The holy grail for the next decades would be to move from ANI to AGI, Artificial General Intelligence, or to an AI that can mimic and best our own general human intelligence.
This seems now within reach and just a matter of time. Some are putting the crossover (machines more intelligent than humans) in the fourth decade, a few years after 2030, well before 2050. This forecast is based on two considerations:
- the past has shown that by increasing processing power, storage and communications we have been able to create ANI that best humans;
- the processing power of machines, their storage and connectivity capacity will continue to grow and will go beyond our comparable brain processing/storage/communications power.
Although the first consideration is a “fact” the second one is not. First of all it is not a given that AGI would be just a matter of processing/storage/communications power (several scientists are convinced this is the case, others oppose this) and, secondly, it is extremely difficult, and so far arbitrary, to measure our brain computation power. It may be worth considering that we have no clue, today, why one brain is “more” intelligent than another, and that most brains look very much alike although their "intelligence" differs.
However, I do not think that a comparison between AI and brain is really the point. I am convinced that in the same way we have built machines that are better (faster, more accurate, more powerful) than our bodies without mimicking our muscles, in a few years, possibly in the next decade, we will build machine that are better thinkers than we are, even though their hardware may not be comparable to ours.
By 2050 many of the results of "thinking tasks" we are normally engaging in will be bested by “thinking tasks” carried out by machines or interconnected machines. Would a machine be able to write a nice article or an engaging book? Definitely Yes! We can discuss on when and what “nice” or “engaging” means but I have no doubt that by 2050 the discussion will be over. Already today we have computers paintings and computer music that is indistinguishable from painting and music created by humans.
So, just looking at the “result” (or at the Turing test) my bet is that we will have AGI well before 2050.
But ….
