[pygy_]

Any computer program is an AI.

What matters is how intelligent they are along various axes.

Automatic programs have been surpassing humans on some dimensions for ages, but we keep insisting that they are not truly intelligent because they can't beat us along all axes. Throughput on simple logic tasks was the elephant in the room, and the scope of "simple" has been expanding at an exponential pace.

Now they are closing the gap or surpassing us on axes that were thought to be bastions of human cognition (TFA, and after chess and go, Google (Alpha Zero) recently beat two Starcraft 2 champions).

Freaking out (err... I mean "not releasing the full model") is understandable, but ultimately misguided as it will only delay the unavoidable... Unless the plan is to enact a global ban on AI research which I don't think is feasible anyway.