Aeon.co has posted an article I worked on for a couple of years. It’s only 2,200 words, but they were hard words to find because the ideas were, and are, hard for me. I have little sense of whether I got either the words or the ideas right.

The article argues, roughly, that the sorts of generalizations that machine learning models embody are very different from the sort of generalizations the West has taken as the truths that matter. ML’s generalizations often are tied to far more specific configurations of data and thus are often not understandable by us, and often cannot be applied to particular cases except by running the ML model.

This may be leading us to locate the really real not in the eternal (as the West has traditional done) but at least as much in the fleeting patterns of dust that result from everything affecting everything else all the time and everywhere.

Three notes:

1. Nigel Warburton, the philosophy editor at Aeon, was very helpful, as was Timo Hannay in talking through the ideas, and at about a dozen other people who read drafts. None of them agreed entirely with the article.

2. Aeon for some reason deleted a crucial footnote that said that my views do not necessarily represent the views of Google, while keeping the fact that I am a part time, temporary writer-in-residence there. To be clear: My reviews do not necessarily represent Google’s.

3. My original first title for it was “Dust Rising”, but then it became “Trains, Car Wrecks, and Machine Learning’s Ontology” which i still like although I admit it that “ontology” may not be as big a draw as I think it is.

Be the first to comment »

Google has just launched a new podcast that Yannick Assogba (twitter: @tafsiri) and I put together. Yannick is a software engineer at Google PAIR where I was a writer-in-residence for two years, until mid-June. I am notably not a software engineer. Throughout the course of the nine episodes, Yannick helps me train machine learning models to play Tic Tac Toe and then a much more complex version of it. Then our models fight! (Guess who wins? Never mind.)

This is definitely not a tutorial. We’re focused on getting beneath the metaphors we usually use when talking about machine learning. In so doing, we keep coming back to the many human decisions that have to be made along the way.

So the podcast is for anyone who wants to get a more vivid sense of how ML works and the ways in which human intentions and assumptions shape each and every ML application. The podcast doesn’t require any math or programming skills.

It’s chatty and fun, and full of me getting simple things wrong. And Yannick is a fantastic teacher. I miss seeing him every day :(

All nine episodes are up now. They’re about 25 mins each. You can find them wherever you get your podcasts, so long as it carries ours.

Podcast: https://pair.withgoogle.com/thehardway/

Two-minute teaser:  https://share.transistor.fm/s/6768a641

4 Comments »

Brink has just posted a piece of mine that suggests that the Internet and machine learning have been teaching companies that our assumptions about the predictability of the future — based in turn on assumptions about the law-like and knowable nature of change — don’t hold. But those are the assumptions that have led to the relatively recent belief in the efficacy of strategy.

My article outlines some of the ways organizations are facing the future differently. And, arguably, more realistically.

Comments Off on

Digital Extremes wants to break the trend of live-service games meticulously planning years of content ahead of time using road maps…’What happens then is you don’t have a surprise and you don’t have a world that feels alive,’ [community director Rebecca] Ford says. ‘You have a product that feels like a result of an investor’s meeting 12 months ago.'”

— Steven Messner, “This Means War,” PC Gamer, Feb. 2020, p. 34

Video games have been leading indicators for almost forty years. It was back in the early 1980s that games started welcoming modders who altered the visuals, turning Castle Wolfenstein into Castle Smurfenstein, adding maps, levels, cars, weapons, and rules to game after game. Thus the games became more replayable. Thus the games became whatever users wanted to make them. Thus games — the most rule-bound of activities outside of a law court or a tea ceremony — became purposefully unpredictable.

Rebecca Ford is talking about Warframe, but what she says about planning and road maps points the way for what’s happening with business strategies overall. The Internet has not only gotten us used to an environment that is overwhelming and unpredictable, but we’ve developed approaches that let us leverage that unpredictability, from open platforms to minimum viable products to agile development.

The advantage of strategy is that it enables an organization to focus its attention and resources on a single goal. The disadvantages are that strategic planning assumes that the playing field is relatively stable, and that change general happens according to rules that we can know and apply. But that stability is a dream. Now that we have tech that lets us leverage unpredictability, we are coming to once again recognize that strategies work almost literally by squinting our eyes so tight that they’re almost closed.

Maybe games will help us open our eyes so that we do less strategizing and more playing.

Comments Off on Games without strategies

Just over twenty years ago, computer scientists were racing the clock to fix a possibly devastating error brought about by an over-estimation of the pace at which tech becomes obsolete, which is an over-estimation of the pace of change itself. It turns out that one of the two popular solutions to the problem made the same mistake. And now we’re paying for it, but mainly through some annoyances, not the sort of world-stopping calamity that the prior error threatened.

The problem twenty years ago was that software developers had with some frequency thought that storing the year could be done with two digits, so that 1970 would be saved as 70. After all, the program wouldn’t still be used in 2000! Would we also still be driving around on earth-bound cars, or giving poodles ridiculous haircuts? Ridiculous!

But, if those apps were in fact still be used as the new millennium began, then the two digits internally representing the year would be taken internally as 00, which would be likely to confuse a computer that would assume – based on the way numbers work – that 00 (2000) comes before 70 (1970). And 2001 would look like 1901, etc.

One approach developers took to preempt the Y2K (year two thousand) bug was to change the way the programs expressed date data, allotting four digits to the year. We shall call this “the right way.” But it’s more complex than it seems. For example, you may have to find every place in a complex, integrated set of programs where the date is referred to. You may have to recompile ancient code, unearthing compilers from ancient crypts guarded by three-headed dogs. It was a freaking nightmare for many organizations.

The second approach was to write a little code that looked for year dates between 00 and 20, and write an except that takes them as referring to 2000-2020. Most applications aren’t dealing with dates going back to the beginning of the 20th century, so that worked. Chris Stokel-Walter (twitter: @stokel) in his excellent, brief explainer in New Scientist, says that an estimated 80% of Y2K solutions took this approach, known as “windowing”, but which we shall refer to it as the “Please don’t do this” approach.

Well, now it’s 2020 and some indeterminate number of windowed apps haven’t updated the fix. Thus, some traffic meters have stopped working. As Chris writes, “The theory was that these windowed systems would be outmoded by the time 2020 arrived…”

So, exactly the same over-estimation of the pace of tech obsolescence has led to exactly the same problem. Surprise?

It’s not at all clear, however, who has made this mistake. The developers implementing the windowing patch were staving off an imminent, plausible crashing of globally crucial systems. Windowing was a reasonable approach to forestalling this crisis … but only if there was a system – a human system – to remember to allocate resources for fixing the problem that the patch postponed.

Conclusion: “Human system” is an oxymoron.

Comments Off on Y2K’s 1% solution

In “The Efficiency-Destroying Magic of Tidying Up“, Florent Crivello (twitter: @Altimor) has written a clear, convincing, and compact critique of the assumption that efficient organization is tidy, neat, and rectilinear. “[E]fficiency tends to look messy, and good looks tend to be inefficient.”

This is because complex systems — like laws, cities, or corporate processes — are the products of a thousand factors, each pulling in a different direction. And even if each factor is tidy taken separately, things quickly get messy when they all merge together

He applies this to management organization, the tool sets used by collaborators, city planning, science fiction visions of the future, parenting, and pizzas.

Not to mention that in one brief, beautiful essay he unites the themes of two of my books: Everything Is Miscellaneous and my new Everyday Chaos.

Comments Off on The inefficiency of order. The efficiency (and beauty) of chaos

I just posted a new page at the Everyday Chaos web site. It lists media coverage, talks, and other ways into the book.

Take a look!

Comments Off on Everyday Chaos coverage, etc.

Joe McKendrick at Forbes has posted a concise and thoughtful column about
Everyday Chaos, including four rules to guide your expectations about machine learning.

It’s great to see a pre-publication post so on track about what the book says and how it applies to business.

1 Comment »