winograd

Here [1] is a list of some Winograd schemata devised for testing artificial intelligence chatterbots.

These Winograd Schemas are basically a subset of the Turing Test, as originally intended.

It also seems like an appreciative backward glance at topics that interested Winograd earlier in his career, but are less fashionable now.

The piece actually seems like a backhanded critique of the optimism shown by the deep learning community (which of course Winograd is aware of).

> Machine learning researcher Hector Levesque of the University of Toronto proposes that resolving such ambiguous sentences, called Winograd schema, is a behavior worthy of the name intelligence.

With these Winograd schemas you could train a very specific machine learning model to guess the right co-reference relations (this is already a task in NLP that people compete for), and such a model is obviously not intelligent in any way, it can only do one thing.

The Coppersmith–Winograd algorithm, on the other hand, is probably "strictly of theoretical interest" for the time being...

The kind of "programming" that you are describing is fundamentally different than what Winograd did, and that was my point.

The Winograd Schema challenge as defined by Hector Levesque poses questions like "The ball could not fit through the hole because 'it' was too big.

At the same time, one thing that's notable about human intelligence is that, in comparison to source code, the learning and the skills gained in one context can be much very easily mobilized for use other contexts (see Winograd and Flores' use of Martin Heidegger's term "ready at hand" in their critique of AGI, etc).

It also disregards the influence on his thinking process from his advisor (Winograd) and fellow graduate students.

Anywhere you see him credit Winograd's influence, you are seeing him mention learning from his PhD.

Fibonacci heaps, AVL trees, KMP, Coppersmith-Winograd, these are all little works of art and theoretically great solutions.