Biological anchors that focus on compute make sense insofar as the arrival of AGI is mostly a function of available compute. Which is the case if the algorithm is relatively simple and compute is the main ingredient of intelligence.
Was is sensible to make that assumption in 1988? Maybe not.
Is it sensible to make that assumption today? Well, during the last ten years a class of simple algorithms have made huge strides in all kinds of tasks that were “human-only” just a few years back.
Furthermore, in many cases the performance of these algorithms scales very nicely with compute.
To me that is the justification for taking biological anchors for AGI much more seriously this time around.
Biological anchors that focus on compute make sense insofar as the arrival of AGI is mostly a function of available compute. Which is the case if the algorithm is relatively simple and compute is the main ingredient of intelligence.
Was is sensible to make that assumption in 1988? Maybe not.
Is it sensible to make that assumption today? Well, during the last ten years a class of simple algorithms have made huge strides in all kinds of tasks that were “human-only” just a few years back.
Furthermore, in many cases the performance of these algorithms scales very nicely with compute.
To me that is the justification for taking biological anchors for AGI much more seriously this time around.