“Computers Are Useless. They Can Only Give You Answers” (2011)
quoteinvestigator.com
quoteinvestigator.com
Questions and answers are not a one-way system of thinking. Who hasn't thought of more questions after being given an answer?
For example, statistics require a computer to calculate. They answer questions like "what is the unemployment rate?". Once you get a number, say a time series, you are going to ask "what affects the unemployment rate", essentially asking what the generating function is. This in turn leads to more and more calculations, a whole web causality.
So computers are not providing questions any less than answers.
However, I think you need also to understand Picasso's answer in the context of his art and what computers were until the 70's.
Picasso was one of the most radical language innovators in art history. Every couple of years he would radically break up with the language and style he mastered in the previous years. He was an obsessive revolutionary.
On the other hand, until the late 70's computers were not really creative tools for artists. True, they were for engineers and programmers. But only after the introduction of the Mac, Amiga, Windows, NeXT, etc they'd be used for visual arts, design, CG, music, architecture, virtual reality, gaming, etc.
So, today, besides of providing answers that lead to other questions (as you mentioned), they now can be used to create real art in the Picasso sense: as new ways to perceive reality and to express it.
Edit: "One-sided" is probably ambiguous. What I mean: You don't ask the computer why it responded the way it did to your initial question, at least in any meaningful sense. You understand, abstractly, that the organization of instructions in the computer's program interacted with the input to produce an output. You might ask yourself why the response is something unexpected, but it seems strange to ask a state machine for a reason in the same way that we might ask a child to justify their dislike of brussels sprout.
But the computer can be coded to do causal calculus. Which is an elegant way for it to reason about the observations it sees. The data tells the machine what is a possible graph of causality, and what is ruled out. Adding more data changes the graph. Now you could say the machine can't dream up what to try to measure that's not already there, but I don't see why it's terribly different from what a person does; a person just happens to have a very wide range of experiences to draw from.
About the child and the brussel sprout, it's because they don't like the taste. They don't like the taste because they're evolved to like certain tastes, where the term "evolved" hides a complex history of genetic interaction with the environment, and "taste" hides parameters discovered during this interaction. Essentially, the child has certain priors in his head about what's good for him. Not all that different from a calculating machine.
Now the child may not understand this, because he can't consciously examine the chemicals in his brain, but the computer is built in such a way that we can, so we tend to think the computer is somehow less complex.
I'm not convinced that number-crunching could be fairly cast into the same pattern of rationality as reasoning. For one, reasoning seems to imply the potential for a multitude of (not necessarily correct or even veridical) outcomes - it is possible to reason as a human being about what to eat for dinner without predicating that on an "acceptable" meal.
In contrast, there is only one (and necessarily veridical) outcome for the sort of "reasoning" that a computer does. You can make that outcome very complex by virtue of a very complex program, but any deviation from it would be something wrong with the machine or program, not just another outcome.
> About the child and the brussel sprout, it's because they don't like the taste.
Possibly. It could also be that they've never had them but have heard their friends complaining about them, or that they just don't like the way they look.
What's interesting is that any of these would be a reasonable response from a child (or an adult). Contrast that to the explanations we would accept for a machine failing to produce the correct outcome - would it be reasonable to say (beyond in a colloquial sense) that my desktop "didn't like" the input I gave it?
Why? Say a human is driving a car, and a kid steps out on the road, you can swerve, or you can stop. Same goes with a computer. Except with a computer, you can have all the stats about the likely outcomes to hand immediately and choose whichever is marginally better. It's only the right choice to the extent that the data says one choice is superior to the other. You can easily imagine that if the choices are close to each other, the machine would just pick one of them. Or even do whatever improves the data gathering, eg if one of the choices had been picked a lot less that the other.
In general there's scope for optimization when you have data and calculation power, but there's also scope for satisficing, ie doing any of several things that will give an acceptable outcome.
>>would it be reasonable to say (beyond in a colloquial sense) that my desktop "didn't like" the input I gave it?
Yes. The only difference between the kid and the computer is the kid evolved naturally, and one of the things he got from evolution is self-interest. Of course a machine can have self-interest as well, but we don't use the language of preferences to describe it.
In each case, it's the researcher that asks the question. They use a computer to calculate an answer, but the next question doesn't jump out of the computer -- it comes from the researcher's mind once they see the results from the computer.
They're strictly answer calculating machines. Not question generating machines.
I would like to see a computer pose an interesting, non-trival question about physics, mathematics, cosmology, logic, etc. Something that would be accepted as a master's or PhD thesis.
My 2 cents on this quote:
1. This quote is about art. Not about computers. Picasso was an artist (and a great salesman, PR-man, self-promoter btw). But he didn't know shit about computers or "calculating machines" or whatever octogenarians called them "mechanical brains" back then. This quote is about art.
2. The quote seems to be widely misunderstood by a surprisingly large percentage of people who don't get it. Firstly, it's just a quote. I'm not sure why it gets as much attention as it does, but everyone needs to chill. I feel maybe people get offended quickly by it or get caught up in who said it and then everyone forgets to read it closely...Here goes:
The quote is illogical. The quote makes no sense. This is prima facie. Something cannot be "useless" and have a use. Even if a thing has one single use, it is not and cannot be called, "useless". Assuming computers only give you answers, that would be a use and so they cannot be called "useless". e.g. can you say "Thermometers are useless, they only tell you temperature" or "Easels are useless, they only hold the paintings"? Hopefully you see illogical structure and fallacy of what he is saying.
Funny how a computer would not accept the logic of the quote. But I don't think that was intentional.
Next: Without going into to much context on Picasso's view on art and technology (which he was not against), I think one way to understand the quote is to flip it: What is the point of art? A question asked by many for a long time..."Art is useless. It just gives you questions".
New knowledge comes from asking novel questions.
I haven't yet seen a computer pose an interest question, something no one has ever thought of yet. I don't mean questions about art, or meaning, or "What happened in 1750?" but novel questions in mathematics, cosmology, physics, etc.
They do seem to be answer-calculating machines.
>Something cannot be "useless" and have a use. Even if a thing has one single use, it is not and cannot be called, "useless".
Thats not true; if its only uses are themselves "useless", then the object is also useless.
To be not useless, it implies useful, in which case its use has value. Not that it has some use whatsoever (since anything can be given a use, the term useless would have no meaning)
The purpose of computing is insight, not numbers.