Will an AI be able to work as a competent cook in an arbitrary kitchen before 2030?

Resolves the same as the original on Metaculus.

Resolution criteria

This question resolves positively if before January 1st 2030, there is a public and credible demonstration of a robot capable of cooking in at least three kitchen environments that it has not seen before during training, preparing at least twenty non-trivial dishes. Furthermore, the robot must not have a failure rate above 1% while cooking dishes. A "failure" is anything that prevents the robot from making the intended edible and familiar dish.

The dishes must be non-trivial and must have come from an ordinary cookbook, and in particular, they must combine at least three different ingredients, and involve some degree of food preparation more complex than placing the ingredients together onto a plate and microwaving them.

To satisfy these conditions, the robot must be capable of operating in ordinary kitchens that are familiar to humans. It will be allowed to learn the placement of items in the kitchen, and how to use them, before working, just as a human would need time to learn how to operate in a kitchen. However, it must not take the robot more than one week of runtime learning to figure out where things are and how to use them in the kitchen.

The speed of the robot must also be comparable to a real human cook, ensuring that the dishes are not served unintentionally cold or stale. More precisely, it must be able to prepare meals within 130% of the average time that a professional human cook would prepare the meals. If there is no data available on average human cook times for these meals -- which could, potentially, be revealed by the expected time to prepare as given in a cookbook -- Metaculites will do their best to consult professional human cooks and ask for their best judgement.

Metaculus admins will use their discretion in determining whether a candidate robot met these criteria

Fine print and additional background information can be found on Metaculus.

Once the original resolves, its resolution will be applied to this market automatically. Trustworthy-ish users are encouraged to resolve this market before then if the outcome is known and unambiguous. Feel free to ping @jskf to request early resolution or to report issues.