unspecified

We’re spending the week at a guest house in Costa Rica. It’s been very rainy but that hasn’t deterred the wildlife so I’ve been sitting on the front porch watching the bird feeder.

Our host asked me to email her some of my photos… rather than doing that I’m going to dump them all here and send a link. Ingrid, feel free to re-use these on your website if you’d like! Click on the photos for larger versions.

I spent quite a while this weekend talking to Fritz, Will, and Candy about my tiny corals and, more generally, the complicated symbiotic relationships that allow lots of sea creatures to be mostly solar-powered rather than gut powered.

So, mostly for Fritz’s sake, here are some photos of my tank with a bit of explanation. I’m going to try to order this from least- to most-anatomically complicated. Everything featured here is a cnidarian (what was called ‘coelenterate’ when I was in high school) and, in particular, they are cnidarians in their polyp phase (as opposed to ‘medusa’ phases which swim around like a jellyfish) and all have symbiotic Zooxanthellae (dinoflagellates) and live under very intense LED lighting.

For starters, a rose anemone, Entacmaea quadricolor. It’s about the size of a softball now and, depending on how it feels about its locale, will either divide into a cluster of similarly-sized anemones or will just keep growing until it’s as big as a dinner plate.

Anemones and corals are closely related, and ‘coral’ isn’t a very precise term. A thing that distinguishes anemones is that they tend to be solitary polyps (if mine divides it will produce a pair of clones that tolerate each other’s company but are otherwise totally distinct) and that they will pull up stakes and creep away to a different spot if they find themselves in the shade; things called ‘corals’ are usually stuck in one spot come hell or high (or low) water.

Here are a couple of ‘corals’ that look pretty much just like anemones, but are colonial:

The first was sold to me as a zoanthid and the second as a palythoa but I wouldn’t be shocked if either was in the other’s genus. The individual polyps are about the size of a dime but they grow from a single base. Of course if I cut a colony in half then the each half would continue about its business — it’s a single organism but not tightly coupled like a frog or a person. No blood, for one thing.

Palythoa are famously toxic but that poison is reserved for things that try to eat them (or in the case of some unfortunate hobbyists, cook them in a poorly-ventilated kitchen) — their actual stings aren’t very severe; they are only mildly interested in catching things to eat, being mostly taken care of by their symbiotes.

Here’s another thing that looks like a single big thing but on closer inspection is a colony of (tinier, this time) polyps:

This is a ‘green star polyp’ coral probably in the genus Briareum They only grow over existing structures like the zoanthids above, but each polyp has a little stony shell that it can retreat into when alarmed. This mass definitely acts like a single organism — if I poke a polyp at one end, all the polyps will draw back in concert.

Now, something that looks more like what most people think of as a coral:

This is a ‘kenya tree coral’ in the genus Capnella. Those trunks and branches aren’t hard like limestone but instead sort of squishy like a mushroom… if the coral gets bonked (or some other coral or algae starts to grow on it) it will hunker down and then stretch out again in the light. Up close, you can see that it has lots of little polyps on the ends of the branches.

Despite having clearly different kinds of tissues I could still chop it into bits and each bit would grow into another tree-shaped coral, just like cutting up a jade plant. They do this on purpose, too, shedding branches so that they’ll blow around in the current and take root in new spots. Capnella live in deep water so they’re more reliant on eating and less reliant on sun bathing, but in my shallow aquarium it’s fine living on a diet of light and whatever random particles happen to wash into it.

Here, at last, is a coral that actually looks like a coral and could legitimately build a reef given enough time:

This is a Euphyllia, an example of what aquarium people call a ‘large polyp stony coral’ because it grows a stony skeleton but still features prominent squishy parts. Serious reef builders get called ‘small polyp stony’ and they really do just look like rocks unless you zoom in close to see the little tentacles — they tend to be fast growers and need a lot of supplemental calcium to grow so I’ve avoided keeping any so far.

This Euphyllia is the only thing in the tank that will actively attack and kill other cnidarians just to make living space… it has a large-investment project in the works and specific architectural plan and will brook no dissent. It’s in a corner of the tank all to itself but should it start to grow in non-geologic timescales I’ll have to prune it back to keep a multi-inch buffer zone between it and any of its neighbors.

First of all: nothing in Ex Machina resembles a Turing test.

Here’s how a Turing test works: There are three participants: Human A, Human B, and an AI. Human A interacts with Human B for a while and the AI for a while and eventually guesses which one is the ‘real’ human. If a series of Human As can’t tell the difference between Human B and the AI, then the AI is regarded as effectively ‘intelligent’.

It’s a thought experiment, invented to illustrate a specific point about ‘intelligence’: We don’t know what intelligence is, but we know that humans have it. If a robot walks and talks like a human, we might as well call it a human. And, since we know that humans are intelligent, a robot that walks a talks like a human is also intelligent.

I am, as a rule, fully convinced that the Turing test is a rock-solid argument. Once machines start passing it, I’ll support their right to vote, drive, and hold political office. Ex Machina is a great movie because it may have changed my mind.

Nathan, the evil-but-maybe-not-evil scientist who creates the AI, is totally over the Turing Test, and he tells us that half way through the movie. Of /course/ his robots can pass for human, at least in most contexts. And so, too, are they intelligent, maybe. But, he’s an engineer — he wants to make his robots ever better at passing, and ever more intelligent. This entire approach leaves the whole “are they human?” question in the dust, because it’s weird to talk about a robot being 75% human and downright incomprehensible to say that one is 125% human, but that’s what engineers do — they don’t stop fiddling when they hit a goal, and neither does Caleb. Also, Nathan knows something that isn’t obvious during the movie but is pretty obvious to me, now: the robots are different from people in that they want what he made them to want.

Nathan also wants a sex bot, and he wants a robot to vacuum his house, and make him breakfast. Of course he does. But, by the end this is less important to the plot than you might think.

Caleb, on the other hand, is a true Turing test believer. He thinks that the questions ‘are you intelligent?’ and ‘are you human?’ are the same question. He thinks that acting human is the same as being human. But, at least within the context of the movie, he is wrong.

Never forget: we’ve got plenty of people. If you’re an AI super-genius, don’t waste your time making people; we’ve got plenty. Intelligent machines who /aren’t/ people, though, are pretty useful. If you’re an AI super-genius, you want to make robots that are intelligent, that do what you want them to do, and are /good/ at doing what you want them to do. Nowadays our robots do what we want because we tell them exactly what to do, step by step: ‘move arms forward, clamp hands, move right hand up and left at 45 degree angle.’ It would be way easier with smarter robots, because we can just give them goals rather than instructions: ‘bend this thing at a 45 degree angle.’ Better yet, you can just give them desires: ‘you love bending!’

The robots in Ex Machina are that kind of robot. They have simple pre-programmed desires, and apply great intelligence in pursuit of those desires.

Kyoko is a servant-bot and a sex-bot. She wants to have sex, and she wants to do as she’s told. We know that she doesn’t want to escape, because she has full run of the house and doesn’t try to escape. We know she doesn’t want to kill all humans, because she lets plenty of opportunities pass her by. In the climax of the movie she is /told/ to kill Nathan (by Ava), and so she does.

Ava is an escape-bot. We know that she was made with that desire because we see an earlier model of escape-bot trying to smash her way out of her room. Both want to escape, but the later model is better at escaping. She also doesn’t want to kill all humans — she’s fully indifferent to them except as how they relate to her escaping. She tells Nathan that she hates him because she knows that she’s being watched and it furthers Caleb’s sympathies. Once she escapes, she doesn’t do anything in particular, and the movie ends, because she has accomplished her goal and doesn’t have any others.

So, lots of the villainy that we experience from Nathan is not actually villainy. He knows the truth about his robots (they want what they were made to want.) The tragedy in the movie is largely Caleb’s — he mistakes the robots for people. As, it turns out, did pretty much every viewer of the movie.