, 47 tweets, 44 min read Read on Twitter
'Bad Bot' (part II) by @DrMurphy11

The story of a dangerously flawed #eHealth AI Chatbot.

A THREAD about #HealthTech #Hype, #AI #Governance & #MedicalDevice #PatientSafety.

Read on, or see the single tweet summary👇0/45
This thread follows on from a prior thread ('Bad Bot' - part I) & picks up the Babylon story from the #BabylonAITest event which was held on the 27th June 2018.

'Bad Bot' (part I) - 👇 1/45
The #BabylonAITest event was held at @RCPLondon & promised to reveal to the world, Babylon's latest "ground breaking research" & "the most exciting advances in Artificial Intelligence".

Dr Murphy was expecting Big Things from @babylonhealth...

The #BabylonAITest event kicked off with Ali Parsa talking about the companies 'mission statement'.

IMO - the companies focus on accessibility & affordability - has been to the detriment of ensuring #PatientSafety - which is the focus of this thread.

Dr Butt took the stage for the BIG #BabylonAITest reveal.

Babylon claimed, their #AI #Chatbot was more capable of passing the diagnostic component of the post-graduate @rcgp exam, than a typical UK GP trainee.

Sounds good doesn't it.
Maybe too good to be true?

Dr Butt 👇4/45
During the event, Babylon also demonstrated the science behind their AI, using some highly advanced wavy lines.

The message was clear. If you want to have good health, you had better follow the guidance of the Babylon AI...

Scientific wavy lines👇5/45
Ali Parsa (Babylon CEO) closed the #BabylonAITest event stating; the Babylon AI Chatbot could diagnosis patients as reliably & safely as a doctor, and the 'diagnostic' Chatbot was live & available to all...

Ali Parsa's sales pitch 👇6/45
Babylon took to @Twitter with their claim that the @babylonhealth AI Chatbot had the "safety & accuracy of a real-life doctor!"

Babylon also released a 'paper' that outlined the 'research' presented at the #BabylonAITest event.

On reviewing the @babylonhealth 'research paper' - significant flaws in the methodology were evident.

Was the 'diagnostic' AI Chatbot actually any good, or was the #BabylonAITest just an elaborate PR exercise?

Dr Murphy's non-academic review 👇8/45
Other twitter users shared their opinion of the #BabylonAITest 'paper'.

Including Professor @EnricoCoiera (trained in medicine with a computer science PhD in Artificial Intelligence), author of "Guide to Health Informatics".

The Enrico review 👇9/45
Dr @EricTopol (currently leading The Topol Review: preparing the healthcare workforce to deliver the digital furture) also provided his opinion.

Eric Topols tweets 👇10/45
Was the new, much hyped, Babylon 'diagnostic' AI Chatbot as reliable & safe as @babylonhealth claimed?

Or was this as misleading as their prior '100% safe' Chatbot claim.

Dr Murphy put the AI Chatbot to the test.

A prior misleading claim👇 11/45
Test 2 #NoseBleed triage.

Chatbot diagnosis = it REALLY doesn't know what to do with a nosebleed.

Result = AI fail.

Bizarre #NoseBleed triage 👇13/45
Test 3 #PE triage.

Chatbot diagnosis = #Costochodritis & #BrokenAnkle.

Result = AI fail.

Another #DeathByChatbot 👇 14/45
The #BabylonAITest was quickly becoming a PR Car Crash.

A statement from the Royal College of General Practitioners @rcgp, rebuked Babylon’s claims as “dubious" - & the online media hype was replaced with more objective reporting.

Media coverage 👇15/45
Despite the negative coverage - the Babylon PR machine continued to disseminate the over-hyped promotional claims.

"Babylon’s AI is on par with doctors"

Misleading Babylon promotional email 👇16/45
There was further embarrassment for @babylonhealth when @RCPLondon - who according to Babylon collaborated in the #BabylonAITest event - also released a statement to distance themselves from the dubious claims.

Another bad day for Babylon 👇17/45
Babylon's PR Car Crash continued - with their Chief Scientist clarifying that the data presented at the #BabylonAITest event was in fact 'preliminary' & from a 'pilot study' - hence NOT practice changing. 🤔

More red faces at Babylon👇18/45
It appeared that both the diagnostic abilities of the Babylon AI Chatbot, & the AI Governance within @babylonhealth were not robust.

Of concern, it appears they had also cut corners with the @babylrwanda AI system deployed in Rwanda.

BBC Click 👇19/45
Then - on the 11th July, the @DailyMailUK dropped this bombshell...

"The new Health Secretary, @MattHancock - uses a controversial smartphone app, @babylonhealth - that offers virtual consultations with a GP."

Dr Murphy needs counselling 😕 20/45
Babylon were also in the news due to their involvement in the controversial @George_Osborne @EveningStandard - “money-can’t-buy” #FutureLondon advertorial scheme.

Babylon's PR panic👇21/45
However, even the well-oiled Babylon PR machine, couldn't keep the Safety Concerns regarding the AI Chatbot, out of the mainstream news media.

The @FinancialTimes reported on the flawed reality of the unvalidated @babylonhealth AI Chatbot...

Babylon's response to the reported #PatientSafety concerns was both dismissive & infantile.

@babylonhealth, claimed the company was the victim of Twitter #Trolls, motivated by vested interests...

#PatientSafety #Troll 👇23/45
At least the Babylon PR team could still rely on the @EveningStandard for some positive media coverage.

The benefits of an Advertorial deal 👇 24/45
Although Babylon had claimed their 'diagnostic' AI Chatbot was 'on par' with GPs.

In reality, the @babylonhealth Chatbot was flawed.

Some of the Chatbot triage algorithms could cause patients unnecessary distress.

e.g. Flawed #UTI triage👇 25/45
Other flawed AI Chatbot triage algorithms were amusing enough to make it into the @MailOnline.

e.g. Flawed #NoseBleed triage 👇 26/45
Worryingly, 18 months after #PatientSafety concerns had first been raised with the @CareQualityComm & @MHRAdevices.

Dangerously flawed AI triage algorithms, continued to be a risk to patient safety.

e.g. #Cardiac #ChestPain triage👇27/45
Given the obvious flaws in the @babylonhealth #ChestPain algorithm - some of the media reports were a little alarming.

#DeathByChatbot👇 28/45
The clinical errors in the Babylon AI algorithms were so fundamental, they raised questions regarding the Corporate & Clinical Governance within @babylonhealth.

Could they be trusted to deliver healthcare?

#DeathByChatbot 👇 29/45
Although Ali Parsa talked of the need for 'testing of AI technology to ensure people aren't harmed'.

He hadn't tested his own Chatbot!

Bizarrely, he also intimated @MHRAdevices regulations & @CareQualityComm were a 'barrier'. 🤔

Listen to Ali👇30/45
The Babylon story then went from Bizarre to Surreal...

The new Health Secretary, @MattHancock, used his first speech to promote #HealthTech & specifically @babylonhealth @GPatHand 😲

Matt Hancock supports Babylon's Bad Bot👇31/45 bbc.com/news/health-44…
However, even with @MattHancock's support - the Babylon AI Chatbot remained dangerously flawed.

Sudden onset #ChestPain & #Breathlessness = #DeathByChatbot

Another Bad Chatbot triage👇32/45
In some cases, the @babylonhealth AI Chatbot would suggest completely random diagnosis.

For example; Productive Cough = #MultipleSclerosis. 🙄

This Babylon AI Chatbot was an #eHealth liability.

Would you trust this Bot👇33/45
Concerns regarding the safety of the Babylon AI Chatbot were raised in parliament - with @sarahwollaston questioning @MattHancock regarding the safety & the regulation of #eHealth Apps.

The Bad Bot debates👇34/45
The Babylon AI Chatbot safety concerns were also discussed in @CommonsHealth.

@MattHancock didn't appear to have a good understanding of the potential risks to #PatientSafety from flawed unvalidated AI #HealthTech.

The Bad Bot debates👇35/45
Meanwhile, the Babylon AI Chatbot continued to struggle with common presentations.

For example; Gastroenteritis =
#LactaseDeficiency 🤔 or
#MultipleSclerosis 🙄 or
#ErectileDysfunction 😕

Don't #AskBabylon👇36/45
The @babylonhealth AI Chatbot also struggled with simple #Cellulitis.

Remember; this Chatbot was promoted as being 'on par' with practicing doctors. Apparently it also helps to deliver healthcare in Rwanda via @babylrwanda!

Trust me, I'm a Chatbot👇37/45
Despite the readily identifiable flaws in the Babylon AI Chatbot. @babylonhealth are reported to have licensing deals/partnerships with @BupaUK, @SamsungMobile, @prudentialplc, @TencentGlobal & most recently @TELUSHealth.

An #eHealth AI🍋? 👇38/45
In stark contrast to the razzmatazz, hyperbole & staged applause of the #BabylonAITest event; in reality, the @babylonhealth 'diagnostic' AI Chatbot appeared to be a flawed, unvalidated #eHealth liability.

#BabylonAITest - Astounding or Snake Oil? 👇39/45
Just 6 weeks after the 'hype' of the #BabylonAITest event; the dangerously flawed 'diagnostic' Chatbot was withdrawn from the UK @babylonhealth App, & replaced with a flawed 'triage' Chatbot.

RIP - Babylon's 'Diagnostic' Chatbot👇40/45
There's been no comment from @babylonhealth to explain why the 'diagnostic' AI Chatbot was withdrawn from the App.

It raises the question - was the Chatbot & the #BabylonAITest event just a #SnakeOil PR gimmick?

#MedicalDevice #AI #Quackery👇41/45
We really need to understand how @babylonhealth managed to promote - a flawed, unvalidated, non-@MHRAdevices approved #MedicalDevice at @RCPLondon with @NHSEngland in support?

Did the Babylon Corporate Culture & Hype stifle Clinical Governance?
Artificial Intelligence has no inherent conscience, nor moral or ethical standards.

It is therefore 'critical' - that those involved in the development of AI technologies do...

How do we regulate for that?
#Corporate #Ethics
#AI #Governance 43/45👂here👇
Babylon appear to have adopted the Silicon Valley Tech Culture of 'Fake it till you make it' & 'Move fast and break things'...

Overblown hype & unvalidated #HealthTech - has no place in healthcare.
#MedicalDevice #Quackery
#Theranos #BadBlood
Babylon’s 'move fast' approach has made them the Poster Child for poor #eHealth AI Governance.

However, we can be optimistic, with many examples of 'truly' innovative AI #HealthTech on the horizon - @DeepMind_Health, @KheironMedical, #IDxDR etc
Thread T&Cs: These tweets are based on publicly available information. Please tweet or DM any queries regarding there factual accuracy.

Thanks for reading. If you have found it useful then please click below & retweet the 'Bad Bot' thread 👇
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Dr Murphy
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!