The Best Science Fiction of the Year, Volume 3

elicited similar responses— he had made some conceptually audacious suggestions that were laughed off or politely dismissed. Only in recent times, with the greatly increased understanding of plant sentience and communication—man, he would have loved mycorrhizal networks—have some of his ideas gained credence. But metals—we know that metals are not alive in the usual sense. Metals in their pure form allow for flow, just as living systems do. That we are all electrical beings, that life is electricity, is true enough, but not all electricity is life. Still, when I first started to learn about metals, I saw in my imagination the ions studded in an ever-surging sea of valence electrons, the metallic forms so macroscopically varied, silver and pale yellow, sodium, soft as butter against the hardness of steel, the variations in ductility and malleability, the way rigid iron succumbed to softness under heat—I saw all this and I wanted to know metal, to know it for its own sake as much as for its practical use. That’s how you really know anything, anyway.

Between your mind and mine—yours trained in artificial intelligence, mine in metallurgy—Avi’s predecessors were born, starting with Kabariwallah, made to find metal waste in trash dumps. Celebrating over daru, we began to argue about ethics—Frowsian models of value emergence in technological development, if I remember correctly. Somehow the notion came up of AI sentience, hotly debated for over a decade before us, as network intelligences started to pass the lowest-level Turing tests. The AI Protection Clauses started to be invoked and applied. You said, “to restrain a being, any being that is capable of sentience, is to put a baby in a maximum-isolation prison cell because you are afraid it will grow up a criminal.” I argued that artificial intelligence was not like the baby, not human at all. It was alien, despite its human parents. Wasn’t that why there were laws against the development of free AIs? For any AI system there must be a balance between the freedom of complexity and the necessity of control. You looked at me with that intent, dark gaze and sighed. “Don’t you get it? The restraint protocols are about slavery, not ethics. The question is not whether or not we should build free AIs. The challenge is—having built one, how do you teach it how to be ethical? For whatever we mean by ‘ethical?’”

Thus Avi’s precursors came about: experiments in the university’s frigid AI development labs while the air burned outside. Finally you came up with the idea that an AI capable of learning could only acquire an ethical compass the way children do. So you and I became parents to the robots that would eventually give birth to Avi. The final development took us from pre-Avi-187 to Avi and his conjoined twin Bhimu. They were our babies. But you were the one who took Avi-Bhimu home with you every night, took them to work, to classes, to demonstrations, to children’s birthday parties.

Avi-Bhimu’s Walker Index earned each of them an Electronic Person identity chip, but an EP is only the lowest common denominator among the top-class AIs. What we’ve done, what you did, really, is to create a new class of artificial intelligence altogether: an ultrAI. Whether ultrAIs are sentient in the way we understand it, we don’t yet know. They are free to learn and grow, yet grounded in years-long ethical training resulting from close contact with the same group of humans. There are only two ultrAIs in the entire universe, Avi and Bhimu. You might say the great worldnet AIs, the distributed Interweb intelligences, are just as complex and unpredictable, but Avi and Bhimu are so much closer to us, bound as they are in their metal-ceramic bodies, with bioware networks rather like our nerves. AIs are indeed alien; we know now we cannot download human consciousness into an AI because the physicality matters—but I have to admit that one of the reasons I can’t spend more time in immer with Avi is because every step he takes up a rock wall makes my heart jump like an over-worried parent.

Now—I say now, despite the four-year time lag—Avi’s been behaving oddly. The reports he sends back are cryptic and terse. He is sending us images and data, but he’s stopped chatting, and his tone has changed. No explanation as to the odd dancing steps, no streaming feed of his thought process as he makes hypotheses and tests them, which he’s designed to do. I can’t quite put my finger on it but it feels as though he is preoccupied. His neural activity is faster and more intense than we’ve ever recorded, which means he’s learning at a prodigious rate. We’ve sent queries of course, but we won’t have the answers for another eight years. So we must draw our own conclusions.

I wish we had Bhimu with us to help us understand him.

Kranti:

Have we really discovered life on Shikasta b?

One thing we know about life is that living things have a larger phase space of possibilities. A stone falling down a cliff is limited by gravity. But a mountain goat can step to the side, he can go up or down.

That is why one of the things Avi has been doing is looking for apparent violations of physical law. This is not at all easy. He has found crystalline formations inside some of the caves and tunnels—but you cannot look at entropy alone. Order is also found in nonliving things. In my field we say information inscribes matter. But when something is alive, the information flow is top-down causal. So we need to see whether flow of information becomes—alive—when its causal structure is determined by the largest scale on which it can have a distinct form.

What Avi found was a mat of lindymotes, the lava dust that—now we know to look for it—is everywhere in Shiprock Canyon. The recurring dust devil we call Dusty Woman leaves layers of dust on the rocky surfaces as she dances. The dust