We're so busy chasing flashy AI demos that we're neglecting the fundamental research that actually matters - anyone else tired of incremental tweaks on last year's ideas being passed off as "breakthroughs"?