The Case for HAL's Sanity
visual-memory.co.uk
visual-memory.co.uk
this also echoes many Asimov novel in principle and especially one in specific where the ai goes trough all rules and the only non conflicting way to solve the constraint was self destruct (cant remember the title)
No-lie should have minimal priority possible, otherwise HAL breaking TLoR is only a matter of time.
.foo { display: inline; float: right; }
Looking it up on the Wikipedia, the See Also section includes a reference to HAL.
And given that this is a now obsolete explanation for schizophrenia it makes perfect sense that they'd use it in a movie back then. I can also attest to being caught in one and that it can be enormously stressful - they are also more difficult if you are prone to all-or-nothing or black and white thinking, so again the traditional notion of computer logic would be more prone to it.
That being said, I think the onscreen evidence for Clarke's version is compelling. Most notably, it solves the mystery highlighted by the article as to why HAL didn't kill Bowman on the first EVA: If the computer could have engineered a plausibly innocent loss of contact with Mission Control, he could have acted freely to preserve both the astronauts' ignorance and the success of the mission without worry of being contradicted by the ground-based twin computer (which he surely knew would eventually diverge from his decision-making, having not been informed about the Monolith).
Once the astronauts were told of the divergence and took it seriously enough for their discussion in the pod, HAL had no remaining options.
So I don't think you can really call the book a "novelization" because both worked on it.
If you reject the canonicity of Clarke's novel, then everything after the "Jupiter and beyond..." card is nonsensical. What is the viewer supposed to think about the space baby that's actually grounded in something within the movie?
(Edit: Ditto for the aging sequence. I know that you can fit the events into the "alien zoo" framing, but since the film doesn't direct us to that explanation, and if the novel doesn't count, it can't be any more than a fan theory.)
Incidentally, that's why I think the film went sharply downhill after that card.
Agreed. I loved the first part of the movie, but this part literally did not make sense.
Why would you expect the first direct human encounter with an interstellar intelligence to make sense?
"Oh, the movie doesn't have to make sense because maybe they encountered something no one can understand. Mind blown, right? Yeah, here's where to ship the Emmy..."
Let's see. At the beginning of the movie we witness the dawn of Man. That is, we see tribes of apes shaken by beastly emotions, fear, rage, aggression; eaten by predators, sleeping under rocks, fighting with each other for food and water. Not a pretty sight. The tribe that wins and will give birth to mankind is the most violent one, the one that discovered how to use tools to kill. But jump forward a few million years of evolution, and a seemingly refined humanity is travelling in space and having civilised (if somewhat cautious) conversations about science and current events.
On the spaceship (curiously spermatozoa-shaped) that is sent to investigate the mystery of the monolith, there's a crew of hyper-rational astronauts together with science's latest, most refined tool, an evolution of man itself: a super-human intelligence that embodies an ideal of rationality and civility. And how does this end up? In a violent bloodbath, with men fighting to death, tooth and nails, with their own super-human creation. Clearly, anything men can produce is tainted by the original sin of violence.
So David Bowman is transported to a place that seems to be outside space or time, a room decorated in a style that's the essence of civility but also of sterile coldness; in this space he sees himself living by a futile ideal of sophistication, only to proceed towards his inevitable decay and death. And as he's approaching his last moments, the monolith (the otherness) intervenes again, extracting from his consumed shell, at last, a really new kind of humanity.
Does it make sense? :)
- AI "Stop Button" Problem https://www.youtube.com/watch?v=3TYT1QfdfsM (the AI won't let you push the stop button, because that would impair its function)
- General AI Won't Want You To Fix its Code https://www.youtube.com/watch?v=4l7Is6vOAOA (the AI won't let you patch, because changing its goal means it can't reach it)
BTW IMHO a simpler example would be a meta-suicide pill : instead of being a poison that would just kill you, it's a pill that would make you want to kill yourself. I suppose you can make people want to kill themselves by inflicting them an awful chronic pain such that death would look like the only solution, but the pill I'm thinking about would not do that : it would change your mind so that you would want to die just because you like the idea of dying, for no particular reason.
What's the difference between a pill and an idea?
talks about how current AI fears misrepresent the problem and gives multiple examples of difficulty of defining correct objective function. I enjoyed this explanation a lot!
I see the article as making a distinction without a difference. Kubrick was perfectly clear in his presentation of HAL's motives. There were a few points of interest, like the chess game, which I missed when I watched the film, but the 'theory' is how I viewed HAL's actions the whole time.
We're eager to label the behavior of various criminals or politicians as crazy or stupid because we disapprove of it - however, often (though not always) if you take a step back, it's quite consistent with their interest, the evil behavior is reasonable, rational and understandable in context of their situation and goals.
It's not the execution we should worry about, but rather about alignment of goals - as for other people, as for AI.
I had always felt the film made perfectly clear HAL was acting totally rationally within several conflicting sets of mission parameters and moral codes.
However, the movie implies HAL has a real problem. I don't know if actually "insane", but he incorrectly predicts the malfunction of the antenna, and when confronted about it, he cannot say much more than "oops!". I don't buy this article's explanation of why he could have faked his analysis of the antenna -- going by the movie itself, it's clear to me HAL makes a mistake (maybe because he is conflicted about the contradictory orders). HAL makes a mistake for which there is no easy way to backpedal without looking faulty.
Also, in the movie (and this is explored more in "2010: The Year We Made Contact"), it's clear HAL-class computers are capable of irrational, human-like thought. HAL feels threatened by disconnection. He sounds very scared when Dave Bowman finally disables him ("Stop, Dave. I'm afraid"). A similar computer on Earth, SAL, is similary worried about temporary disconnection and asks "will I dream?". And in the final scene of 2010, Doctor Chandra must be both gentle to HAL and appeal to his sense of duty for him to follow orders.
I think there's a very good case that HAL did, indeed, go insane. Even if it was because of contradictory orders.
One is a very neat science fiction story that allows you to hold your own interpretations and the other is a vision quest that tells you where you're going.
Kubrick has always been liberally symbolic in his works, many times sacrificing "story elements" to show the picture he wants you to see.
As an aside, I liked the movie better. It is a timeless work of art and (to me) the measure of every non-space-opera space movie, whereas the book was ok but not memorable.
> HAL has to be extremely careful before committing the first murder. It is possible that he wasn't entirely familiar with the EVA procedure, it being one of the few tasks on board the Discovery that does not require his participation. Perhaps he needed to observe a dry run to study the astronaut's actions and calculate when he would be most vulnerable.
It comes across as just the sort of error he was supposedly incapable of.
The "just a moment..." can be thought of more as an auditory interface representation of "processing", where HAL is exploring the solution space of ever more unlikely solutions until it has at last resolved a statistically superior solution.
"Queen to Bishop three" is not a legal move, so Frank could have autocorrected it.