Transfixed by a terrible tall tale
In 2017, technology research firm Gartner ranked virtual assistants and machine learning at the “peak of inflated expectations” on its much-quoted emerging technologies hype cycle, but it also named AI as one of three big megatrends that will provide unrivalled intelligence and create profoundly new experiences in the next five to 10 years.
We were pondering this projection this week having read up on a stunning yarn that has utterly transfixed the health IT and the wider IT industry in New Zealand. In two seriously good stories in Kiwi independent site The Spinoff – here's the first, and here's the follow-up – Auckland journalist David Farrier took a look at a purported artificial intelligence technology called Zach, the creation of a supposed charity called The Terrible Foundation and its founder Albi Whale, which was set to revolutionise healthcare as we know it.
As Mr Farrier recounts, the Terrible people are allegedly trialling Zach as an AI assistant in a Christchurch medical practice, where it is being trained through natural language processing to create a summary of a doctor-patient consultation so the doctor no longer has to physically write clinical notes during or afterwards. The story was told in an article in New Zealand Doctor last year, and reading that article, it all sounded – on the surface at least – reasonably plausible.
The GP was named, as was his practice, and he himself sounded convinced and convincing in the interview. He heard about the technology while at a meeting with his PHO, Pegasus, when they were discussing the roll out of the South Island's HealthOne shared care record system. The GP decided to give Zach a go in his practice, obtaining patient consent for consultations to be recorded, and he was adamant that although he was corresponding with Zach by email, no identifying data was exchanged.
At that stage, a few eyebrows should have been raised. Email? To whom are these sent, and to where? There were also strange spellings and capitalisations in the resulting report, which appeared on second thought to be suspiciously human. But despite some eyebrow-raising claims – how does custom silicon, liquid nitrogen cooled supercomputers and the Cray XC50 as a “firewall” suit you? – the idea sounded promising and everything seemed reasonably legit.
That's until Mr Farrier started poking around, and what he found was a yarn for the ages. We recommend you take a look and behold the splendour of the peak of inflated expectations come crashing down into the trough of disillusionment right in front of your eyes.
Thankfully, it seems the Terrible tale was told before it got too tall, unlike the other big news this week, that of the disgrace of Elizabeth Holmes and her “revolutionary” pathology company Theranos. This one-time Silicon Valley darling promised to be able to run hundreds of tests using just a tiny pinprick of blood, cheaply and quickly. It signed up Walgreens at one stage and by the time the skivvy-wearing Ms Holmes admitted to massive fraud this week, it has raised $US700 million from deluded investors.
“The Theranos story is an important lesson for Silicon Valley," a US Securities and Exchange Commission official said in a statement. "Innovators who seek to revolutionise and disrupt an industry must tell investors the truth about what their technology can do today, not just what they hope it might do someday."
The digital disruption of the healthcare industry is a term many like to bandy about with the very best intentions, but we think it's hyped far too much. There is certainly the potential to do remarkable things with technology tools – take a look at our chat this week with Sydney Children's Hospitals Network CEO Michael Brydon, for instance – but it helps if they are real in the first place.
That brings us to our poll for this week: Is the potential of AI overhyped in healthcare?
To vote in our weekly polls, sign up for our weekend edition or leave your comments below.
Our poll last week asked: Do you think ADHA's Framework for Action is too ambitious? Straight down the middle on this one: 50% yea, 50% nay.