While having a quick lunch today, I happen to be thinking about an old topic.
Humans and computers are each good at different tasks. Humans are very good at finding associations. For example, we might realize that ice is related to cold, cold (temperature) is related to cold (disease), cold is related to a runny nose. Identifying these simple relations seems simple for us. Yet, for computers, this is a major task with no fast solution. This is one of the main reasons for computers not being able to plan as well as humans under changing circumstances. Almost all forms of computer planning, must be done on very well defined problem domains or, with naive assumptions.
On the other hand, computers score well in accuracy (i.e. perfect memory), repetition, calculation speed, and large amounts of perfectly retrievable data. Humans "store" large amount of information too, albeit imperfectly due to memory loss. Furthermore, it has been hypothesized that our memory is not absolute, but approximated (i.e. it is reconstructed when needed from associations and belief). Worse, belief can actually shape this reconstruction and we end up thinking that we remember, often inaccurately. Following this, my hunch is that the more one tries to remember, the more the recollection is inaccurate. Kind of like the movie Big Fish.
Humans are always trying to perfect computers to be like themselves. After all, since they are already so good at doing those previously mentioned tasks, if they can think like us, they will be perfect! Hence, the many Sci-Fi movies and novels, depicting computers as being superior to humans. However the myriad of variables uncertainty brings makes this a daunting task (even calculating conditional probabilities of an event given 10 other events is already a ridiculous amount of computation). Maybe if computers sacrifice their accuracy they can become like us, but hey! Then humans will be back at square one, making babies will probably be more fun.
From my limited reading experience, only one series of novels presents an interesting alternative - humans learning to do what computers can do, hence becoming superior. This series is Frank Herbert's Dune novels and subsequently the prequels written by his son. I did not have time to read every book so I just browsed the summaries on Wikipedia. These are what I gathered.
In the Dune universe, there are powerful machines but AI is tabooed. This is due to some eons ago, humans were enslaved by intelligent machines that they created. So how did they free themselves from their machines? Apparently, one day the machine Overload (i.e. big boss) made a wager with his adviser. The Overload believed humans were inferior but its adviser believed that humans can be trained to do what computers can do. Hence to prove its (damn everything is it, I mean the adviser) point, it created a human computer, a human capable of blindingly fast computation and can accurately store and retrieve vast amounts of information at the same time retaining the traditional abilities of humans. This was called the first Mentat (i.e. a human computer).
The result was revolution, the humans took over again and AI is outlawed. Humans began to augment their abilities through genetic manipulation and a form of drug abuse of something called spice to allow heightened awareness. Mentats were advisors to empires and were basically all knowing super planners. Space guild navigators were so warped they had brains so large they swam in tanks. Their job was to perform immeasurable amount of computation at unimaginable speeds for spacecraft to navigate folded space. All rather interesting but also rather gross.
So which will it be first? Computers like us or us like Computers? Thus begins the battle between the Computer Science and Life Science disciplines with Psychologists sitting on the fence. It is amazing the amount of thoughts that can arise from a quick lunch...
Humans and computers are each good at different tasks. Humans are very good at finding associations. For example, we might realize that ice is related to cold, cold (temperature) is related to cold (disease), cold is related to a runny nose. Identifying these simple relations seems simple for us. Yet, for computers, this is a major task with no fast solution. This is one of the main reasons for computers not being able to plan as well as humans under changing circumstances. Almost all forms of computer planning, must be done on very well defined problem domains or, with naive assumptions.
On the other hand, computers score well in accuracy (i.e. perfect memory), repetition, calculation speed, and large amounts of perfectly retrievable data. Humans "store" large amount of information too, albeit imperfectly due to memory loss. Furthermore, it has been hypothesized that our memory is not absolute, but approximated (i.e. it is reconstructed when needed from associations and belief). Worse, belief can actually shape this reconstruction and we end up thinking that we remember, often inaccurately. Following this, my hunch is that the more one tries to remember, the more the recollection is inaccurate. Kind of like the movie Big Fish.
Humans are always trying to perfect computers to be like themselves. After all, since they are already so good at doing those previously mentioned tasks, if they can think like us, they will be perfect! Hence, the many Sci-Fi movies and novels, depicting computers as being superior to humans. However the myriad of variables uncertainty brings makes this a daunting task (even calculating conditional probabilities of an event given 10 other events is already a ridiculous amount of computation). Maybe if computers sacrifice their accuracy they can become like us, but hey! Then humans will be back at square one, making babies will probably be more fun.
From my limited reading experience, only one series of novels presents an interesting alternative - humans learning to do what computers can do, hence becoming superior. This series is Frank Herbert's Dune novels and subsequently the prequels written by his son. I did not have time to read every book so I just browsed the summaries on Wikipedia. These are what I gathered.
In the Dune universe, there are powerful machines but AI is tabooed. This is due to some eons ago, humans were enslaved by intelligent machines that they created. So how did they free themselves from their machines? Apparently, one day the machine Overload (i.e. big boss) made a wager with his adviser. The Overload believed humans were inferior but its adviser believed that humans can be trained to do what computers can do. Hence to prove its (damn everything is it, I mean the adviser) point, it created a human computer, a human capable of blindingly fast computation and can accurately store and retrieve vast amounts of information at the same time retaining the traditional abilities of humans. This was called the first Mentat (i.e. a human computer).
The result was revolution, the humans took over again and AI is outlawed. Humans began to augment their abilities through genetic manipulation and a form of drug abuse of something called spice to allow heightened awareness. Mentats were advisors to empires and were basically all knowing super planners. Space guild navigators were so warped they had brains so large they swam in tanks. Their job was to perform immeasurable amount of computation at unimaginable speeds for spacecraft to navigate folded space. All rather interesting but also rather gross.
So which will it be first? Computers like us or us like Computers? Thus begins the battle between the Computer Science and Life Science disciplines with Psychologists sitting on the fence. It is amazing the amount of thoughts that can arise from a quick lunch...
No comments:
Post a Comment