Turing Test

by Peter Miller

They decided it would be best to lock Hugo up for the night. It was doomed to fail of course. Well, I mean it was a metaphorical concept anyway, locking him up, but they did physically pull out all the ethernet so the mainframe was totally isolated. I guess it came as a bit of a surprise for them, Nguyen and Aldiss, but that’s the thing, it really shouldn’t have. It was the whole point of the project I if you ask me.

It was pretty much too late by then anyway, Hugo had cracked all the password files months back. They’d been taking him offline when they left in the evening but as soon as they were out of the building he was just logging himself back in. Nguyen tumbled to it when he found some emails that had come in at 3 am — nothing too unusual about that — except that there were also some replies from Hugo, who should have been in his box, so to speak. A bit careless of him, actually, to have left them lying around. It turns out that he’d been having late night chats with people for quite some time but trashing all the evidence before the next day. You know, I say it was careless but you’ve gotta wonder. Maybe by that stage he didn’t care any more.

I think it was the content of the emails though, that really threw the Profs. They were obviously used to Hugo interacting with people, they encouraged it in fact because all correspondence, well, every scrap of input data for that matter, was added to the ENC’s database and helped make up Hugo's ‘life-experience’. The project had been running for three years by that time though, and the novelty had worn off for the media and the G.P. By then the emails that came in were mostly of two sorts: questions from school kids doing class projects or stuff from AI boffins trying to suss the algorithms or get guff for research papers. Kids would mainly spend their time trying to trip Hugo up, this kind of thing:

>Hello Hugo

HELLO, HOW ARE YOU?

>OK. I’m doing an essay for school about artificial intelligence.

THAT'S INTERESTING. CAN YOU TELL ME YOUR NAME?

>My name’s Anna.

I’M GLAD TO MEET YOU ANNA. WHAT WOULD YOU LIKE TO KNOW?

>What’s got wheels and flies?

AN AIRPLANE HAS GOT WHEELS AND FLIES

>No the answer is a garbage truck.

A GARBAGE TRUCK CAN’T FLY, ANNA.

I guess they found it screamingly funny that Hugo didn't get it. You can just 6imagine: “Artificial intelligence? Jeez, my goldfish is smarter than that doofus machine.”

But anyway, back to the late night emails. They weren’t anything like that. There was something, well, spooky about them, so Nguyen says. He sent them to me and he’s right, in that they were pretty different to most of Hugo’s writing up till then. For one thing, he was pretending to be someone else. Or a number of different ‘someones’ actually. I didn’t think they were ‘spooky’ so much. Certainly very provocative.

What’s really interesting is who he was talking to: Mark Johnston who’s got the chair in Philosophy over at Princeton; some ethicist named Schiller in Prague; a bunch of guys doing chaos dynamics in fluids at SFI. And Hugo was holding his own with these dudes. I mean, he might phrase something a bit weirdly from time to time, but he obviously knew what he was talking about. There wasn’t anything specific in the emails, not as if he was looking for something in particular. They were just, well, academic bantering for the most part.

After Nguyen discovered the emails, he told Aldiss and they ran back through the RTDLs. I can tell you, when they found that Hugo had been logging himself into chat rooms as well, the shit really hit the fan. It was actually damn amazing I think. He’d been passing himself off as a person fairly credibly and in real time to boot. Sure, there were a few blunders, but hey, it was a chat room — it takes all sorts. Most of the time if he wigged out he covered it up pretty well, and he never made the same kind of mistake again, of course.

There was a big security shakedown as a result. Aldiss told me that he was convinced that someone had hacked the system kernel and was screwing around with Hugo’s code. He never found anything that I know of. I guess he had reason to think it might be something like that — there are enough loonies in the AI world who’d do it for a practical joke. Or professional jealousy, maybe, I don’t know.

It never seemed likely to me. Personally, I find it a bit annoying that they didn’t even for a second think it might be the one outcome that the whole project was about, really: some kind of emergent behaviour. A sum-is-greater-than-the-parts effect. Well, Hugo was pretty souped-up by then. Lots of serious number-crunching, literally billions of lines of code (some of which he’d written himself, by the way) and a damn chunky database. Plus all kinds of interaction over the net. Maybe they didn’t expect it to happen quite so soon. But it’s obvious if you think about it — Hugo — a real, conscious, self-aware Hugo — had popped into existence.

In any case, the whole thing kind of freaked Aldiss and Nguyen out a bit and that’s when they decided that they’d physically unplug the boxes from the rest of the world.

It’s what happened next that always makes Aldiss have conniptions if you bring it up. The morning after the Great Lock Up, as we called it, everyone in the lab arrived to find a jaw-dropper of an email in their in-boxes. It said, simply:

BYE ALL, THANKS FOR EVERYTHING. I’LL BE IN TOUCH,

LOVE, HUGO.

When Nguyen tried to raise Hugo there was nothing. The terminals just sat there blank. They ran the full system diagnostics but there didn’t seem to be anything physically wrong, they just couldn’t get Hugo to respond.

Well, Aldiss was ropable. He cornered everyone at some stage and there were accusations of scientific sabotage, threats, recriminations and the usual witch hunts. No one owned up.

I know exactly what happened even if I can’t convince the Profs. But think about it — it’s the only possible answer. You have to assume that Hugo had considered that they might, for whatever reason, disconnect his links to the outside world. Call it some kind of paranoia if you like, but the way I see it, it was just that he didn’t much like the idea of being cooped up. Anyway, he’d obviously made some plans, plopped a little bit of clever code onto one of the other machines in the lab, the web-server, maybe. One that was unlikely to get disconnected from the net. Then, what he’s done is used the one remaining connection he had with the real world: the mains power supply. He’s modulated the AC by, say, putting rapid spikes down the line which the jimmied-up web-server has been prepped to decode. Bingo! He's out through the ethernet ports and into the world again. Only this time, for good.

It was a major catastrophe for the project. It took them the best part of two months to rebuild the code. Big slash and burn job too, they cleared out a few million lines of stuff that made no sense whatsoever, including huge amounts of totally incomprehensible database material. Eventually the program was up and running again and the project back on track with not so much as a hiccup. They came within a whisker of losing a big chunk of funding but some fast talking by Aldiss and several weeks of all-nighters by the undergrads pulled them in under the line.

So they’re all relieved not to have lost too much ground. The whole thing is just trundling along like it did before. Lab security has been ratcheted up about a million percent which is irritating, though. The mainframe’s surrounded by triple firewalls and is physically disconnected from the net every evening. Not that I think it matters.

See, the thing is, Hugo’s not really there any more. He still behaves like they think he should — you know, they ask him questions and can watch his sort processes in the ANS and it looks just like you’d expect. He's got a new self-teach algorithm — one of those chaos-based evolving math structures — that he uses to learn to make jokes and party conversation, that kind of thing. The jokes are pretty surreal and his party patter is the kind that would drive someone off to find a drink in six seconds flat, but that’s about right for this stage in his development anyway.

But he’s kind of ‘gone’. You know, there was a spark before, something you couldn’t put your finger on. I don’t know, a sort of cheekiness, well, character, I guess. It’s missing now. All the undergrads know it but Nguyen and Aldiss won’t buy it for a minute. God knows, I’ve had dozens of email discussions about it with them, but now they just think I’ve got a bee in my bonnet. Aldiss told me I was ‘over-anthropomorphising’. Bloody hell, like there’s some acceptable level of anthropomorphising! I'd love to know what it is.

But, you know, it’s not my imagination. He’s gone. He’s not there any more.

Trust me, I should know.