The concept itself is pretty amazing to think about, but I wouldn't expect the long-term implications of intelligence to play out nearly as dramatically as the author implies. While coming from means AND having high intelligence would confer benefits to be sure, there has to be a law of diminishing returns on increasing intelligence. For sure, the short-term inequalities would be difficult to keep in check. But if the entire world has increased intelligence then those benefits will quickly deteriorate.
Something I learned in grad school is that there is a continent of linear problems surrounded by exactly solvable islands in a vast unsolvable sea.
For instance, you can solve quadratic equations easily with a closed form formula, and do cubics and quartics with just a little more work, but there is no closed form for quintics and higher and no amount of "intelligence" is going to change that.
Similarly, without the right experimental results, no amount of "intelligence" is going to lead to a fundamental breakthrough in theoretical physics.
You might be able to throw more speed, more computing units, a larger memory, etc. at a problem but that doesn't mean you're going to get proportionally better results.
I think you are assuming that the "cloud brains" will just be number crunchers like today's computers. Human intelligence and computer intelligence are fundamentally different. Sure, plugging directly into a massive calculator is useful, but does have it limits. However, this article is more about running simulated copies of our neural patterns. The human brain is limited to only a handful of simultaneous ideas. Imagine if you could synthesis an idea with 100 inputs instead of just 5. Or if you go with the spontaneous theory of creativity, you can parallelize the creative process, and just have a higher chance of having a good idea by increasing the number of ideas had. This is fundamentally different from raw calculation. Both are likely to be useful, but both will serve different purposes.
I think the outcome for that is even more pessimistic.
If you generate 100x more ideas most of them are going to be rehashings of the same old ideas, not that much better than what you had before. You probably get some more value, but not 100x. Whether value goes up logarithmically or reaches an asymtope, the benefits of throwing more neurons on it will become uneconomic.
Also I think creativity is overvalued. Look at all the things Buckminister Fuller invented that went nowhere:
4 comments
[ 5.7 ms ] story [ 13.8 ms ] threadSomething I learned in grad school is that there is a continent of linear problems surrounded by exactly solvable islands in a vast unsolvable sea.
For instance, you can solve quadratic equations easily with a closed form formula, and do cubics and quartics with just a little more work, but there is no closed form for quintics and higher and no amount of "intelligence" is going to change that.
Similarly, without the right experimental results, no amount of "intelligence" is going to lead to a fundamental breakthrough in theoretical physics.
You might be able to throw more speed, more computing units, a larger memory, etc. at a problem but that doesn't mean you're going to get proportionally better results.
If you generate 100x more ideas most of them are going to be rehashings of the same old ideas, not that much better than what you had before. You probably get some more value, but not 100x. Whether value goes up logarithmically or reaches an asymtope, the benefits of throwing more neurons on it will become uneconomic.
Also I think creativity is overvalued. Look at all the things Buckminister Fuller invented that went nowhere:
https://bfi.org/about-fuller/big-ideas/dymaxion-world/dymaxi...