"I think we might have AI that is smarter than any human by the end of this year, and no later than next year, probably by 2030 or 2031, AI will be smarter than all of humanity collectively." — Elon Musk
While generally sharing Elon Musk's concerns regarding the unpredictable consequences of artificial intelligence implementation, I perceive a fundamentally different strategic outlook for its evolution.
Firstly, to compare artificial intelligence with human intelligence, one must operate with unambiguous, quantifiable definitions.
As of today, not a single methodology has been developed worldwide capable of comprehensively and reliably evaluating human intellect. Moreover, classical psychotherapy posits that the subconscious drives over 90% of human activity (a fact corroborated by foundational cognitive neuroscience research on the automaticity of higher nervous activity). If this core component of the psyche remains unperceived by the individual, how can it possibly be algorithmized and, much less, measured by a machine?
Secondly, computers have long surpassed humans in chess, and a precision-guided cruise missile has become exponentially more efficient within its flight program than any living infantryman. Accordingly, if we are to evaluate intellectual parameters at all, it is more logical to begin with the specific individuals who govern these missiles, algorithms, and artificial intelligence systems.
Thirdly, turning to the conscious domain of human mind, there is likewise no unified scale to extrapolate its evaluation methods to AI. Yet it is precisely at this intersection that the projection of AI onto human reality will manifest.
All breakthrough classes and methods of AI operate effectively strictly within deterministic spaces—areas where training datasets can uncover mathematical patterns, or where these patterns can be derived from the established laws of physics and natural sciences.
Any living individual perceives themselves in reality not merely through "discrete" intelligence—which can, to some extent, be compared to machine logic—but fundamentally via their "analog" senses that forge an inimitable individuality. These are best categorized into physiological and moral senses.
Physiological senses: vision, touch, smell, hearing, and taste discrimination. These senses are objective, as they can be modeled and measured by technical means with high precision. This is exactly where the colossal, yet far from fully realized, potential of AI lies, even under current computational capacities.
However, everything pertaining to the concepts of human morality does not constitute a deterministic space—if only due to the presence of free will and the capacity for unpredictable moral choices made by any individual at any given moment.
Replicating a "noble terminator" or a "noble cinderella" through AI is achievable only if the training model is fed a highly representative array of historical precedents of such personalities.
For instance, fully overcoming the "Uncanny Valley" phenomenon marks the next major milestone in generative models. It is evident that in the near future, these methods will allow for the seamless localization of TV series and entertainment shows. Yet, crafting a screenplay capable of captivating and genuinely resonating with a self-actualizing individual will remain the sole domain of other self-actualizing humans.