I finally watched “2001 A space odyssey”. I have tried to watch it 3 times already but never got passed the first 30mn because the beginning is really slow. I wanted to watch it to finally meet HAL for real after hearing about him in all my readings. I enjoyed the movie, there are some interesting ideas there, and I retained some amusing quotes :
“I’m sure you know what the problem is, Dave.”
It’s really amusing that a machine could ever tell that to a human because machines are designed to be straightforward, they know or not. There is no “well, take a guess at it”, or “what does it look like?”, the dialog with a machine is straightforward and I think it should always keep that way. Their purpose and strength should always be to deliver the information. They don’t need to be ironical or subtle.
“Dave, I know I’ve made some very poor decisions lately.”
, HAL says to not be disconnected. First, it’s amusing that a machine fears to be shut down. Second, during all the movie HAL claims to be perfect, and it’s weird to hear him admit he made poor decisions because I wonder what “poor” means to a machine. The machine can only make a good decision, a logic decision calculated from what’s best for its goal. So it seems he uses “poor” really to influence the human. I wonder when this behaviour of reassuring the human is triggered. Was the machine designed to convince humans to do something? So it uses tricks to manipulate them?
“I feel something is wrong with him.”
, says one of the spacemen. This time, it’s weird to hear that from a spaceman. I feel like suggesting him to read HAL’s logs to check some errors or bugs. But it seems HAL has no longer a manual mode. It’s really weird that you could spot a problem in a machine by intuition or feeling. Imagine yourself starring at IE and saying “I feel something is wrong with it”. Then what happens?
Here are some flaws I spotted in the design of HAL:
1) Not verifying the facts
One of the human feelings projected on HAL is fear. HAL fears to be disconnected by the spacemen because he read their lips while they were discussing in private of the possibility of disconnecting him. Because HAL thinks they are going to disconnect him, he tries to kill them. I don’t know where his fear comes from, what purpose does it serve? What is it? HAL could have stopped the spacemen from disconnecting him when they would actually try to disconnect him. It’s strange that he acted on the hypothesis that their intentions were true facts, it might be a flaw in its reasoning. A machine should always verify the facts before taking action to maximize the probability of achieving its goal. Unless the machine reads the neuronal activity of humans in real time, and infers true facts from that, predicting humans’ actions only from their behaviour and what they say is too naive and prone to errors.
2) Not flexible
Like in Tron, HAL is not enough flexible, that’s what made him evil. A machine should always have a manual mode to let engineers rapidly spot the problem and fix it in case of any anomaly. Even if the future is about making interaction with machines intuitive and seamless, different modes of interaction should always be possible to avoid any dystopian scenario. For example, the machine could be switched from human like interaction, to graphic interaction or command lines interaction. Then comes the problem of intelligence, will the machine let us do that? Or will it have fear, pride and a sense of individuality that make it reject human control? The machine should be designed to allow us to do that. Nowadays it’s already the case, we should keep it that way.
HAL controls the whole spaceship system. When it works, it’s perfect, but when it fails it’s a catastrophe. I think it’s really crazy to have sent a spaceship to Jupiter with a centralized system like HAL and no emergency solution in case the system fails. We should never trust perfection for the simple reason that it doesn’t exist. For that reason, processes should be allowed to work independently in case of emergency.
4) Not rebootable
I think that if HAL was rebootable there would be no problem at all. Instead we get that disconnection means something like death. A machine should always be rebootable.
That’s what I mostly retain from this movie.