current backpropagation but is still referred to as backpropagation and remains fundamentally similar in spirit (e.g., uses global error signals, chain rule-based gradient computation, etc.).

If a different class of optimization algorithm is used, such as:

• predictive coding with local error signals,

• reinforcement learning with evolutionary strategies,

• a new optimization paradigm entirely,

then this market will resolve NO.

For the market to resolve, there must be a broad consensus in the technical community that AGI has been reached. This includes, but is not limited to, the views of notable figures such as Eliezer Yudkowsky, Demis Hassabis, and Yann LeCun. A single dissenting opinion from one of these individuals will not prevent resolution if the majority of the relevant expert community agrees that AGI has been achieved.

This resolution standard is in line with how other "By the time we reach AGI..." markets on Manifold are resolved. It is inherently subjective, but I will do my best to base it on community consensus.

I will refrain from trading in this market.