6 ms·

Alan Kay: Power of Simplicity [video]

120 points by tnorgaard 11y ago

mpweiher 11y ago
Around 9:30: "What's the expense of getting simplicity? [..] You get simplicity by finding a slightly more sophisticated building block to build your theories out of. It's when you go for a simple building block that anyone can understand with common sense that you start screwing yourself right and left."
- duncanawoods 11y ago
  I found that interesting as well. After thinking about it for a while, I struggled to get anything profound out of it. I feel his point is equivalent to saying "a higher level of abstraction can make problems more tractable" which is something we do all the time in all contexts. Finding good abstractions is an art but I don't think this really maps to insights like Copernicus though. I see a slightly different process happening i.e. discovering a new perspective on a problem. A new perspective can be a different way to decompose a problem but I don't think it necessitates anything about the sophistication of the building blocks. Instead, a revealing change is often because it changes the type of relationships between the building blocks in a radical way that enables new inferences.
  - mpweiher 11y ago
    I see where you're coming from, and in a very, very abstract sense it might be true. However, I think there are two different kinds of "abstraction" at play here. The more common is where you have a basic set of building blocks and build abstractions with them. The second is where you adjust the building blocks. That's the one that's profound...and profoundly difficult, I found. For example, I believe to have discovered that we need a slightly more complex notion of identifier/reference/variable (I call them Polymorphic Identifiers [1][2][3]). Smalltalk says that we only need messages (procedural abstraction) to be polymorphic, identifiers and things like assignment are simple/monomorphic. SELF (another "Power of Simplicity" [4]) essentially said that you don't really need separate identifiers for things at all, all the things that look like identifiers are really message-sends. This makes a lot of sense at first glance, and at second and third. However, there are subtle problems related to the Uniform Access Principle [5], and my contention is that to solve these problems you can't layer abstractions on top, you actually have to go against the grain and make one of the fundamental building blocks (variable access, both read and write) slightly more complicated. And that you actually need the concept "variable access" :-) Having worked with PIs for a bit now, it looks like this slightly more complicated primitive really does help solve/reduce a lot of complexities in real world systems. As far as I can tell, a lot of these complexities are fundamentally "storage" related, but that unifying abstraction is lost because the mechanisms we have available are procedural. Once you have it, you get building blocks similar to shell filters that you can combine to create variations as one-liners that would otherwise be many pages of code, similar to McIlroy's shell solution to Jon Bentley's word count challenge (6 lines of shell vs. Knuth's 12 pages of Pascal/Web) [6] However, communicating this appears to be nearly impossible. For example, David Ungar (of Self fame) reviewed my paper and just tore it to shreds, misunderstanding not just the solution, but also the problem, and completely incorrectly claiming that Self already solves those problems. He stopped responding when I showed him the (very simple) examples demonstrating the problem in Self. [1] http://dl.acm.org/citation.cfm?id=2508169 http://dl.acm.org/citation.cfm?id=2508169 [2] https://www.hpi.uni-potsdam.de/hirschfeld/publications/media/WeiherHirschfeld_2013_PolymorphicIdentifiersUniformResourceAccessInObjectiveSmalltalk_AcmDL.pdf https://www.hpi.uni-potsdam.de/hirschfeld/publications/media... [3] http://objective.st/URIs/ http://objective.st/URIs/ [4] http://bibliography.selflanguage.org/_static/self-power.pdf http://bibliography.selflanguage.org/_static/self-power.pdf [5] https://en.wikipedia.org/wiki/Uniform_access_principle https://en.wikipedia.org/wiki/Uniform_access_principle [6] http://www.leancrew.com/all-this/2011/12/more-shell-less-egg/ http://www.leancrew.com/all-this/2011/12/more-shell-less-egg...
  - calibraxis 11y ago
    Don’t know if this applies to your post, but I got a lot more out of Alan Kay talks when I stopped trying to look for profundity, and assumed he's mostly talking about how we're being morons.
    - mpweiher 11y ago
      I'd say the profundity is in how we're being morons...and how we could stop if we wanted to do so. Alas, those insights aren't easy to apply.
    - wodenokoto 11y ago
      I think that too is wrong way of looking at it. What he basically says is, you need an organizational framework that works 10 years at a time in order to get a result in 7 years. The people working in this framework, needs 5 years windows and you need to hire absolute geniuses and have the trust that they'll do the right thing. If you do that, you will get long term, world changing inventions relatively cheap. I don't think anyone is being a moron for utilizing a staff of less than stellar geniuses to get work done in a time frame less than 7-10 years.
      - calibraxis 11y ago
        I believe there's a lot of evidence we're collectively morons. Consider how close we are to a transformed world. Hunger, wage slavery, environmental destruction, etc could be a thing of the past if enough of us decided right now to be serious... countering absurd ways of thinking. Are people really afraid of death? Then we should invite people to organize to solve that. Support each other regardless of gender or other insignificant differences, to accomplish such goals; any menial work can be shared equally. What huge percentage of humanity would go for that? But currently we organize ourselves in moronic ways. Alan Kay's vision seems actually quite narrow and corporate-friendly; many go further in seeking to liberate imagination. http://thebaffler.com/blog/david-graeber-defiantly-having-fun-at-the-baffler-no-22-release-party http://thebaffler.com/blog/david-graeber-defiantly-having-fu...
        11y ago
  - papaf 11y ago
    a higher level of abstraction can make problems more tractable I think of an abstraction of something that makes things simpler: https://en.wikipedia.org/wiki/Abstraction_%28computer_science%29 https://en.wikipedia.org/wiki/Abstraction_%28computer_scienc... I understood Alan Kay to be saying that a lot of abstractions are actually making things too simple. I would understand the point to mean "a lower level of abstraction can make the big picture more tractable".
  - codeshaman 11y ago
    Suppose you need to get from A to B and back. You need to do it often, you need to do it fast. Your cargo - written notes which two kings (from A and B respectively) exchange between each other. So you try to solve the problem by getting faster cars, building better roads, hiring the best drivers - all with the goal of getting the notes from king A to king B as fast as possible. --- The 'abstraction' which I think Alan is talking about is not about how to make the car go faster between A and B - it is about stepping back and actually thinking about the context - what do we actually need ? We need a way to transfer messages between A and B. And pretty soon you come up with the idea of using a membrane to convert voice into electrical signals and send those over a wire. So you install a cable between A and B and now the kings can chat all day long. That's a much simpler and powerful solution, but not to the problems you were trying to solve (making the car go faster), but to the problem in the upper scope - sending messages from A to B. This was the gist of Alan's talk imho.
staunch 11y ago
> "The big problem with Xerox is they only wanted to make billions. And that's the problem with most companies. When you're doing kind of stuff you're actually in the trillion-dollar range."
- wodenokoto 11y ago
  Yeah, he is really selling his new research division hard.
  - sp332 11y ago
    The quantity of world-changing tech that came out of the late 70's has never been matched. In fact it's hard to come up with more than a handful of innovations in tech of the same magnitude that have happened since. http://stackoverflow.com/questions/432922/significant-new-inventions-in-computing-since-1980 http://stackoverflow.com/questions/432922/significant-new-in...
    - hga 11y ago
      The World Wide Web. Yep, plenty of antecedents, I personally start with Vannevar Bush's Memex for the idea, and there were many attempts with varying levels of un-success, until Tim Berners-Lee put together a winning, massively network effects positive solution. Which, along with search engines, have resulted in significant, qualitative changes in information publishing and access; in the '70s (or perhaps earlier) Jerry Pournelle was looking forward to the day when you could quickly find the answer to any question that had an answer, and this is one of the biggest steps towards that. To the point I rank it near computers per se; from Wikipedia (https://en.wikipedia.org/wiki/General_purpose_technology https://en.wikipedia.org/wiki/General_purpose_technology): General-purpose technologies (GPTs) are technologies that can affect an entire economy (usually at a national or global level). GPTs have the potential to drastically alter societies through their impact on pre-existing economic and social structures. Examples include the steam engine, railroad, interchangeable parts, electricity, electronics, material handling, mechanization, control theory (automation), the automobile, the computer, and the Internet. This was the step that made the Internet into a "monster". So even if it only counts as one invention, it's as big as many of those preceding GPTs.

120 points by tnorgaard 11y ago

Around 9:30: "What's the expense of getting simplicity? [..] You get simplicity by finding a slightly more sophisticated building block to build your theories out of. It's when you go for a simple building block that anyone can understand with common sense that you start screwing yourself right and left."

duncanawoods 11y ago
I found that interesting as well. After thinking about it for a while, I struggled to get anything profound out of it. I feel his point is equivalent to saying "a higher level of abstraction can make problems more tractable" which is something we do all the time in all contexts. Finding good abstractions is an art but I don't think this really maps to insights like Copernicus though. I see a slightly different process happening i.e. discovering a new perspective on a problem. A new perspective can be a different way to decompose a problem but I don't think it necessitates anything about the sophistication of the building blocks. Instead, a revealing change is often because it changes the type of relationships between the building blocks in a radical way that enables new inferences.
- mpweiher 11y ago
  I see where you're coming from, and in a very, very abstract sense it might be true. However, I think there are two different kinds of "abstraction" at play here. The more common is where you have a basic set of building blocks and build abstractions with them. The second is where you adjust the building blocks. That's the one that's profound...and profoundly difficult, I found. For example, I believe to have discovered that we need a slightly more complex notion of identifier/reference/variable (I call them Polymorphic Identifiers [1][2][3]). Smalltalk says that we only need messages (procedural abstraction) to be polymorphic, identifiers and things like assignment are simple/monomorphic. SELF (another "Power of Simplicity" [4]) essentially said that you don't really need separate identifiers for things at all, all the things that look like identifiers are really message-sends. This makes a lot of sense at first glance, and at second and third. However, there are subtle problems related to the Uniform Access Principle [5], and my contention is that to solve these problems you can't layer abstractions on top, you actually have to go against the grain and make one of the fundamental building blocks (variable access, both read and write) slightly more complicated. And that you actually need the concept "variable access" :-) Having worked with PIs for a bit now, it looks like this slightly more complicated primitive really does help solve/reduce a lot of complexities in real world systems. As far as I can tell, a lot of these complexities are fundamentally "storage" related, but that unifying abstraction is lost because the mechanisms we have available are procedural. Once you have it, you get building blocks similar to shell filters that you can combine to create variations as one-liners that would otherwise be many pages of code, similar to McIlroy's shell solution to Jon Bentley's word count challenge (6 lines of shell vs. Knuth's 12 pages of Pascal/Web) [6] However, communicating this appears to be nearly impossible. For example, David Ungar (of Self fame) reviewed my paper and just tore it to shreds, misunderstanding not just the solution, but also the problem, and completely incorrectly claiming that Self already solves those problems. He stopped responding when I showed him the (very simple) examples demonstrating the problem in Self. [1] http://dl.acm.org/citation.cfm?id=2508169 http://dl.acm.org/citation.cfm?id=2508169 [2] https://www.hpi.uni-potsdam.de/hirschfeld/publications/media/WeiherHirschfeld_2013_PolymorphicIdentifiersUniformResourceAccessInObjectiveSmalltalk_AcmDL.pdf https://www.hpi.uni-potsdam.de/hirschfeld/publications/media... [3] http://objective.st/URIs/ http://objective.st/URIs/ [4] http://bibliography.selflanguage.org/_static/self-power.pdf http://bibliography.selflanguage.org/_static/self-power.pdf [5] https://en.wikipedia.org/wiki/Uniform_access_principle https://en.wikipedia.org/wiki/Uniform_access_principle [6] http://www.leancrew.com/all-this/2011/12/more-shell-less-egg/ http://www.leancrew.com/all-this/2011/12/more-shell-less-egg...
- calibraxis 11y ago
  Don’t know if this applies to your post, but I got a lot more out of Alan Kay talks when I stopped trying to look for profundity, and assumed he's mostly talking about how we're being morons.
  - mpweiher 11y ago
    I'd say the profundity is in how we're being morons...and how we could stop if we wanted to do so. Alas, those insights aren't easy to apply.
  - wodenokoto 11y ago
    I think that too is wrong way of looking at it. What he basically says is, you need an organizational framework that works 10 years at a time in order to get a result in 7 years. The people working in this framework, needs 5 years windows and you need to hire absolute geniuses and have the trust that they'll do the right thing. If you do that, you will get long term, world changing inventions relatively cheap. I don't think anyone is being a moron for utilizing a staff of less than stellar geniuses to get work done in a time frame less than 7-10 years.
    - calibraxis 11y ago
      I believe there's a lot of evidence we're collectively morons. Consider how close we are to a transformed world. Hunger, wage slavery, environmental destruction, etc could be a thing of the past if enough of us decided right now to be serious... countering absurd ways of thinking. Are people really afraid of death? Then we should invite people to organize to solve that. Support each other regardless of gender or other insignificant differences, to accomplish such goals; any menial work can be shared equally. What huge percentage of humanity would go for that? But currently we organize ourselves in moronic ways. Alan Kay's vision seems actually quite narrow and corporate-friendly; many go further in seeking to liberate imagination. http://thebaffler.com/blog/david-graeber-defiantly-having-fun-at-the-baffler-no-22-release-party http://thebaffler.com/blog/david-graeber-defiantly-having-fu...
      - 11y ago
- papaf 11y ago
  a higher level of abstraction can make problems more tractable I think of an abstraction of something that makes things simpler: https://en.wikipedia.org/wiki/Abstraction_%28computer_science%29 https://en.wikipedia.org/wiki/Abstraction_%28computer_scienc... I understood Alan Kay to be saying that a lot of abstractions are actually making things too simple. I would understand the point to mean "a lower level of abstraction can make the big picture more tractable".
- codeshaman 11y ago
  Suppose you need to get from A to B and back. You need to do it often, you need to do it fast. Your cargo - written notes which two kings (from A and B respectively) exchange between each other. So you try to solve the problem by getting faster cars, building better roads, hiring the best drivers - all with the goal of getting the notes from king A to king B as fast as possible. --- The 'abstraction' which I think Alan is talking about is not about how to make the car go faster between A and B - it is about stepping back and actually thinking about the context - what do we actually need ? We need a way to transfer messages between A and B. And pretty soon you come up with the idea of using a membrane to convert voice into electrical signals and send those over a wire. So you install a cable between A and B and now the kings can chat all day long. That's a much simpler and powerful solution, but not to the problems you were trying to solve (making the car go faster), but to the problem in the upper scope - sending messages from A to B. This was the gist of Alan's talk imho.

staunch 11y ago

> "The big problem with Xerox is they only wanted to make billions. And that's the problem with most companies. When you're doing kind of stuff you're actually in the trillion-dollar range."

wodenokoto 11y ago
Yeah, he is really selling his new research division hard.
- sp332 11y ago
  The quantity of world-changing tech that came out of the late 70's has never been matched. In fact it's hard to come up with more than a handful of innovations in tech of the same magnitude that have happened since. http://stackoverflow.com/questions/432922/significant-new-inventions-in-computing-since-1980 http://stackoverflow.com/questions/432922/significant-new-in...
  - hga 11y ago
    The World Wide Web. Yep, plenty of antecedents, I personally start with Vannevar Bush's Memex for the idea, and there were many attempts with varying levels of un-success, until Tim Berners-Lee put together a winning, massively network effects positive solution. Which, along with search engines, have resulted in significant, qualitative changes in information publishing and access; in the '70s (or perhaps earlier) Jerry Pournelle was looking forward to the day when you could quickly find the answer to any question that had an answer, and this is one of the biggest steps towards that. To the point I rank it near computers per se; from Wikipedia (https://en.wikipedia.org/wiki/General_purpose_technology https://en.wikipedia.org/wiki/General_purpose_technology): General-purpose technologies (GPTs) are technologies that can affect an entire economy (usually at a national or global level). GPTs have the potential to drastically alter societies through their impact on pre-existing economic and social structures. Examples include the steam engine, railroad, interchangeable parts, electricity, electronics, material handling, mechanization, control theory (automation), the automobile, the computer, and the Internet. This was the step that made the Internet into a "monster". So even if it only counts as one invention, it's as big as many of those preceding GPTs.