Apparently, we've just invented AI

The other evening I overheard a conversation about AI and its implications, now that we have it. This morning I read a post saying that it was hard to believe that ChatGPT was released only four years ago and now AI is everywhere, and how that might be the quickest adoption of a new technology in history.

I'm sure you can find plenty of examples the story yourself: Around four years ago, we didn't have AI. Then ChatGPT happened, and we do. What is AI? It's what LLM-based systems like ChatGPT, Claude and Gemini do, especially now that these systems are agentic, that is, they can break down tasks and coordinate a collection of agents and subagents in order to complete them.

I spent some time looking for a word (and asking an AI to find one) to convey the feeling that something is completely predictable and yet still somewhat surprising and came up with droll, ironic, paradoxical and a few others, but none of them quite fit.

A quick look at the AI label on this blog (and the other one) will tell you I've been watching this space for a while. The second Intermittent Conjecture post, back in 2010, mentioned AI in the past tense. The thesis was that there were plenty of applications around that would have been considered to unquestionably be AI in the early days of AI research, but they weren't in 2010 because they just didn't have that "I'll know it when I see it" quality that would cause the general public to say: Yes, that's it, that's Artificial Intelligence.

I gave the examples of "a neural network mining some pile of data, or even a chess program [i.e., that could regularly beat the best human players], or voice-enabled phone," all of which were around by 2010. Facial recognition and machine translation of natural language were also widely available at the time, although current versions work better, as one would hope.

The mechanisms behind the current generation of AI have their roots in research into connectionism in the late 1980s, which is generally considered the second wave of research into computing based on neural networks, the first having run from the 1940s through the 1960s. The general idea of machines mimicking human behavior is ancient, but in the context of modern computers it goes back to Turing's 1950 paper Computing Machinery and Intelligence, which introduced the idea of the "imitation game," which we now call the "Turing test." Systems that people took for human and communicated with accordingly first appeared in the late 1960s and early 1970s and more have appeared in the years since.

So it's natural to have a certain sense of "Wait, now everybody thinks we have AI?" but on the other hand it doesn't seem surprising at all, because something significant did happen when ChatGPT came out. LLM-based systems are simply much better at conversing in natural language than anything that came before.

One reason is fluency. An LLM-based system can answer question after question in fluent natural language. It can spit out page after page of prose that looks for all the world like a person wrote it, not just in the syntax and vocabulary but in tone and structure. It's not hard to write a hand-coded Markov chain that will produce at least mostly grammatical sentences, but it won't really sound human and it won't be able to answer questions. LLMs are the first approach that allows an extended two-way conversation without "the man behind the curtain" appearing in short order.

The second reason is the range of topics you can converse about. The first 'L' in LLM is for large. LLMs are trained on trillions, or even hundreds of trillions of tokens (words, more or less). By comparison, the entire US Library of Congress collection comprises a few trillion words. Without digging too deeply into what the numbers actually mean, it's not unreasonable to compare the sheer volume of training data to "all books ever published." Whatever topic you might want to ask about, there's a good chance that something related to it is in the training data, in sufficient quantity that the model can produce a few coherent paragraphs about it.

Before ChatGPT there was nothing widely available -- and hardly anything available at all -- that could converse with an average person in human language and appear knowledgeable on a wide range of topics. Artificial systems were able, and had long been able, to do a variety of tasks that can reasonably said to require some form of intelligence, but until ChatGPT everything was in the category of "Yeah, that's interesting, and even useful, but it's clearly not really intelligence." The particular task of conversing with a human on an open-ended variety of topics was the first AI task that really fit the "I'll know it when I see it" bill.

All this is fine with me. In everyday usage, LLMs are AI and nothing before them was. Common usage makes sense in terms of people's expectations of what artificial intelligence ought to mean. Everyday usage doesn't have to match up with technical definitions. 

Still, I'm uneasy with a couple of aspects of all this. First is the idea that AI is one single thing, whether LLMs or something else. Just as intelligence comprises a number of abilities, artificial intelligence is many different things, each capturing some portion of what we call intelligence in living things.

Tying into this is the idea that artificial intelligence necessarily means artificial general intelligence (AGI) or superhuman intelligence, or even the creation of a new, sentient species. Yes, ChatGPT crossed a threshold into something that people could generally agree could be called AI, but only in one particular sense. It did not cross a threshold into everything that AI could possibly mean.

Fortunately, after spending time actually conversing with the things and observing their behavior, people seem to have generally concluded that, even though LLMs are AI, they're not all-powerful or superhuman, except in the literal sense that they can do some things that humans can't -- as can many things.

How do we anthropomorphize software, and why?

To anthropomorphize something is to attribute human qualities to it, whether it actually has those qualities or not. Compugeeks do this a lot, maybe not everyone but many of us. It's quite common to hear someone talk about what a particular piece of code thinks, or to say that a particular server is unhappy, or that a system is confused, and so on.

This was going on long before there were LLMs and I'm not even sure people do it more with LLMs than with ordinary code, though it has a distinctly different flavor with LLMs ("No, stop trying to patch the kernel so my parser's unit tests will pass. Why do you want to do that? What are you even thinking??")

We do this even though we know perfectly well that fifty lines of Python aren't thinking anything at all, and that an LLM-driven coding agent isn't actually trying to be annoying. Anthropomorphizing isn't treating something non-human as an actual human. We attribute some human qualities but not all.

To take an example, sorting a list of things into a given order, say, alphabetically, is a very useful operation because it enables fast algorithms like binary search. It's much faster to find a particular item in a sorted list than in one in a random order. If you give such an algorithm unsorted input, it won't behave correctly. If you track down a particular case of this happening, you might say "It expected the input to be sorted" and even "That's not its fault. We need to fix whatever gave it the unsorted input".

In a case like this, the code is standing in for the author. The author used an algorithm that required sorted input, with the expectation that anyone calling it would actually provide sorted input. As long as that's clearly noted, the author did their job and it's not their fault if the code fails on unsorted input. In other cases, we interpret behavior in terms of human behavior.

To take another example, if a server that normally runs without incident suddenly starts reporting a lot of errors, you might say "That server is angry," because it is very visibly reporting that something is going wrong. If a server tends to get bogged down due to some internal issue and start rejecting requests, you might say "It's in a mood", as in something like "Our service had to reject requests because the server it calls for FooService was in a mood". Or you could say it's "feeling unwell", or any of a number of other things. The common thread is that the server is not behaving as it usually would, and the reasons aren't clearly understood.

As far as I know there's not really a geek-standard way of saying such things, and what people say may well depend on the details of how a service tends to act. In other words, servers can have personalities.

What sort of human traits do we tend to attribute to software? Some of the more common ones are

• Knowledge: "It doesn't know whether this list could be shared, so it has to make a defensive copy"
• Goals: "It's trying to call FooService, but FooService is down, so it just sits there in a loop"
• Emotional and physical states: "FooServer is angry/unwell" "This code will get confused and panic if you give it unsorted input"
• Communication: "FooServer and BarServer talk to each other" "The executor asks the coordinator for the next task to execute"

Are there any human traits that can't be attributed to software? Probably not. Metaphors are fundamental to human thought and people can be very creative in applying them. Nonetheless, some examples come across as deliberately fanciful

• "FooServer and BarServer are scheming together to make sure this query fails"
• "This system lives in a happy world of unicorns and rainbows"
• "FooServer is lonely because no one wants to talk to it"
• "This code has decided that existence has no meaning and all results are equally valid"

The more I go over this, the more I think there's nothing special here about computing. We anthropomorphize all sorts of things: Other animals ("That cat is happy"), vehicles ("It's not a pretty car, but it's been a loyal friend through the years"), the weather ("The sun is trying to poke its way through the clouds") or really anything that can act on its own, or even appear to. Software fits very comfortably into that category.

As with metaphors in general, anthropomorphizing can be nearly invisible, as with the executor asking for tasks, or deliberately vivid, as with deciding existence has no meaning, or somewhere in between, as with not knowing whether a list could be shared. When a metaphor is conspicuous, we recognize it as such. When it's not, we just use it.

Postscript: In re-reading, it struck me that the examples I gave, like "parser's unit tests" and algorithms that expect sorted input, are more than a little dated. While it's not impossible that a working software engineer would end up writing a parser or a function that requires sorted input, if you're designing a web service you're much probably more concerned about things like UX on the front end and scaling and security on the back end. I stuck with the examples anyway for familiarity. Topics like parsing and binary search have been taught continuously for generations, so they're universally familiar, unlike whatever I was working on last week.

What is the pound measuring?

How much does the Orion capsule (that is, the Crew Module) that splashed down on April 11 weigh? According to NASA's reference guide for Orion, 22,900 pounds.

The guide specifically lists "liftoff weight", and there are a couple of reasons for that. One is that the capsule has reaction control thrusters, which are small rocket engines that allow for fine-tuning the attitude of the craft and small-scale maneuvering, and their propellant is part of that liftoff weight. For this and other reasons, the capsule did not have the exact same contents when it splashed down as when it took off.

The other reason, of course, is that the weight of the capsule depends on where the capsule is in its trajectory. For most of the mission, that weight was essentially zero, since the capsule was coasting in freefall except at a few key points. Units of weight, like pounds, measure force, not mass. At least that's what I was taught in high school physics.

For most practical purposes, though, the pound is a unit of mass. If the door of a bank vault weighs a ton (2,000) pounds, you know it will be a little hard to move, even if it's perfectly mounted on bearings with very low friction so that when you push on it you're not trying to lift its mass. That inertia is due to its mass. If you weigh out a quantity of something, you're interested in how much of it you're getting, that is, the total mass. The force that it exerts on the scale is just a way to deterimine the mass.

You're measuring that mass by way of how much that mass weighs on Earth, but it's still mass that you're measuring. Except in specialized applications like calculating load limits or foot-pounds of torque, the amount of force something exerts under gravity is secondary to how much of it you have.

Yes, it matters that a 22lb bag of something is easier to lift than a 44lb bag, but it matters just the same that a 10kg bag is easier to lift than a 20kg bag. You don't need to know the amount of force involved (about 98 and 196 Newtons, respectively) to make that determination and no one is thinking "Hmm ... that 20kg bag will require 196 Newtons to lift" before trying to pick it up.

There are units, the pound-mass and pound-force, that make the distinction between mass and weight. The pound-mass is now defined as exactly 0.45359237 kg, and the pound-force is the weight of this mass under standard Earth gravity of 9.8m/s².

No one uses this. Well, maybe not absolutely no one, but you won't find anything on a supermarket shelf that says it weighs, say, 1.5 lbm, because no one at the supermarket cares. If you're doing precise engineering or scientific work where the distinction matters, you're not using pounds, but kilograms and Newtons. This is just an example of the distinction I previously discussed between everyday units of measure, which can be pretty much anything, and precisely-defined scientific units of measure.

There are several reasons that SI (metric) units work better than imperial units for scientific work (and why, for example, the telemetry feed that NASA put up during Artemis II showed both SI and imperial units, with SI units first as I recall). One is the consistent use of powers of ten and standard prefixes like mega- and milli-. Another is that SI units have been standard for generations, so anything you're referencing in a scientific context is almost certainly using them. Another is the body of very careful definitions of what each unit means.

A less obvious reason is that SI units carefully make distinctions that we gloss over in everyday use, particularly the mass-weight distinction. During re-entry, when a capsule may be pulling on the order of 5g, it matters quite a bit that the forces on the body of the capsule are much higher than when the capsule is on the launch pad. You want to be talking about Newtons of force and not kilograms of mass when you do those calculations. Using pound interchangeably for pound-mass and pound-force in everyday speech makes good sense when you're buying groceries. Trying to use mass and force interchangeably in mechanical engineering is a recipe for disaster.

To make the distinction completely clear, the Newton is defined as a kilogram-meter per second squared, with no reference to Earth's gravity. A pound-mass weighs a pound-force under standard gravity because we don't really care about the distinction when using pounds. A kilogram weighs about 9.8 Newtons, which helps keep the distinction clear when it matters.

NASA is happy to quote the weight of Orion in pounds and show its speed in miles per hour because the US audience is used to those units. Trying to point out that actually the mass is about 10.4 tonnes and the weight varies is just going to get in the way unless you're specifically talking about the effects of acceleration or microgravity. Using pounds interchangeably for mass and weight is only incorrect if you're doing engineering or science, but then you shouldn't be using pounds at all.

Back to the space age

Yesterday, Artemis II launched four people on a flyby of the Moon, the first such crewed mission in 56 years. I have dim memories of the Apollo program, not so much the missions themselves -- I don't remember whether I heard Neil Armstrong's "One small step" live or later, for example -- but I do remember details like drinking Tang, because astronauts drank it (and still do), and a print on the back of a cereal box (I think?) that you could cut out and fold up into a model of the lunar lander.

The original Apollo program was a truly remarkable engineering feat, particularly considering how much progress there has been since then in fields like materials science and, of course computing. Today, we build massively powerful datacenters (at least, they seem massive now). At the start of the Apollo program in 1961, computers were much, much smaller and the field was so new that the word software had only been coined three years before.

It would be tempting to say that Artemis is just a retread of 50-year-old technology. In the years since the Apollo missions, space flight has become routine. There were 330 orbital launches in 2025, 317 of them successful. The ISS has been in continuous operation for 25 years. A dozen countries have launched satellites into orbit. Spacecraft have gone to all eight planets, Pluto, the Kuiper Belt object Arrokoth and to within about 6 million kilometers of the Sun (harder than it might sound). There were even two lunar landings last year, not to mention ongoing missions on Mars.

Except ...

The vast bulk of space activity has been launches to Low Earth Orbit (LEO for short). An orbital launch is not nothing. It means accelerating whatever you're launching to about 8 kilometers per second (17,500 mph) and handling all the details of tracking exactly where the launch vehicle is at all times, deploying the actual satellite and plenty of stuff I'm leaving out because I don't know any better. Nonetheless, as far as space travel is concerned, it's easy mode.

Everything else in the last 50+ years has been uncrewed. No human has been past LEO since Apollo 17 splashed down in 1972.

There are several reasons for this, not all of them technical, but the technical obstacles are considerable. For one thing, crewed missions are much heavier. Besides the mass of the people themselves, you need a life support system, food and water, equipment for a cabin and so on. More mass means a bigger rocket. 

Missions beyond LEO need a significantly higher delta-v budget, which is the total of all speed changes for the maneuvers the mission needs to do. LEO needs about 8 km/s of delta-v. Artemis II will use around 13km/s, about 1.6 times as much. Since it takes more fuel to lift more fuel, that means significantly more than 1.6 times as much fuel. In all, the Space Launch System (SLS) that launched Artemis II was the most powerful rocket that NASA has every laucnhed. The Saturn V rockets used in Apollo are not too far behind.

The stakes are also higher. 13 of the 330 launches, or about 4%, failed. If you leave out LEO launches (crewed or uncrewed), that number is much higher. There were two successful lunar landings last year, but also at least three failures. Of the two successes, one landed on its side. This sort of thing is OK if it's just expensive equipment getting destroyed, so you can afford to take more risks. For a crewed mission, nothing major can go wrong, and even minor problems like toilet malfunctions require serious attention.

Nonetheless, Artemis II is still pushing the envelope a bit, and not only in the power of the SLS. When Artemis II splashes down (assuming everything goes well up to then), it will be traveling at around 40 km/s, breaking the previous record held by Apollo 10. It will also go a bit further from Earth than the Apollo missions, so the Artemis crew will be further from Earth than anyone has ever been before.

From a strictly economic perspective, crewed missions make very little sense. The real reason to send people around to the Moon is that we want to send people to the Moon, either for its own sake, or so that we can establish a presence there and eventually send people to Mars and beyond. Whether that's a worthwhile goal is a matter for debate that I'm not going to take a position on here.

My point here is that a crewed mission to the Moon, or anywhere beyond LEO, wasn't just a major engineering feat 50 years ago. It's still a major engineering feat now. Practically all of the progress in the past 50 years ago has been aimed at solving different problems: Getting equipment and people to LEO, and getting equipment beyond LEO. Crewed missions like the ISS have told us a lot about what happens to people in space, and the Artemis mission reflects that, but not much about how to get them there that we didn't already know.

Suborbital crewed missions like Virgin Galactic's and Blue Origin's are pretty much irrelevant to all this.

As I write this, Artemis has successfully executed all but the last major maneuver in its mission. Before long, if all goes well, it will do the trans-lunar injection burn that will put it on a path to swing by the Moon and return directly to Earth. At that point, the crew needs to survive a bit more than a week in space and splash down safely. They also have a long list of mission goals to accomplish, of course.

I'm a bit surprised by my feelings about this. I've studied enough about the Apollo missions and spaceflight in general to know how much can go wrong, even with the most careful planning. NASA itself has lost crew members on multiple occasions. So while I'm excited for the crew and the many people on the ground, I'm also more nervous about it than I expected to be.

Beyond that, though, is a strange feeling of being in two timelines at once: a young kid in the late 1960s curious about all the moonshot stuff going on, and an adult watching nearly the same things happen 50 years later, almost as though for the first time.