...

Stagger onward rejoicing

Tag: computing (page 1 of 1)

the irrelevance of thinking

Mary Harrington:

As I argued here, it would would be more accurate (if less snappy) to describe AI as “powerful modelling and prediction tools based on pattern recognition across very large datasets”. It is, in other words, not a type of cognition in its own right, but – to borrow a term from Marshall McLuhan – one of the “extensions of man”: specifically a means of extending cognition itself.

I don’t think this is correct; what LLMs do is not the extension of cognition but rather the simulation and commodification of the palpable products of cognition.

The people who make LLMs have little discernible interest in cognition itself. Some of them may believe that they’re interested in cognition, but what they’re really focused on is product — that is, output, what gets spat out in words or images or sounds at the conclusion of an episode of thinking.

Seeing those products, they want to simulate them so that they can commodify them: package them and serve them up in exchange for money.

This doesn’t mean that LLMs are evil, or that it’s wrong to sell products for money; only that thinking itself is irrelevant to the whole business. 

UPDATE: From a fascinating essay by Dario Amodei, the CEO of Anthropic:  

Modern generative AI systems are opaque in a way that fundamentally differs from traditional software.  If an ordinary software program does something—for example, a character in a video game says a line of dialogue, or my food delivery app allows me to tip my driver—it does those things because a human specifically programmed them in.  Generative AI is not like that at all.  When a generative AI system does something, like summarize a financial document, we have no idea, at a specific or precise level, why it makes the choices it does—why it chooses certain words over others, or why it occasionally makes a mistake despite usually being accurate.  As my friend and co-founder Chris Olah is fond of saying, generative AI systems are grown more than they are built—their internal mechanisms are “emergent” rather than directly designed.  It’s a bit like growing a plant or a bacterial colony: we set the high-level conditions that direct and shape growth, but the exact structure which emerges is unpredictable and difficult to understand or explain.  Looking inside these systems, what we see are vast matrices of billions of numbers.  These are somehow computing important cognitive tasks, but exactly how they do so isn’t obvious. 

UPDATE 2: Essays by Melanie Mitchell of the Sante Fe Institute — one and two — on what LLMS do instead of thinking. The “bag of heuristics” idea is a vivid one. 

my forever lament

Also: Every time I get to the copy-editing stage of a book project I want to write a long angry post about how much I hate Microsoft Word. But I have done it, and other people have done it. Here I am in 2016.

And here’s Charlie Stross back in 2013:

The reason I want Word to die is that until it does, it is unavoidable. I do not write novels using Microsoft Word. I use a variety of other tools, from Scrivener (a program designed for managing the structure and editing of large compound documents, which works in a manner analogous to a programmer’s integrated development environment if Word were a basic text editor) to classic text editors such as Vim. But somehow, the major publishers have been browbeaten into believing that Word is the sine qua non of document production systems. They have warped and corrupted their production workflow into using Microsoft Word .doc [.docx] files as their raw substrate, even though this is a file format ill-suited for editorial or typesetting chores. And they expect me to integrate myself into a Word-centric workflow, even though it’s an inappropriate, damaging, and laborious tool for the job. It is, quite simply, unavoidable. And worse, by its very prominence, we become blind to the possibility that our tools for document creation could be improved. It has held us back for nearly 25 years already; I hope we will find something better to take its place soon. 

(Fat chance of that.) And Louis Menand, all the way back in 2003:  

When, in the old days, you hit the wrong key on your typewriter, you got one wrong character. Strike the wrong keys in Word and you are suddenly writing in Norwegian Bokmal (Bokmal?). And you have no idea how you got there; you can spend the rest of the night trying to get out. In the end, you stop the random clicking and dragging and pulling-down and have recourse to the solution of every computer moron: with a sob of relief, you press Ctrl/Alt/Del. (What do Control and Alt mean, by the way? Does anyone still know?) A message appears: “You will lose any unsaved information in all programs that are running.” O.K.? Cancel? End task? End life? The whole reason for rebooting was that you didn’t have access to your information, so how can you save it? You can always pull the plug out of the wall. That usually ends your “session” (a term borrowed — no accident — from psychoanalysis).

I could use some psychoanalysis right about now. 

the file system

The Verge:

“I grew up when you had to have a file; you had to save it; you had to know where it was saved. There was no search function,” says Saavik Ford, a professor of astronomy at the Borough of Manhattan Community College. But among her students, “There’s not a conception that there’s a place where files live. They just search for it and bring it up.” She added, “They have a laundry basket full of laundry, and they have a robot who will fetch them every piece of clothing they want on demand.” […]

To a point, the new mindset may reflect a natural — and expected — technological progression. Plavchan recalls having similar disconnects with his own professors. “When I was a student, I’m sure there was a professor that said, ‘Oh my god, I don’t understand how this person doesn’t know how to solder a chip on a motherboard,’” he says. “This kind of generational issue has always been around.” And though directory structures exist on every computer (as well as in environments like Google Drive), today’s iterations of macOS and Windows do an excellent job of hiding them. (Your Steam games all live in a folder called “steamapps” — when was the last time you clicked on that?) Today’s virtual world is largely a searchable one; people in many modern professions have little need to interact with nested hierarchies.

“Search, don’t sort,” Google says, but increasingly I’m wondering whether I ought to spend more time sorting and less time searching. I’m always looking for ways to introduce constructive friction into my working practices — see this post for an example — and maybe going Old Skool with some kind of rational filing system would be more helpful than either (a) searching or (b) using the inconsistent and ad hoc system of folders I now have. If I designed a new system Hazel would help me implement it. 

Lamport 5

Quanta Magazine interview with Leslie Lamport:

One last thing, about another side project of yours with a sizable impact: LaTeX. I’d like to finally clear something up with the creator. Is it pronounced LAH-tekh or LAY-tekh?

Any way you want. I don’t advise spending very much time thinking about it.

zine!

Julia Evans makes really cool zines for people who want to know more about computer programming, or, more generally, about being a power user of computers. Her most recent zine is called How DNS Works, and it’s excellent — plus, there are some leftover pages about registering and maintaining your own domain. For example:

Buying domain

Other images with more detail here.

Longtime readers of this blog will know that I am a big fan of owning your online turf — or coming as close to it as you can get — for reasons I explain in detail in this essay. Evans’s zine-within-a-zine about domain registration and maintenance does a great job of explaining exactly how it works — and in the process, I hope, makes it seem less intimidating than it otherwise might.

getting a new Mac up and running

Things I do when I get a new Mac, more or less in order:

  • install Homebrew
  • use Homebrew to install pandoc
  • install BBedit
  • install MacTex
  • type this into the terminal: defaults write com.barebones.bbedit FullScreenWindowsHogScreen -bool NO
  • type this into the terminal: defaults write com.apple.dock single-app -bool true (followed by killall Dock)
  • enable Night Shift
  • install TextExpander
  • install Alfred
  • install Hazeover
  • install Hazel

Everything else can wait; once I have the above in place — plus of course syncing all my existing TextExpander snippets — I can do almost everything I really need to do on a computer, with maximum focus and speed. 

Knuth, Lutheran

This is a nice — not a great, but a nice — profile of one of my heroes, Donald Knuth, but it does have an odd little moment: 

Dr. Knuth lives in Stanford, and allowed for a Sunday visitor. That he spared an entire day was exceptional — usually his availability is “modulo nap time,” a sacred daily ritual from 1 p.m. to 4 p.m. He started early, at Palo Alto’s First Lutheran Church, where he delivered a Sunday school lesson to a standing-room-only crowd. Driving home, he got philosophical about mathematics. 

Hmmm, isn’t that interesting? Knuth is the deepest and most wide-ranging of computer scientists; plus, “many consider Dr. Knuth’s work on the TeX computer typesetting system to be the greatest contribution to typography since Gutenberg”; and he’s a Sunday-school teacher? Might it not be worth our time to explore that a little bit? Apparently not. 

But if you, unlike the NYT, wanted to explore these matters, then you might take a look at the book of calligraphy and commentary that Knuth put together called 3:16: Bible Texts Illuminated; or, if you’re really interested, listen to or read his lectures on religion and computer science, Things a Computer Scientist Rarely Talks About

operating systems & the Reformation

I wrote this some years ago in a post that has now been taken down, but with Umberto Eco’s death the topic is fresh again. Thus this excerpt: 

Have you heard the one about computer operating systems and the Reformation? You probably have. Most people got the story from the Italian semiotician and academic superstar Umberto Eco. In his telling, from an article he published in 1994, it goes like this:

The fact is that the world is divided between users of the Macintosh computer and users of MS-DOS compatible computers. I am firmly of the opinion that the Macintosh is Catholic and that DOS is Protestant. Indeed, the Macintosh is counter-reformist and has been influenced by the ratio studiorum of the Jesuits. It is cheerful, friendly, conciliatory; it tells the faithful how they must proceed step by step to reach — if not the kingdom of Heaven — the moment in which their document is printed. It is catechistic: The essence of revelation is dealt with via simple formulae and sumptuous icons. Everyone has a right to salvation.

DOS is Protestant, or even Calvinistic. It allows free interpretation of scripture, demands difficult personal decisions, imposes a subtle hermeneutics upon the user, and takes for granted the idea that not all can achieve salvation. To make the system work you need to interpret the program yourself: Far away from the baroque community of revelers, the user is closed within the loneliness of his own inner torment.

More on this interpretation in a moment, but I can’t go any further without commenting that my friend Edward Mendelson — professor at Columbia, literary executor of W.H. Auden, and occasional character in the novels of Alexander McCall Smith — made the point six years before Eco did, and just as wittily and incisively:

In the 16th century the printed book helped make possible the split between Catholics and Protestants. In the 20th century this history of tragedy and triumph is repeating itself as farce. Those who worship the Apple computer and those who put their faith in the IBM PC are equally convinced that the other camp is damned or deluded. Each cult holds in contempt the rituals and the laws of the other. Each thinks that it is itself the one hope for salvation.

Each of these cults corresponds to one of the two antagonists in the age of Reformation. In the realm of the Apple Macintosh, as in Catholic Europe, worshipers peer devoutly into screens filled with “icons.” All is sound and imagery in Appledom. . . . A central corporate headquarters decrees the form of all rites and practices. Infallible doctrine issues from one executive officer whose selection occurs in a sealed boardroom. . . .

As in Protestant Europe, by contrast, where sects divided endlessly into smaller competing sects and no church dominated any other, all is different in the fragmented world of IBM. That realm is now a chaos of conflicting norms and standards that not even IBM can hope to control. . . . When IBM recently abandoned some of its original standards and decreed new ones, many of its rivals declared a puritan allegiance to IBM’s original faith, and denounced the company as a divisive innovator. Still, the IBM world is united by its distrust of icons and imagery. IBM’s screens are designed for language, not pictures. Graven images may be tolerated by the more luxurious cults, but the true IBM faith relies on the austerity of the word.

My first thought on re-reading Mendelson’s extended metaphor, which I went around quoting for several years after it first appeared, is: How much has changed! It is growing increasingly difficult to remember that IBM was once a colossus striding the earth, and that people spoke of almost any non-Apple computer as an “IBM machine.” In 1988, the major players were hardware manufacturers.

Six years later, when Eco develops the same conceit, there is one subtle but important shift: he doesn’t speak of “IBM” but rather “MS-DOS” — not a maker of computers but an operating system. And this is the path the conflicts would take: not Apple versus IBM, but Mac (conceived more as an operating system, as a way of organizing and presenting data, than as a physical machine) versus Windows.

But in the domain of zero to one, not-something to something, Pointsman can only possess the zero and the one. He cannot, like Mexico, survive anyplace in between. Like his master I. P. Pavlov before him, he imagines the cortex of the brain as a mosaic of tiny on/off elements. Some are always in bright excitation, others darkly inhibited. The contours, bright and dark, keep changing. But each point is allowed only the two states: waking or sleep. One or zero. “Summation,” “transition,” “irradiation,” “concentration,” “reciprocal induction”—all Pavlovian brain-mechanics—assumes the presence of these bi-stable points. But to Mexico belongs the domain between zero and one—the middle Pointsman has excluded from his persuasion—the probabilities. A chance of 0.37 that, by the time he stops his count, a given square on his map will have suffered only one hit, 0.17 that it will suffer two… .

“Can’t you … tell,” Pointsman offering Mexico one of his Kyprinos Orients, which he guards in secret fag fobs sewn inside all his lab coats, “from your map here, which places would be safest to go into, safest from attack?”

“No.”

— Thomas Pynchon, Gravity’s Rainbow
css.php