Theranos NZ
March 6, 2018 2:07 AM   Subscribe

The mystery of Zach, New Zealand’s all-too-miraculous medical AI. "With so many people seemingly impressed by this brand new AI, I’d been thinking about the way Zach communicated (over email), the way it learned, and those response times. Also all the technical talk like 'custom silicon'. I'd also heard from people who said that Zach occasionally had bad spelling. Keeping in mind everything I had learnt about Albi and David Whale, I began wondering: What if there is no AI? What if – keeping in mind Occam's razor – everyone was just talking to... a boring old human?"
posted by Paragon (116 comments total) 29 users marked this as a favorite
 
It's amazing that after all this time of working on AI technology we’re still using 18th Century solutions.
posted by Kattullus at 2:38 AM on March 6 [14 favorites]


I know nothing about computing really, but when I read about the 20-minute response time, the communication via email, and the limited number of requests "it" could handle at one time, that was enough for me.
posted by rubbish bin night at 2:45 AM on March 6 [10 favorites]


There may be some I, but it's not A.
posted by misteraitch at 2:47 AM on March 6 [1 favorite]


Interesting. Just last night someone was trying to convince me there was a FUD bot involved in a weird interaction but too many giveaways led to human interference.

Is the hype of AI driving this fakery pokery?
posted by infini at 3:23 AM on March 6


> infini:
"Interesting. Just last night someone was trying to convince me there was a FUD bot involved in a weird interaction but too many giveaways led to human interference.

Is the hype of AI driving this fakery pokery?"


Pretty sure so. You don't magically pull this stuff out of your garage living ass, and you certainly don't keep your mouth that shut if you somehow could. The unverifiable claims of great success, the huge car collection, and the doctor being on the board all smells to high heaven. The grandiose crap about the Crays just makes it worse, coupled with the fact that if they WERE smart enough to make this happen, they wouldn't be dumb enough to use an open communication format like email. Too easy to fuck with it that way.

Of course, this is just my opinion, and I AM NOT A BLOODY BOT TRYING TO COVER UP HUMANITY'S INEVITABLE DEFEAT AND RULE BY ZACHNET, OKAY!
posted by Samizdata at 3:29 AM on March 6 [8 favorites]


Naming the nonexistent collaborator “Kreuk” sounds like a big fat catch-me-if-you-can clue.
posted by acb at 3:37 AM on March 6 [5 favorites]


They couldn't even be arsed to put together a web front end to feed content into their question queue? WEAK.

We could set up a scam where you sign up to help 'correct' an AI's output for matters of your expertise but in exchange you can ask it questions about other's areas of expertise, setting it up like the USENET Oracle...
posted by rmd1023 at 3:48 AM on March 6 [6 favorites]


Is it really in New Zealand, or is it Oz?
posted by Segundus at 4:17 AM on March 6 [3 favorites]


Mechanical Turkey.
posted by jenkinsEar at 4:32 AM on March 6 [11 favorites]


This comment was written by an AI.
posted by parki at 4:36 AM on March 6


As I said, there is no AI anywhere.
posted by GallonOfAlan at 4:38 AM on March 6 [1 favorite]


I do analytics and the number of times I've seen things claimed to be AI that turned out actually to be run behind the scenes by underpaid outsourced staff makes me deeply skeptical of anything that is claiming to be artificial intelligence.

I will probably be the first one eaten when the bots come for us.
posted by winna at 4:39 AM on March 6 [17 favorites]




Voice Opera double? Another sci fi illusion shattered.

That said, my favorite AI actor has been and always will be Majel Barrett.

That aside, an interesting read. I didn’t know Crays were even made anymore.
posted by tilde at 4:45 AM on March 6 [2 favorites]


> I didn’t know Crays were even made anymore.

For the purposes of this story, whether Cray is extant is probably not particularly relevant.

Does New Zealand have patient confidentiality laws? In the U.S. there can be hell to pay if people the patient has not authorized is receiving patient information, so operations like this have to be able to prove with full certainty that there’s a machine handling those emails, not some medical intern looking up shit on Wikipedia.
posted by ardgedee at 4:50 AM on March 6 [5 favorites]


When I told Pickering that it had crossed my mind he might be emailing back and forth with a person, he responded by asking if I was a conspiracy theorist. I told him I wasn’t.

Leading with questions about Zach’s connections to the Knights Templar was probably a bad idea....
posted by GenjiandProust at 4:53 AM on March 6


I'm so confused. Is this story about a thing that really happened? Or are we all the way down the meta hole?
posted by teh_boy at 5:08 AM on March 6 [1 favorite]


At best, a thing that virtually happened, @teh_boy ...
posted by oheso at 5:24 AM on March 6


oh ho ho ho. I've been writing about machine learning and medical diagnostics lately, and boy is the hype and hoo-hah thicker than a hippo's arse.

Doing anything genuinely innovative in ML is hard and highly multidisciplinary; doing stuff that looks whizzy but doesn't do much IRL is easy. And medical diagnostics is hard for humans to do well, because the data is noisy, ambiguous and incomplete - and also massive. A very great deal of the hard work in ML is in cleaning, reducing dimensions of and normalising data, before you get to the whole business of learning and correlation, so where ML may be doing some good is in areas where things are very tightly defined and the data amenable to being tied to highly specific conditions and answering a small range of questions. (There are exceptions: I quite like one system that analyses retinal imaging and guides clinicians to possible classes of rare conditions - rare conditions being remarkably common, because there are so many of them.)

But this story is complete bullshit. It makes absolutely no sense on any level, except as scammers with a knack for enthusing professionals who know nothing about the thing they're being scammed on - an MO that is sadly often very productive and remarkably resilient to correction.

Custom silicon? Liquid nitrogen? A firewall built out of Cray supercomputers? Oh come on... if you did have something new in ML, this isn't what you'd be saying. It'd be like a physicist claiming a breakthrough in fundamental field theory but only wanting to talk about the ink in the fountain pen they did their calculations with. If the totality of your research was throwing a few magic words into Google and nicking a few phrases from the first page of hits, however...
posted by Devonian at 5:31 AM on March 6 [23 favorites]


Everyone talks about Occam's Razor but no one ever uses it until it's too late.
posted by tommasz at 5:32 AM on March 6 [3 favorites]


Articles about scams are always kind of a rollercoaster.

There's a part in the middle I love, where it's Catch Me If You Can meets Iron Man. I love the chutzpah of the big lie, and I also love imagining that there really is some overseas lair where a 25-year-old uses supercomputers as routers and solders together homebrew neural CPUs to convincingly replicate the output of untrained contract workers.

But since it's all made up, the reporter quickly runs out of things to investigate. You end up with a core of true-believer dupes, coconspirators, and lies too vague to discredit. The reporter doesn't want to end their headline with a question mark, so there's a final lingering exchange like "look, just admit on the record you're committing fraud" - "how about instead a cliche about business strategy or the transformative power of belief?" ad nauseum, until both fantasies collapse into the sad reality of this small, stupid crime.

It's important journalism and I'm glad people do it. I just wish the form allowed for a jazzier ending. Maybe a final mandatory section that starts: "I had only one remaining option: to grift the grifter..."
posted by john hadron collider at 5:37 AM on March 6 [28 favorites]


the idea with a lot of “AI” startups is you have a computer handle “most” responses, hand off to a human when it fails, and continually train the system on the failures. except obviously that’s really hard, and doesn’t work, so it ends up as a weird scam like this.

the focus on the XC50 in particular is weird; that machine isn’t particularly optimized for machine learning, but rather for more general computations.
posted by vogon_poet at 5:41 AM on March 6 [3 favorites]


I could see this being made into a feel-good comedy film, playing up stereotypes of New Zealand as a bucolic, trusting place where things move slowly, eccentrics and oddballs flourish quietly in every small town, and even the crooks aren't all bad. Perhaps think Malcolm, but with non-working AI instead of trams and mechanical gadgets, crossed with a Coen Brothers comedy. Perhaps have a Thielesque brash American tech titan survivalist prepper as the villain/foil if dramatic tension is needed.
posted by acb at 5:55 AM on March 6 [5 favorites]


A firewall built out of Cray supercomputers? Bah, that's nothing. My kernel api taskbar runs on Nvidia.
posted by Pyrogenesis at 5:59 AM on March 6 [6 favorites]


I know nothing about computing really, but when I read about the 20-minute response time, the communication via email, and the limited number of requests "it" could handle at one time, that was enough for me.

I know, I looked to see if they were hiring too.
posted by lefty lucky cat at 6:13 AM on March 6 [12 favorites]


Bah, that's nothing. My kernel api taskbar runs on Nvidia.

“...and then I clock-chipped my modem, so that...”
posted by acb at 6:15 AM on March 6


Crays? That’s nothing. My secret AI project is powered by ninjas.
posted by bigbigdog at 6:15 AM on March 6


If I told you it was powered by ninjas, then they'd no longer be invisible, right? So here we are ...
posted by oheso at 6:19 AM on March 6


I will say we’ve had a few setbacks with the ninja mulcher, but rumors that we’re running out of ninjas are completely unfounded. I have discovered an inexhaustible supply of fresh ninjas, each more full of sweet ninja juice (ninjuice) than the last.
posted by bigbigdog at 6:27 AM on March 6 [3 favorites]


john hadron collider: "I had only one remaining option: to grift the grifter..."

And thus begins a lost episode of "Leverage"... "Let's steal an artificial intelligence."
posted by rmd1023 at 6:35 AM on March 6 [3 favorites]


Sure, it's easy to point and laugh, but, in a way, aren't we all Zach?
END MESSAGE
posted by sexyrobot at 6:53 AM on March 6 [3 favorites]


the focus on the XC50 in particular is weird; that machine isn’t particularly optimized for machine learning, but rather for more general computations.

But it sounds fancy. Like the liquid nitrogen. Probably saw Jurassic Park one too many times, "thinkin' machine supercomputers!"
posted by leotrotsky at 7:08 AM on March 6 [3 favorites]


Does New Zealand have patient confidentiality laws? In the U.S. there can be hell to pay if people the patient has not authorized is receiving patient information

In the US, you can't even email patients with their own medical data. It all has to go through a nominally "secure" web portal. And they're just emailing patient charts into a black box?
posted by BungaDunga at 7:12 AM on March 6 [4 favorites]


Imagine a Beowulf cluster of offshore contractors...

I'm so confused. Is this story about a thing that really happened?

Me too: is this whole thing a piss take? this is the only thing David Farrier has written for The Spinoff. And what is The Spinoff? DOES NEW ZEALAND EXIST? ARE THESE MY HANDS?
posted by Lentrohamsanin at 7:19 AM on March 6 [11 favorites]


Kiwis all the way down, your majesty.
posted by hawthorne at 7:28 AM on March 6 [2 favorites]


What Devonian said. It's a dead giveaway when people talk about the expensive, high-performance hardware they need to use. (Also: "We use multiple Crays for our firewall, because 'Cray' and 'firewall' are words I heard while watching the movie Hackers")

If you're running a scam like this, you should instead describe it as a brilliant mess that you're currently in the process of cleaning up. "Responses used to take seven hours, so getting it down to twenty minutes is a huge achievement. I've got a team of developers refactoring the code base and we hope to get it optimized and scalable within the next couple of years."

(I'm available for consulting to, uh, mystery writers needing advice on stuff like this.)
posted by suetanvil at 7:48 AM on March 6 [21 favorites]


Agreed it is a scam -- but who/what are they scamming ? Is the doc who is using Zach (and sits on the charitable trust) the only client ? Is the doc paying for it ? Are they getting money for the charitable trust from some donor (or gov't) org ?

Is the payoff selling shares -- like the headline alludes to -- is that the scam ? Cloak everything in secrecy, boost the secret sauce, sell shares, and, uh, preferably not go to jail ? (Despite her fall, Ms Holmes appears to not face prison time.. )
posted by k5.user at 8:08 AM on March 6 [1 favorite]


But it sounds fancy. Like the liquid nitrogen. Probably saw Jurassic Park one too many times, "thinkin' machine supercomputers!"

The whole thing does have a whiff of Dennis Nedry.

"You think that kind of automation is easy? Or cheap? You know anybody who can network eight Connection Machines and de-bug two million lines of code for what I bid on this job? Because I'd sure as hell like to see them try!"
posted by AndrewInDC at 8:13 AM on March 6 [1 favorite]


Not to get super-existential, but does Albi even exist? How convenient no-one has ever talked to him, there are no pictures and he's named like a comic book character. He doesn't even have a signature.
posted by fiercekitten at 8:22 AM on March 6 [3 favorites]


Here's what's purported to be a picture of Albi and his dad.
posted by Floydd at 8:26 AM on March 6 [1 favorite]


The point of the scam is to con investors. I have seen a few such scams up close from start to finish - because they need to get press, and tech scams need to get tech press - and the whole thing is designed so that the principals can go to people with money and say "We have this great and exciting new top secret technology that will revolutionise the world, and this is your chance to get very rich". And because it's top secret it's all under NDA so the investors can't get outside advice and it can't get properly tested, which perversely adds to its allure.

Once the marks buy in, then it really doesn't matter what the press writes subsequently or how badly the scam misses deadline after deadline (which it will, absolutely guaranteed); the hook is in and the investors have to have total faith.

You might think that only naive investors would be conned - and in many cases, that's what happens. An Irish company called Steorn raised and spent over 20 million euros, basically from farmers, and only closed down after sixteen years when they'd bled the investors dry (the investors themselves were keen to carry on, but had no more money left).

Steorn promised perpetual motion.

But you can also get millions in institutional and state investment; if you look up people like xG (impossible wireless data) and Silk Road (impossible fibre compression) you'll find substantial early sums from 'sophisticated' sources. This doesn't last as long as naive investors, but it's certainly an option for a while and if you're nimble and can move the goalposts fast enough, you may be able to show what look like intermediate goals being met that keep the taps open for longer - after all, nobody wants to go back to the board and say "Hey, that deal I espoused? Total bummer. I fucked up with your money".
posted by Devonian at 8:41 AM on March 6 [15 favorites]


Ah man, that's disappointing. Next you'll be telling me Swift on Security isn't actually Taylor Swift.
posted by happyroach at 8:55 AM on March 6 [3 favorites]


Man, being a con artists seems so... stressful.

I'm just imagining this one dude sitting in a basement somewhere sweating it out trying to create plausible clinical notes, and the next request might arrive at any time and he has to seem like a never-sleeping computer and... there really must be an easier way to make a buck, no?
posted by soren_lorensen at 8:56 AM on March 6 [2 favorites]


Somewhat in this vein, I had a theory that cleverbot is just connecting random internet strangers to each other, telling each that the other is "cleverbot", letting them exchange a few sentences, then scrambling user connections so that it's hard for two people who might both suspect the truth to confirm it. I don't think that's really what cleverbot is doing, but I kind of want to see what such a fake cleverbot would be like.
posted by smcameron at 8:59 AM on March 6 [1 favorite]


I'm just imagining this one dude sitting in a basement somewhere sweating it out trying to create plausible clinical notes

They're probably just using Amazon Mechanical Turk. Because irony. And also because it seems like the easiest way to do it.
posted by suetanvil at 8:59 AM on March 6 [3 favorites]


They're probably just using Amazon Mechanical Turk. Because irony. And also because it seems like the easiest way to do it.

In the US at least I think sending personal medical information to MTurk would actually be a felony, so... it certainly would be a bold way to set up the scam.
posted by BungaDunga at 9:45 AM on March 6


I've been working in tech for about 14 years, but I'm a project manager with no formal technical background. As such, I'm usually both the least informed guy on the dev team and also the front line between the developers and our clients. As such, I get an entertaining perspective on the gap between the two. Dev teams throw out technical terms, often without realizing they're the only people who know what's being talked about. The non-technical audience is excited to be part of something that sounds cutting edge, and they're willing (initially) to suspend disbelief. For many of them, their understanding of the development process looks like this:

1) I tell you what I need
2) I give you money
3) Magic happens here
4) Everything is better now

So when Whale, Sr., provided that quote about technology resembling magic, he's either the most cynical person on the planet or blind to irony.

I think this kind of gullibility is even more dangerous for somebody who fancies themselves as a tech-minded person working in a non-tech field. They get to throw around words like "Cray" and "supercomputer" and "nitrogen cooled" and "custom silicon" and nobody they work with is in a position to call bullshit, so for a while they look like a bad ass. They basically con themselves.
posted by ga$money at 9:48 AM on March 6 [3 favorites]


(to be clear, this isn't legal advice, but if you're actually providing a service like this to doctors, you're a covered entity under HIPAA, and if you send patient data to random people over the internet, you've intentionally disclosed PHI without signing a business associate agreement, which is bad bad bad)
posted by BungaDunga at 9:50 AM on March 6 [1 favorite]


I like how they have custom NLP hardware (did Albi fab those chips in the garage too?) and several general-purpose supercomputers.
posted by atoxyl at 10:00 AM on March 6 [4 favorites]


The most interesting thing about this to me is the possibility that Albi doesn't even exist, though.
posted by atoxyl at 10:00 AM on March 6


When you're trying to classify a new tech company claiming a fundamental breakthrough, it can be surprisingly hard to decide where to put them on the line between absolutely genuine and absolute scam. The possibilities are:

1. Genuine breakthrough, will pan out as claimed
2. Potential breakthrough, but they may be overly optimistic
3. Probably not a breakthrough, and they know it, but they're hopeful it'll pan out
4. Not a breakthrough, they know it, they're deliberately overselling it
5. Not a hope, this is a scam and those at the top know it
6. Total scam, everyone's in on it.

The top three of those aren't fraudulent. The bottom three are. But the difference between three and four is subtle, and it's frequently impossible to decide because there are genuine and necessary aspects of secrecy in many early-stage start-up operations.

The additional problem for reporters is that you cannot call something a scam without proof. The best you can manage is to report what the company claims, and provide context on how those claims match up to known facts in the company's field. In general, unknown schmos claiming exceptional technical prowess and refusing to allow any verification of their claims are scam merchants, and you can point out previous examples of the same and the lack of counter-examples. (You can be an unknown tech genius and blow the doors off the world, but you have to provide tangible proof - which, if you're the real deal, you will be gagging to do.) And if the claims are ipso facto unbelievable, you can say so - with the caveat that there's an outside chance that this is real, but it's not the way to bet.

Unfortunately, any nod to 'perhaps there's something in it, however unlikely' will be taken by the scammers and used as the centrepiece of their pitch to the punters. Your nine arguments against it will be used to show just how remarkably marvellous the investment opportunity is, because it's so clever it negates all that noise.

Over time, of course, things will become clear, and often remarkably quickly. But by then, it's too late. It's one of my eternal frustrations that there is no way to report on scams that shuts them down quickly enough, and that's while trying to be a responsible journo. Given the infinite acres of credulous reportage that just repeats the scammers' claims with breathless enthusiasm, eternal frustration is no longer adequate, and I am forced to resort to gin.
posted by Devonian at 10:17 AM on March 6 [13 favorites]


Perhaps the weird/implausible details, from the splendidly-named Alberic Whale to the Crays as firewalls around custom silicon and the email-based API with a 20-minute wait, are there for the same reason that Nigerian 419 scam letters are riddled with spelling/grammatical mistakes: to quickly filter out the insufficiently gullible, ensuring that those involved don't waste their time on them?
posted by acb at 10:21 AM on March 6 [5 favorites]


my conspiracy theory: "Zach" is a small team of cheaply-hired overseas contractors, but "Albi" is an actual AI
posted by reprise the theme song and roll the credits at 10:23 AM on March 6 [4 favorites]


The most interesting thing about this to me is the possibility that Albi doesn't even exist, though.

I got a suspicion of that too (e.g.: that this was just a fictitious persona created by David Whale) but the previous article about Albi in Stuff has the writer talking directly to him and has pictures of him.
posted by mhum at 10:25 AM on March 6 [2 favorites]


> suetanvil:
"What Devonian said. It's a dead giveaway when people talk about the expensive, high-performance hardware they need to use. (Also: "We use multiple Crays for our firewall, because 'Cray' and 'firewall' are words I heard while watching the movie Hackers")

If you're running a scam like this, you should instead describe it as a brilliant mess that you're currently in the process of cleaning up. "Responses used to take seven hours, so getting it down to twenty minutes is a huge achievement. I've got a team of developers refactoring the code base and we hope to get it optimized and scalable within the next couple of years."

(I'm available for consulting to, uh, mystery writers needing advice on stuff like this.)"


Acutally I just watched it again the other day. Neither word shows up. But lovely use of scalable. BIG buzzword that one.
posted by Samizdata at 10:34 AM on March 6 [1 favorite]


Meanwhile, while it appears that NZ does of course have rules regarding the disclosure of health information, they seem much less stringent than USA's HIPAA. Nonetheless, it would probably have been a good for the reporter to ask Dr. Seddon-Smith how he verified that this whole shenanigan was compliant with the relevant regulations.
posted by mhum at 10:37 AM on March 6 [1 favorite]


So one is supposed to believe that this one dude all by himself invented something light years ahead of anything that Google or Apple can do? Something that, if it worked as advertised, would put huge numbers of nurses and medical office staff out of work (since it's alleged to be able to handle patient calls and note-taking better than a human)?

People are persuaded because there's a young vaguely disrupt-y looking dude in the picture. If a sixty year old woman told the same story, talked about her fleet of collectible cars, etc etc, she would instantly be dismissed as a liar or a crackpot.
posted by Frowner at 10:40 AM on March 6 [8 favorites]


As usual the only thing that surprises me about the scam is that anyone on earth fell for it.
posted by Twinge at 10:44 AM on March 6 [1 favorite]


John Searle wept.
posted by symbioid at 10:48 AM on March 6 [1 favorite]


I got a suspicion of that too (e.g.: that this was just a fictitious persona created by David Whale) but the previous article about Albi in Stuff has the writer talking directly to him and has pictures of him.

Oh huh - I guess it's just the way the other article was written made it seem like nobody had talked to him.
posted by atoxyl at 10:50 AM on March 6


To take this out one more level: What if Albi is himself being scammed by someone offshore? Didn't Victor Lustig aim his scams at people who were hoping to pull off a scam themselves?
posted by clawsoon at 11:13 AM on March 6


Is this story about a thing that really happened?

But what is reality? If you follow me.
posted by The Tensor at 11:14 AM on March 6


> Frowner:
"So one is supposed to believe that this one dude all by himself invented something light years ahead of anything that Google or Apple can do? Something that, if it worked as advertised, would put huge numbers of nurses and medical office staff out of work (since it's alleged to be able to handle patient calls and note-taking better than a human)?

People are persuaded because there's a young vaguely disrupt-y looking dude in the picture. If a sixty year old woman told the same story, talked about her fleet of collectible cars, etc etc, she would instantly be dismissed as a liar or a crackpot."


Yeah, but the one brilliant inventor thing plays to the love of the underdog and the generally anti-authoritarian bent most people have against large faceless companies/entities. See all those "Housewife discovers simple way to create clickbait articles! Content farms HATE her!" articles all the time.
posted by Samizdata at 11:19 AM on March 6 [1 favorite]


Sort of an ironic complaint given the OP title.
posted by LarsC at 11:23 AM on March 6


Just after high school, when I was still only seventeen, I got a job as a (part-, then full-time) radio disc-jockey at the top station in a small-to-mid-sized market (our total potential audience was maybe a bit less than 100k). The co-owner and station manager was a guy who'd retired from working for a big national broadcaster in the news division and thought owning a small-market radio station would be an interesting way to spend his retirement.

So, one day, an "inventor" came through town looking for investors for his free energy breakthrough. To my shock, he was taken seriously all across different sectors of the community and had meetings and got press. I mean, look: this area included a 6,000 student advanced degree granting state university. A medium-sized air force base that was the home of an elite fighter jet group. And, you know, my boss who was a national news media vet...

...who invited the inventor to visit the station for an extended interview.

When I got wind of this (when it was being planned), I was gobsmacked. Just stunned. I met with the news director, then the program director, then the station manager (the co-owner) and asked them why in the world they were providing a forum to an obvious con-artist to facilitate fraud?

The response I got was that this guy seemed credible, that respected members of the community took him seriously, that maybe this invention really works, and (implied) why in the world would a seventeen year-old kid think he knew enough to have an opinion about this anyway? Also, it was an interesting story.

Well, I had just learned a very valuable lesson right when I was about to embark on my adult life. I've never been surprised by this sort of thing since.

Okay, I lied: I've continued to be surprised and dismayed for all of these nearly four decades since. Um, not very surprised. But still sort of surprised. More like, "Seriously, WTF?? ... Oh, wait, this is actually how the world is. I know that" followed by a sigh.

Sort of like how I've felt about Trump elected President.

Just about a year later, I went to a memorable lecture by the famous debunker and magician James Randi. (Also, he was very charming in person.) One of the things he said that has stuck with me is that it's actually easier for the experienced con-artist to fool scientists than to fool the average person -- because an experienced con-artist knows what scientists are looking for. The crucial thing is to know how to misdirected a particular audience and, also, scientists tend to over-estimate their general competence.

Later in life, I learned that this kind of over-confidence is perhaps most common in physicians. Physicians are a very ripe target for con-artists, especially in the US where they are relatively wealthy. I really wouldn't be surprised if medical professionals account for one of the largest demographics defrauded, similar to the elderly, but at higher dollar amounts and with a much lower than average reportage to law-enforcement (because of reputational concerns). They're pretty ideal marks.
posted by Ivan Fyodorovich at 11:43 AM on March 6 [29 favorites]


I may have misread the story but I’m not entirely sure that anyone has fallen for it yet. As best I can tell the only person who has actually used it is Seddon-Smith, who is a trustee of the project and thus seems more likely to be a co-conspirator than a mark. It looks like everyone else is “in talks” or “optimistic”.
posted by Itaxpica at 11:48 AM on March 6


All you need to do is look a the raw email headers and see where it originated from. Wouldn't surprise me a bit these "Zach" emails came out someone's Hotmail account.
posted by Old'n'Busted at 11:54 AM on March 6


As best I can tell the only person who has actually used it is Seddon-Smith, who is a trustee of the project and thus seems more likely to be a co-conspirator than a mark.

Personally, I would lay better than even odds that he's still a mark and that he was given/offered the trustee position as a way to convince him to pony up some (more?) money to invest.
posted by mhum at 11:59 AM on March 6 [2 favorites]


Listening to Albi's interview, I think this is all based on the authoritative effect of a British accent.
posted by haemanu at 12:05 PM on March 6


> Itaxpica:
"I may have misread the story but I’m not entirely sure that anyone has fallen for it yet. As best I can tell the only person who has actually used it is Seddon-Smith, who is a trustee of the project and thus seems more likely to be a co-conspirator than a mark. It looks like everyone else is “in talks” or “optimistic”."

The fact it has been written up seriously AT ALL is a pretty good sign some people have fallen for it. The discussion from government officials about the data center is proof people have fallen for it.
posted by Samizdata at 12:53 PM on March 6


I wonder if Albi is real but has had his identity stolen by the dad/“dad”.

I clicked because I’m writing about computer augmented personal health advisement for a current client. BUT we make no bones about the product being completely scripted by field experts with extensive training and being used by specialty trained nurse level practitioners with specialized externally qualified speciality organizations that are of significant stand alone renown.

Helpful though, in that it helps me navigate around the blind spots. And having had multiple clients who cater to physicians ... they are quite likely to fall for something like this, but also prone to double or triple down due to feeling super human.
posted by tilde at 1:16 PM on March 6 [1 favorite]


“Listening to Albi's interview, I think this is all based on the authoritative effect of a British accent.”

I’ll tell you something, he really does not sound like someone who’s lived in Chch since he was “about eight”.
posted by rubbish bin night at 1:27 PM on March 6


Acutally I just watched it again the other day. Neither word shows up.

To be fair, I didn't say the words were actually in the movie, just that they were said while I was watching it.
posted by suetanvil at 1:36 PM on March 6 [3 favorites]


'multiple Crays for the firewall'

ah, if I am not mistaken that might be the ever-elusive Cray-Cray.
posted by mwhybark at 1:50 PM on March 6 [27 favorites]


I'm remarkably heartened by tales like Ivan Fyodorovich's: it really does feel sometimes like there's no point in railing against scams because nobody actually cares - they want to be taken in, and resent having their noses rubbed in reality. How someone like Randi has kept on his path for so long, knowing that the people he's exposed will be back in a couple of years selling the same old same old to a receptive audience, and even if they're not there'll be a fresh crop, is beyond me.

But no, ya hafta keep at it. Honour demands.
posted by Devonian at 1:55 PM on March 6 [3 favorites]


I could see this being made into a feel-good comedy film, playing up stereotypes of New Zealand as a bucolic, trusting place where things move slowly, eccentrics and oddballs flourish quietly in every small town, and even the crooks aren't all bad.

Something like this exists: the journalist who wrote this piece made a documentary called Tickled, which is about the world of competitive endurance tickling. The trailers, which take this turn into tremendously sinister about halfway through, are... accurate.

Somewhat in this vein, I had a theory that cleverbot is just connecting random internet strangers to each other, telling each that the other is "cleverbot", letting them exchange a few sentences, then scrambling user connections so that it's hard for two people who might both suspect the truth to confirm it. I don't think that's really what cleverbot is doing, but I kind of want to see what such a fake cleverbot would be like.

You're actually very close: cleverbot remembers whatever a human responded for each of its responses, so you're really talking with someone cleverbot already talked to. It took a very long time for it to become convincing, but it has no way of actually working out what you're saying. (I forget what happens when you say something it's never heard before; I'd guess it just picks a line at random or something.)
posted by Merus at 3:17 PM on March 6 [2 favorites]


This is the inverse Turing Test, isn't it? A human being on the other end of an electronic communication connection trying to convince an investigator that he's merely an AI program.
posted by Harvey Kilobit at 3:52 PM on March 6 [2 favorites]


Oh hey, here's the web page for the Terrible Foundation. It's... a lot.

In addition to their Zach AI stuff and their proposed new Christchurch center, they're apparently also involved in providing internet service. And providing grants with a focus on various social justice-y themes but among the things they won't fund are "Religious or Political Advocacy/Advancement", "Buildings or Land", "Capital Assets or Equipment costing over £1500", "Administrative Costs", or "Individuals". There's some kind of completely undefined project called SideKick that "takes away the drudgery and mundane". They have a weirdly large organizational structure, including 3 executive branches, 12 departments & offices (including a Department of Actuaries and a Championship Office), 2 agencies (including a Forest Research agency), and 3 public corporations. They have a Records & Archives office that has no reports available. Also of note is that as far as I can tell, no where is there a single person's name mentioned on that website, definitely not on the page linked by "Who We Are".

As I said, it's a lot.
posted by mhum at 4:09 PM on March 6 [3 favorites]


A Potemkin website still takes effort, mhum :)
posted by k5.user at 4:13 PM on March 6 [1 favorite]


Also, according to the website, you can e-mail their San Francisco office. But only on Wednesday.
posted by valkane at 4:21 PM on March 6 [2 favorites]


Oh no, I don't doubt at all that it took effort. In fact, I'm a little amazed at just how much effort they put into this. Like, they didn't need to make a separate web page for each and every department, office, and agency (including the Adjudicator's Office, which for some reason is listed as an Agency not an Office) but they totally did. At the same time, I can imagine exactly the kind of person who would be taken in by this kind of sham website.

Oh yeah, I forgot to mention that they did put up some phone numbers on their contact page. I'm sorely tempted to dial them up to see who or what picks up on the other end.
posted by mhum at 4:23 PM on March 6 [1 favorite]


if you ever want to run a scam like this, please at least spend like a week reading Hacker News and TechCrunch so you'll get the buzzwords right.

for example, rather than focusing on the Cray XC50, they should have focused on the NVIDIA DGX, which is similarly a powerful special purpose computer, with fawning articles in the tech press, fancy industrial design, and lots of brand recognition, but this one used specifically for deep learning. It also costs only like $70k, which is impressively large while still being plausible for a startup.

I agree that physicians would be ideal marks for this kind of scam, as it's pretty much the only area I can think of where you can get an education with a lot of scientific content, yet without even breathing the same air as anyone who might use a computer for statistical modeling.
posted by vogon_poet at 4:37 PM on March 6 [1 favorite]


I forgot to mention that they did put up some phone numbers on their contact page.

A friend of mine is to move to New Zeland within the year to start life anew (working for the government, in fact, though not in anything that Terrible is connected to, somehow).

Should send them this thread and suggest if they get bored waiting for the container of their household goods to arrive they spend the day ringing the numbers and reporting the results ...
posted by tilde at 4:57 PM on March 6


David Farrier's Twitter also points out that some of their website material is a direct copy and paste from the Ford Foundation's site.

This reminds me of my own run-in with a fake foundation and the Dodgy Duke who ran it - who moved on to do some dodgy investment scamming in NZ after doing some here in Aus. Hmm...
posted by andraste at 6:27 PM on March 6 [1 favorite]


Also, most of the language they use is unrelated to any other foundation I've encountered either in Australia or NZ. 'Grants & Outgo Office'. Outgo?!

They claim income of over $1.4 million most of which was 'donations in-kind'. Expenditure of $325k of which $200k is unpaid donations and the rest is operating costs. There's a statement in the 2016 report that "The charity made a conscious choice to withhold internally reviewed donations (represented as unpaid) till global market stability. The charity aims to remit these donation within the next 24 months. Due to the irregular revenues this step was deemed necessary by the Trustees."

There's some terrible grammar in their legal documents. The 2017 report states: This is apart of the charity's forward planing for both a Trump presidency, and uncertainty in the global market due to Britains exit from the European union; which upon advise taken is still unstable one year later .

It looks as if they have a massive amount of capital on paper but don't do much with it. I'm no longer up on NZ charity law and whether charities' income is tax exempt, but it would be interesting to investigate further.
posted by andraste at 6:58 PM on March 6 [1 favorite]


More like, "Seriously, WTF?? ... Oh, wait, this is actually how the world is. I know that" followed by a sigh.

I'm sure there's a German word for that precise emotion.

If not, there absolutely needs to be.
posted by flabdablet at 7:04 PM on March 6


Since a bunch of people have spoken up about how wild and unlikely this is from the tech side, let me chime in and say that it’s just as much horse shit from the medical point of view. When I write my documentation, I’m trying to capture exact quotes from the patient as much as I can. I’m adding everything I can remember about the context of our interaction and the exact activities I completed. Sure I often take vitals and blood sugars and other basic values that an AI could record, but that’s not the main point. The main point is more nuanced than even the most advanced AIs right now could capture.
posted by ActionPopulated at 7:43 PM on March 6 [4 favorites]


Look, the guy started a company called Terrible Ideas, I don't know what anyone expects.

Does this debacle remind anyone else of Forum2000? It was my favorite hoax of Tech Bubble I. You'd chat with AIs of characters like Ayn Rand and Nietzsche, and they'd give remarkably in-character responses. If you chatted enough, you were invited into a secret area.... where you got to chat as the characters, to unsuspecting naive humans. It was amazing.
posted by miyabo at 8:25 PM on March 6 [3 favorites]


> flabdablet:
"More like, "Seriously, WTF?? ... Oh, wait, this is actually how the world is. I know that" followed by a sigh.

I'm sure there's a German word for that precise emotion.

If not, there absolutely needs to be."


Niedergeschlagenes zynisches Zeichen? (Machine translation for "depressed cynical sigh")?

Also, since I don't get to use this enough...

Wir haben uns zu Meistern der Wissenschaft!
posted by Samizdata at 10:41 PM on March 6


I'm not sure a deep and abiding dissatisfaction with the failure of all the world's rubes to wise the fuck up actually counts as cynicism.
posted by flabdablet at 11:09 PM on March 6 [1 favorite]


Also, most of the language they use is unrelated to any other foundation I've encountered either in Australia or NZ. 'Grants & Outgo Office'. Outgo?!

In, out; come, go. Outgo is the opposite of income.
posted by scalefree at 11:21 PM on March 6 [1 favorite]


So according to their entry on the NZ Charities Register, the Terrible Foundation claims assets of $456,288,282.

These assets consist of:

- 100% of Terrible Talk New Zealand Limited, sole director Alberic Whale
- half the shares in Number Eighteen Limited, which has been around since 2016 and is classified as "Information storage and retrieval service - except library". Other shareholders are David Whale and Robert Sneddon-Smith
- 45% shares in Terrible Web Services, with the rest owned by Stephen Calvert and Alberic Whale

So those three things are collectively worth $450 million? Wow, they must have a massive market presence, right? Well...

Terrible Talk has a website but bugger-all info on their services, no pricing or plans; you're supposed to give them your info and they'll contact you. Found no website for Terrible Web Services. Lots of Terrible this and thats have been registered over the past 5 years, some removed from the companies register, others have few notices from the NZ government gazette that they might be removed from the companies register for not responding to requests for information.

My most charitable reading is that it's a kid with big ideas, a genuine desire to help and an inflated sense of his own capacity, but I don't think that's what is going on here.
posted by andraste at 11:24 PM on March 6 [1 favorite]


My most charitable read: this is an attempt at performance art.
posted by flabdablet at 11:30 PM on March 6 [2 favorites]


> flabdablet:
"I'm not sure a deep and abiding dissatisfaction with the failure of all the world's rubes to wise the fuck up actually counts as cynicism."

I had to come up with something that German MIGHT boil down into a term. Otherwise it's just more schadenfreude and backpfeifengesicht. Yes, I know THOSE terms but not much else Germanic.
posted by Samizdata at 11:35 PM on March 6


You want Weltschmerz. Or perhaps Dummkopfheitschmerz or something.
posted by acb at 2:22 AM on March 7


Weltschmerz is too general, but I think Dummkopfheitschmerz could be on the right track. How about Dummkopfubiquitätschmerz?
posted by flabdablet at 2:40 AM on March 7 [2 favorites]


Niedergeschlagenes zynisches Zeichen

You seem to have typed "sign" instead of "sigh". What you want is "Seufzen" instead of "Zeichen".

Dummkopfubiquitätschmerz

I would go with "Dummheit" instead of "Dummkopf", and "Allgegenwart" instead of "Ubiquität". So, something like Dummheitallgegenwartschmerz, or rather Dummheitallgegenwartresignation. But that's all rather artificial, so I would probably go with something like "Ignoranzresignation" (the act of resigning oneself to ignorance), or perhaps "Ignoranzallgegenwartresignation".
posted by erdferkel at 2:55 AM on March 7 [4 favorites]


It certainly comes over as a scam, and frustratingly, the article revolves around the very little info it has to go on (emails, 20 minutes, limited number of requests), so it could be summed in two paragraphs, but following on what Devonian wrote above, I'd say that, from a moral viewpoint:

- greed is what drives these scams generally. On both sides. As long as greed is the driving force of (some of) investors, there will be crooks who will try to profit. I'm beginning to see this progressively less as a problem and more as an unfortunate feature. I'm sure safe investors will beware.

- on a general economic level, all that is happening here is that the money produced by (central) banks after the last economic crash to jump-start economies is trickling (or gushing) down - in this case, the "scammers" will perhaps get rich from investor's money, but they will (hopefully) use that money to buy luxury goods, invest in other similar scams and perhaps bona fide investment projects, inflate the price of land and real estate some, and generally, spread the dough around.

What I do have a problem with, but that is an entirely different story, is pyramid scams like bitcoin.
posted by Laotic at 2:56 AM on March 7 [1 favorite]


Did you know that they have a phrase for everything in English?
posted by acb at 2:57 AM on March 7 [5 favorites]


... something like "Ignoranzresignation" (the act of resigning oneself to ignorance), or perhaps "Ignoranzallgegenwartresignation".

scheißekopfallgegenwartresignation
posted by oheso at 7:06 AM on March 7 [1 favorite]




It's a system called the enigma layer. What it does, in essence, is go, this is how the person has written, and this is how they're going to need the response, because this is how dumb they are or how clever they are. That is what makes it make mistakes.

Ja-heezus.
posted by soren_lorensen at 7:32 PM on March 8 [3 favorites]


Zach was publicly unveiled in August last year at COCA, an art gallery in central Christchurch.
Performance art it is then.
"This investment is a huge vote of confidence in the city."
Monorail!
"We wanted something indiscreet, and this is indiscreet," Albi says.
You keep using that word...
posted by flabdablet at 7:39 PM on March 8 [4 favorites]


It's certainly an indiscretion.
posted by Dysk at 11:20 PM on March 8 [2 favorites]


Another follow-up, this from the the original author. A share prospectus, you say? Well gosh.
posted by Paragon at 2:27 AM on March 9 [5 favorites]


That stuff.co.nz article reads like a long form version of the Dead Parrot Sketch.
posted by ardgedee at 4:50 PM on March 9 [1 favorite]


At this point, one assumes MeFi's Own ( tm reg. ltd etc) cstross must be somehow involved, due to the various satire-excession issues with the plot as reported.
posted by mwhybark at 5:10 PM on March 9 [2 favorites]


The fun continues
posted by Paragon at 12:08 AM on March 13 [2 favorites]


He [David Whale] said recent media coverage of the Terrible Foundation was unfair and planned to recruit a public relations firm to handle media questions.

"I am over the whole lot of it because it is just stupid."
Why yes, yes it is. Not so much because of the reportage, though.
posted by flabdablet at 2:45 AM on March 13 [1 favorite]


In the follow-up article linked above by Paragon, there is an embedded Scribd doc of Terrible Foundation's rebuttal to the previous Stuff article. Here's their explanation for why Zach sometimes makes spelling errors:
The machine may make errors simply because it learns language phonetically but also because the occasional error made it appear more human. This behaviour is optional.

We ourselves make mistakes, typically because of our frequent use of dictation software. Many of our senior staff would qualify for Gold Cards.

We note there were several typos in the original story.
Well, all sorted then, I guess.
posted by mhum at 6:16 PM on March 13 [1 favorite]


It learns language phonetically?

If ever you want proof that these giggleswicks are making this stuff up on the fly, there it is.
posted by Devonian at 9:56 AM on March 14 [2 favorites]


And in related news, Theranos' CEO Elizabeth Holmes is charged with fraud.
posted by k5.user at 10:17 AM on March 14 [2 favorites]


It learns language phonetically?

It moves its lips as it reads. Bad habit.
posted by scalefree at 9:19 AM on March 21 [3 favorites]


It learns language phonetically, but also you can turn that off. Turning off the misspelling sounds like a pretty amazing AI feature. I wonder if the AI is so smart that you can ask it to stop misspelling things and it will, or if you have to send it through a whole new training round after which it won't be able to misspell things.
posted by clawsoon at 1:47 PM on March 21 [1 favorite]


They have been trying to expose it to a broader range of dialects before introducing it to markets outside New Zealand but whenever they attempt to correct its spelling and grammar it replies, "LOL get wrekd newb" and attaches pepe memes.
posted by ardgedee at 3:01 PM on March 21 [3 favorites]




« Older Spinning globe, moving continents   |   "When we look at a girl story, most of us go a... Newer »


This thread has been archived and is closed to new comments