07:00 "On the Fact that the Atlantic Has Two Sides" Kay: "On the Fact that Most of the World's Software Is Written On One Side Of The Atlantic!" 07:30 Two types of math: computers = practical math structures of a much larger kind, consistent 08:55 Debugging goes on in math as well 10:00 What's being done out in the world under the name of "OOP"? 10:33 Infamous C++ quip 10:50 Same feelings about Smalltalk! Meant to be improved. Not syntax, or classes. Taken as given... 13:00 Math as gears 13:35 Analogy to 60s programs: SCALING A DOGHOUSE! 18:00 Pink plane / blue plane 21:30 Data abstraction in 1961 US Air Force 23:25 Contrast that with what you have to do with HTML on the internet! Dark Ages! Presupposes that there should be a "browser" that should understand its format. "This has to be one of the worst ideas since MS-DOS" What happens when physicists decide to play with computers :P Browser wars, which are 100% irrelevant [to the big picture] 25:30 The internet is starting to move in that direction as people invent ever more complex HTML formats, ever more intractable. This same mistake made over and over again 26:00 Sketchpad: very much an object-oriented system 26:50 Simula 27:50 Better Old Thing vs. New Thing 29:20 Molecular biology: cell physiology and embryology (morphogenesis) 30:00 First assay of an entire living creature: E.Coli; biological info processing 34:10 Only takes 50 (40) cell division [cycles?] to make a baby! 35:20 Computers: slow, small, stupid. How can we get them to realise their destiny? 36:00 Doghouses, clocks don't scale 100x very well. Cells scale by 1,000,000,000,000x. How do they do it, how might we adapt this for building complex systems? 36:40 Simple idea that C++ has not figured out. "No idea so simple and powerful that you can't get zillions of people to misunderstand it" Must not allow the INTERIOR to be a factor in the computation of the whole Cell membrane keeps stuff OUT as much as it keeps stuff IN Confusion with objects from noun+verb-centered language. "Our process words stink" 37:50 Apologies for "Object-Oriented". Should have been Process/Message/間 (ma) -Oriented Stuff IN-BETWEEN objects. What you DON'T see. 39:25 An object can act like anything. You have encapsulated a COMPUTER!! The universal simulator! Take the powerful thing you're working on and not LOSE it by PARTITIONING UP your design space! That's the bug in data/procedures. Pernicious thing in C++ and Java by looking like the OLD THING as much as possible. 40:35 Virtual Machine 41:00 UNIX processes: like objects, but too much overhead. "3" ought to be an object; but it can't be a UNIX process. 41:50 Bio-cells can't share their DNA. OUR "cells" can! No need to e.g. engineer a special virus to change the DNA (e.g. cystic fibrosis) 43:00 "An object is a virtual server" Every object should at least have a "URL" I believe every object should have an IP 44:10 So far we've been CONSTRUCTING software; soon we'll have to GROW it Easy to grow a baby 6 inches; never have to take it down for maintenance!! Can't grow a Boeing 747! Simple mechanical world; only object was to MAKE the artefact, NOT to grow it 44:55 How many people STILL[1997] use a dev system that forces you to develop OUTSIDE of the lang? Edit/compile/run? Even if it is "fast". Cannot possibly be other than a DEAD END for building complex systems! [20 years later, in 2017: has anything changed?] 45:50 ARPANET, the precursor to the Internet. From the time ARPANET started running, expanded by 8 orders of magnitude. Not ONE PHYSICAL ATOM in the internet today that was in the original ARPANET! Not ONE LINE OF CODE of the original remains in the system! System expanded by 100,000,000x, has changed every atom and every bit and has NEVER HAD TO STOP! That is the metaphor we ABSOLUTELY MUST apply to what we THINK are "smaller" things! When we think programming is "small", that's why our programs are so BIG. 47:45 LISP, meta-reflection. "Maxwell's Equations" of software Saddest thing about Java (possibly before reflection API?) Java originally for programming toasters, not internet, but still -- How can you hope to cope with stuff without a meta-system? Represents a real failure of people to understand what the larger picture is / will be. 50:10 Meta-programming. Bootstrapping on top of the language itself. Tyranny of a single implementation. 52:15 Dozens and dozens of different object systems, all with very similar semantics, but different pragmatic details. Think of what a URL is, a HTTP message is, an object is, an obj-oriented pointer is... Should be clear that objects should not have to be local to the computer. Interop is possible basically for free under this stance. Things like JavaBeans and CORBA will not suffice. Objects will have to DISCOVER and NEGOTIATE what each other can do. Prod and poke each other to see response. Must automate this!! 55:20 1970s: Abstract Data Types. Assignment-centered thinking. C++ like MS-DOS: no-one took seriously, because who would ever fall for a joke like that? ... 57:35 McLuhan: "I don't know who discovered water, but it wasn't a fish!" 58:10 3 stages of new ideas: Dismissed as mad. Then: totally obvious. Finally: original denouncers claim to have invented it! Smalltalk, sadly, went this way. We don't know how to make systems yet! Don't let what we DON'T know become a religion! Before coming out into the world, Smalltalk was very good at making obsolete previous versions of itself. Squeak: bootstrap something better than Smalltalk! Think of how you can obsolete the damn thing! 1:01:30 Pipe organ: "Play it grand." Let us combine all this with Bret Victor's work, and lay the foundations for the *real* computer revolution yet to come.

“After growing wildly for years, the field of computing seems to be reaching its infancy.” -- John R. Pierce

The collection of sick burns Alan Kay lays on the Computing field never get old and is still just as relevant. I love this talk and so many of his others

An absolute visionary, and seems one of the more exceptional and charismatic Turing Award winners. I am going through many Allan Kay videos with my brother. His "context" idea to get leverage from other domains to address issues of scalability is truly a great concept and seems to be far more powerful in scalability as the more conventional engineering metaphors of building bridges. It seems in fact to me an iconic example of Bacteria for the loose coupling minimal cohesion principle. Cheers, K

Thank you a lot for sharing this!!! For people being far away from people like Alan and surrounded by very different kind of people these interviews and thoughts of Alan Kay is like a fresh air and really enriching. Thank you very much. Having possibility to listen people like this really makes a difference and .I am really grateful to you for your efforts to make it available.

thanks for keeping this alive, great talk.

Thanks for being an unofficial archivist. This talk still applies today despite being more 15 years old. Back then we had PCs and browsers. Today, we have smartphones, browsers, and e-book readers... all using the same obsolete ideas and architectures.

21:03 basically explains most of the problems in modern computer industry, people learn one language, and then just mess around in that language optimising their use of that language but never expanding their ideas into generally new (to them) concepts. We end up with knowledge being slowly lost and a constant pathetic reinvention of the wheel

"I made up the term object-oriented, and I can tell you I did not have C++ in mind"

"Play it grand." That's great advice to end a great talk.

the book he showed at 54min is The Art of the Metaobject Protocol by Gregor Kiczales

Best Alan Kay quote @10:34"I made up the term object-oriented, and I can tell you I did not have C++ in mind." Reminds me of: "Those who forgot the past, are condemned to reinvent it, badly."

Thank you for this. I realise I graduated with a CS degree right after this talk happened and I've been stuck in the old "lone hacker" meme never really embracing message oriented/oo. A light has been lit for me and now I'm looking for any reading material and other media I can get my hands on.

This man is so genius that now in 2018 looks like he speaks about a future, PS: i'm anxious to learn Smalltalk now :D

oh my god, the music at the beginning is amazing

Well it's a very good talk overall, despite quite some flippant side-comments, because it makes one thinking. And that's what good talks are about. For me this talk wasn't about programming languages or paradigms. It was about two things: a) what does a computer do and b) present a compositional view of computing to recreate complex systems. Programming is just the way to make a computer do anything from a techical perspective. I feel like the talk adds a lot of noise over programming languages and operating systems and perhaps elicites some lower level disagreements present during that time. Let's start my thinking with the critique on the notion of object orientation. If object oriented programming wasn't supposed to mean what it has become, and instead an object is like a server, then by all means why connect it with the term programming? Building servers, setting them up, running them, tearing them down, replicating them, duplicating them is not programming. At least we have not come to think of it as programming. Maybe that is more obvious from a hindsight perspective than in 1997. The configuration and operation of thousands of servers is not typically called programming. What we have today in form of the internet is a very large, very complex and very heterogeneous beast that has grown a lot more than it has been designed. But sure, there is design and there are protocols and standards and such. And as the talk suggests it wasn't compiled and shipped in its whole form. But nevertheless, despite it's complexity it is still the cog-wheel machine and not the living and transforming biological entity. More on that later. Web applications are often that, servers that are setup somewhere and that perform some task. If you will that's like a cell. It doesn't matter what programming language you use for describing how that cell works, because nobody on the outside will perceive it. Programming languages are for inner cell workings and not for describing how the whole organism works. Putting up a new server, tearing down the server IS the change of a system while it's running. While it's a laudable move to embed within a programming language (Smalltalk) this notion of growing, running and evolving, it still remains an egineer's perspective for me. Now, the intriguing thought of the complexity of biological systems and the comparison to computational or information systems. In my opinion, the comparison is not fair. It still requires a human programmer to change and grow and run "compuinformational" system. Living creatures don't grow by a third party that adds stuff to it (remember the fish that hasn't discovered water). To have a computational system that grows is one where you have to take the human out of the loop in which the fundamental question remains of what all this computing is then supposed to do? Because, nothing what computers do is real. The reality of a video is a careful alignment of atoms in some specific places which can, by some electromechanical machine, be interpreted as a stream of bytes, which can, by some virtual machine, be interpreted as a series of visual and audible frequencies. That's what we do with software. The only purpose computers serve for us is in forecasting the future physical universe to adapt (to) the real physical uinverse now and second, to organize, exchange, collaborate and coordinate or synchronize human activity. Nature just doesn't have a purpose (remember the fish that hasn't discovered water), which is why have bacteria and humans. But computers and computational/informational systems are built for a purpose and it's a very specific purpose and it's _only for humans_. Animals have no use for computers, plants have no use for computers, the universe has no use for computers. Only humans care about what computers do. And they do basically two things as stated above a) compute some model of existing or hypothetical physical realities or b) communicate and exchange information and instructions, collaborate, synchronize, coordinate, control. I say this, because there's parts of this talk on how babies grow with 40 steps of cell division and of bacteria and organisms and life and such. The discussion of how a bacteria is a phenomenal biological entity and all that is very intriguing, but at the same time, what would we do in computing it? We'd simulate it, in order to forecast and understand it, to dissect and analyze it and then to finally build a machine that does the same thing as the bacteria. And then use that machine for whatever good or bad purpose. Computing is a cog-wheel in that process. There's no universal grandness, all we do is building better and more complex machines. Regarding the construction of software vs growing it. I think that deserves some additional commentary. We have abstraction and composition as too fundamentally different approaches. The first may be related to a top-down view, the second may be related to a bottom-up view. Programming languages are mostly perceived as abstractions, but of course their products can also be composed to form more complex systems. Alan suggests that Smalltalk fundamentally differs from these in that, yes it's a programming language, but also more like it's a system already. I am not so convinced of this. I think the level of composition realized within a lanuage is too low. If you think about a database server and an application then these two interact, yes, but do they need to be in the same language to be composed? Why? Also performance... sometimes it doesn't matter, sometimes it does matter a lot. The one language to rule them all is simply never going to work. I think that's where a lot of the noise and flippant commentary in this talk comes from. So what's the essence for me? I suppose what the talk reveals is that a compositional approach is a viable path towards complex systems. Let's view an abstraction like a projection to a low-dimensional space. So with abstraction you take an object like a doghouse and you abstract it by projecting into the low-dimensional space (or perhaps even just the visible space that you perceive the doghouse to be in). Then you would scale that low dimensional space and then un-abstract to perceive the resulting higher-dimensional reality (I am thinking here behavioural dimensions, social, economic, etc.). And that would not work. First, because you may not even be aware of all the dimensions that went into an abstraction, because they were so small from the object that you started the projection with, e.g, think of the social dimension of a single person organization. And second, because it requires to invent the expressions in the higher dimensions (that you not even know all of as stated just before). So then composition and there the talk is about growth. And that's where it gets a bit convoluted, because abstraction scaling is associated with programming and computing (doghouse to building) and compositional scaling is associated with biological organisms. However, reality is that compositional scaling is just natural for computing and programming too, regardless of the programming language. So, for me the comparison of a cell to a bacteria and a doghouse to a building shows that the first two are functionally different and requires composition and the other is functionally similar, but nevertheless requires a lot of invention and new things. But system development doesn't necessarily be only just one of those. A system can be composed of software components, e.g. database, e.g. web server, etc. and it can be of various abstractions, e.g. the database system itself that have to scale (single table vs multiple tables and complex joins).

He attacks Dijkstra for arrogance but fails to perceive it in his own exceptionalism, that there's something so special about his discipline that it can't be described by known mathematics.

1997 seems like it was 1950...

Really your diligence is applaudable.

His description of the ossification and canonization of smalltalk (around 59:00) reminds me of the following Nietzsche quote: "As long as a man knows very well the strength and weaknesses of his teaching, his art, his religion, its power is still slight. The pupil and apostle who, blinded by the authority of the master and by the piety he feels toward him, pays no attention to the weaknesses of a teaching, a religion, and soon usually has for that reason more power than the master. The influence of a man has never yet grown great without his blind pupils. To help a perception to achieve victory often means merely to unite it with stupidity so intimately that the weight of the latter also enforces the victory of the former."

39:43 "What you have encapsulated is a computer" The most important idea in the last 50 years of computer science. Objects are abstract computers. You decouple the notion of a "computer" from the physical hardware on your desk. The paradigm holds from a single object representing "1" right up to the internet (an object orientated network). Its objects all the way up and all the way, which is just another way of saying that it is "computers" all the way up and all the way down. A desktop PC can hold a thousand "computers" just as the internet can hold a million computers (each holding a millions computers). This is the most misunderstood idea in computer science right now, and the lack of understanding is killing the industry. It is much easier for some to simply think of "1" not as a computer representing the number one, but simply as data, decoupled from any computing ability. The computation happens some where else, "1" is just one. And also the performance in the 70s of representing an small thing like one as a computer was considered unnecessary waste. But we are now stuck with systems that cannot scale, and despite the millions of times faster our PCs are we are hitting this brick wall of the limits of procedural programming.

Around 52:24 holy shit that sounds like the json apis that are all over the place these days

html is dark ages, presupposes browser understands everything

Great job on the timecoding by the way

Thank you so much for sharing!

Yes! Thank you!

18:31 his 2015 talk on "Simplicity" has lots of elements common with this one talk.

Here we are 25 years later. Has the biological metaphor won out yet?

The header of every HTML page should contain the source code for the web browser that can render it.

Perhaps it is not objects saying what they can do but a soup of events and objects that can do something with/about those events.

Thank you.

thank you for this video

Brilliant

Set playback speed to 1.5x or 1.25x.

I just realised that Google's new OS under development is called Fuschia, Apple's attempted OS was called Pink. Wow

The notion of the 'tyranny of a single implementation' obtains in all domains. Regarding human language, avoiding the tyranny of a single implementation is a strong argument in support of learning Esperanto - even as merely an object of private study.

11:40 OOP произносит "УП"

25 - can't agree with that. Mixing representation and interpretation of content makes sense only if there's only one sensible one of the latter. In particular, mixing in code only makes sense if the author of that code is trustworthy from the consumers perspective, the code they're writing will work on the users machine (the second you're trying to make it machine independent you're starting to expect a browser) and either there's only only one possible interpretation (so no mobile devices, screen readers, automated processing etc) or there are many and you're expecting the original encoder to account for all of them... Including the ones that don't exist yet somehow. This doesn't even go into any of the hard problems browsers solve; yes that makes them impenetrable monoliths, but the reward is that clients can get away with just worrying about much simpler application problems rather than, say, text rendering (which before modern browsers basically no software did correctly, not that even they aren't still struggling with it).

Thanks for sharing. But are OOPLAs complying to this today? And which of them?

I wonder if the man had something to smoke before he gave this talk.

So, he meant microservice architecture by OOP but misunderstood?

23:48 HTML on the Internet has gone back to the dark ages, because it presuppouses that there should be a browser that should understand it's formats. This has to be the worst idea since ms-dos.

7:43 > _"practical math"_ yes, i have always called computer science as applied maths. wow.

And then Sun created EJB....

do you have transcript? since i think i need one

Okay, so what is a _"Texas Tent-Beating"_ ? 45:25 Google doesn't seem to know... so either it's very obscure or, more likely, my ears no longer work properly : /

51:45 "Microsoft won't be able to capture the internet", this is the best prediction, they failed in mobile market, they failed in browser market, search engine, services etc, only apple and google were able to do it.

Atrthur koesler 18:28

His message is good, his expertise is fantastic, but his speaking style is difficult to follow. Reading a transcript of this is so much easier.

23:30

It's funny how he says that Dijkstra is arrogant, but he disses browsers and HTML. Who's laughing now? :D Btw, he's a genius man really.

rewrite the art of the metaobject protocall in a general message passing programming sense

Ok, we send you the data and also the program code how to process the data. This is really optimal especially when I want to send you a virus. This shows that this speech is from a more innocent time.

A possible bright side: if humanity implemented what is being communicated here, we might have destroyed ourselves because our computers and software would be much more powerful; perhaps too powerful. The self-limiting nature of our conceptions may be some kind of built-in fail-safe survival mechanism.

powerful oo? skip java but see www.prevayler.org and how it can work in lisp or smaltalk or haskell, process oriented term for it! not object oriented, but message passing?

Jeff Gonis Please pin "Joel Jakubovic " comment b/c he has all the timestamps GTG!

Something fishy about this guy

I'm not a huge fan of C++ either

1501 operating systems BTFO

Too bad culture, and market systems drive tech instead of innovation in the west. Maybe that's why the best coders are from Austria or eastern Bloc former coldwar countries. Though I'm not sure pointers was a good thing

'90s web... seriously ?

"On the Fact that Most of the World's Software Is Written On One Side Of The Atlantic" Great talk, but that sentence is stupid. Great software is written everywhere, but big money is just on one side of the Atlantic.

Is it just me or does he talks like Donald Trump ?

so why doesnt kay show a broswer that does not do anything much but offer bits of memory and output grpahics and show us how it looks? can it beat icewm on freebsd and firefox I am using? I think gopherspace is kinda cool idea and a network drive idea is cool and all browser doe sis read something lieka file or set of files and show you something and then accpt typing or clicking inputs.....

39:55 hurting the programmer!!