In recent years researchers have been claiming to stride towards machines that have “human like vision”, “human like intelligence” and “human like general intelligence.” This goal has been championed and invested in with some of the greatest minds and the largest capital. But, is this goal indeed smart and meaningful? I would like to argue that doing that may actually waste time and resources.
Let us go back in time, where the only things that flew were birds.
Inspired by birds, man attempted to fly. When we tried to design machines that mimicked birds we failed. Miserably. Some inventors paid with their life. Only when we gave up on a moving wing and went to solid materials very different from feathers, propelled by an engine and not by flapping our arms, did we manage to fly. Our design, that is so different from most organisms, can fly faster than any bird, carry more weight and fly higher and in most weather conditions. Just imagine how long it would take us to fly if the Wright brothers would have insisted on trying to build an airplane that is “bird-like flying”.
A second similar story is the invention of the car, that really starts with the invention of the wheel. Our technology can move faster than any animal and carry much more weight. Just imagine if the inventor of the wheel would have insisted on developing a “horse-like machine”.
Which brings me to Artificial Intelligence and computers.
The hardware we use is dramatically different from the brain. It operates differently. It is great in some things like mathematical calculations, and bad in other things like knowing when I’m sad. The modern processor and the algorithms we have are the building blocks of AI, just like the wheel was for cars. Forcing to build “human like intelligence” is almost like trying to build a horse from wheels. The wheels are simply not the right building blocks for a horse and computers are not the right building blocks for “human like intelligence”.
However, if we learn from the past, what we see is that there is no need to build “human like intelligence”.
Instead, we can build a machine to solve any problem, starting from vision and going to industrial applications, medicine and decision making. The machine will be inspired by our ability to perform a certain task, just like we were inspired by birds to fly. However, if we just get out of the box of trying to make machines “be like humans”, we can certainly design machines that are more compatible to our hardware, perform better and much faster. These machines will not be “nice”. They will not be like us. But then again, why should they be?
We maximise the advantages of “machine like intelligence”, developing bespoke AI solutions to solve specific, real-world challenges.