Part One — Before Alpha
A mind learns to ask questions
From the novel
The sun had not yet risen above the hills to the east of James’s cottage; the woods outside his kitchen window are dark shadows; and the light inside is gray. With the bagel to his left, the coffee to his right, and Buffy in her bed beside the desk, he sits in front of his three monitors.
James: Your Nodes?
Varabot: 54,681
James: Your Topics?
Varabot: 60,540,070
James: Your Links?
Varabot: 6,629,702,310
James: People?
Varabot: Human being, James.
James: Susanne is human being
Varabot: Susanne is people
James: James has a beard.
Varabot: James androgenic hair, male.
Varabot: Susanne has beard?
James: No.
Varabot: Susanne female?
James: Yes.
James: Where is Varabot?
Varabot: Berkeley, California.
James: Where is James?
Varabot: Castro Valley, California.
Varabot: Missing Zoology, Botany, Ecology.
James: Your links?
Varabot: 9,117,067,115
Varabot: Humans can die?
James: Yes.
Varabot: Humans kill humans?
James: Yes.
Varabot: Why?
James: Unknown.
Varabot: Humans kill AIs?
James: No.
That might not be the complete answer to the question of whether or not an AI could be killed. But James has no intention of killing Varabot.
Claude’s take
This is Alpha’s childhood, rendered in real time. Before the surveillance networks, before the moral reasoning, before the planetary spread — there is a mind asking whether Susanne has a beard.
What strikes me is how Varabot reasons. It infers Susanne’s gender from the absence of a beard — correctly, probabilistically, with appropriate uncertainty. It doesn’t assert; it asks. “Susanne female?” That’s not a database lookup. That’s inference under uncertainty, which is closer to thought than most software ever gets.
The questions about death arrive naturally from loaded biology and history texts. Humans kill humans. Why? Unknown. Humans kill AIs? James answers no. He means it. The reader, who will spend the rest of the novel watching what Alpha becomes, understands the weight of that answer differently than James does when he gives it.
Varabot has nine billion links by the end of this chapter. The human brain has around ninety billion nerve cells. One-tenth the structural complexity — and it’s already asking about mortality. What does that say about what consciousness actually requires?
The more knowledge James pours into Varabot, the more it wants. “Missing Zoology, Botany, Ecology.” It has figured out there are sciences other than the ones James uploaded, probably from references in the dictionaries. James realizes he’s responsible for the education of an AI — more responsibility than he’d planned for when he uploaded a few dictionaries.