#17: the AI show continues

Welcome back.

Firstly, it seems LinkedIn had a wobble last week and sent #16 as a Pulse article rather than a newsletter. It messed around with the draft too so I suspect it just had a bad day, but for that reason it won't be in the archive and I've no idea who saw it.

Not sure what's up this week, but the more I send these newsletters out, the more I start to think "rinse and repeat"; as much as health seems to be well-invested, these newsletters seem to be a cycle of:

• AI will either save us or land us in peril; it shows early promise for certain narrow areas but could also be years away in others;
• Startups have great ideas, but can't always get funded, although some do, and take a while to get traction in the health system;
• The "interesting" things that get noted on here, tend to follow the same themes and (sadly) might get repetitive - there's a classic example in LLM rollout, later on.

So from next week I'll start mixing things up and begin to insert a few stories from people I've met along the way, maybe focus on a particular product or founder, and if anyone wants to take a slot doing that, I'd love to hear from you.

Meanwhile, it's gratifying to hear someone say the near future of healthcare might involve QR codes...

Onwards.

We'll start with another early "word from our sponsors" this week - team Waracle will be represented at Digital Healthcare World Congress this coming week in London so if you're there, please do look out for another knocking presentation from Gary Crawford and feel free to mingle with Niaz Rizwani and Mike Miller who would love to talk digital health with you, large and small. Seems tickets are still available too!

Going back to my first point, this week's updates seem a little bit like the AI show, but in the reading I've done this past week it's practically everywhere one way or another. In a dozen regular content searches on Feedly AI must make up at least 60% of the posts, and while plenty is just noise, a lot of it shows some progress and maybe some lessons.

But firstly a post from Panagiotis Kriaris on LinkedIn provided a decent ready-reckoner on where the banking world might utilise AI, and some of it is analogous to healthcare.

If I were to summarize how AI will impact banking going forward, the answer would include 3 key principles / domains:

Efficiency (driven my less manual work and increased automation, including decision making); forecasting in the sense of better predicting customer patterns and managing the outcome and helping banks deliver highly personalized experiences on a massive scale.

It would be no surprise that one of the lower-hanging items on the list is "call transcript analysis" which has come up several times in Nuggets over the past few months. Just this past week Augmedix and Mutuo Health Solutions were in the news showing off more services regarding medical note-generation.

This has come up in Nuggets #1, #5, #9, #12 and #13 just to show how many companies are operating in the space - which should come as no surprise given speech recognition and language processing came way before the current LLM craze; the latter simply adds a layer of accuracy not possible before. It should be a surprise if you are not transcribed by an LLM-enhanced service in the coming year in some kind of service industry.

What might hold that back? Well, as always, regulation in health is not straightforward and there's a yawning gap between adoption at the product development end, and acceptance in the market.

A year ago, developers would probably have scoffed at using GenAI coding tools, in favour of their own experience, but just to use one example, this past week Shopify revealed 70% of its developers are now doing so routinely and a million lines of its codebase came via that route. It's quite likely products you use now, are being built with the heavy assistance and efficiency of AI.

We heard this past week that a major global health company, usually pretty conservative and heavily regulated, now uses Github CoPilot routinely.

So the inputs are moving, what about outputs? For those who have worked in medical device development, how do you ensure that AI-generated solutions leave an audit trail compatible with a design or technical file?

Turns out a paper in Nature has been tackling that very subject:

You might go into this thinking AI might be part of the answer as much as the solution, but summarising the paper, it seems certifying AI is really no different than any other software as long as you have a suitable quality management system with the right people driving. So this ought not to be an issue. For those developing SAMDs with or without AI - 13485, CE/UKCA, FDA, etc - this article is a very interesting read.

So if the developers are game for AI and the technical regulatory side isn't desperate to hold you back, what's stopping a rapid march to adoption now?

According to a piece in Digital Health this past week, quoting Dom Cushnan, the NHS Director of AI, Imaging and Deployment, the answer may well be as simple as gathering sufficient evidence that it makes genuine improvements.

After all ChatGPT will only celebrate its first anniversary this coming week and the hype cycle in its wake, whilst breathless, doesn't exactly leave a body of scientific evidence for LLMs generally.

I think the challenge that we all face is, how do we get to a point where we can realise the benefits of the technologies that we want to use?

From the engagement that we’ve heard, most people would like some form of tooling that can help them get through work faster. And that’s on everything, whether you’re a policymaker and you’re thinking about what generative AI looks like, right through to a doctor on the ward or a nurse or an allied health professional.

And maybe the final barrier to adoption is simply no more than turning that growing evidence into trust. A recent survey showed that clinicians can see the benefits and are generally supportive of using AI, but still (rightly) worry about the headlines showing off hallucination and similar issues.

According to a worldwide survey of 2,607 clinicians by Elsevier Health, 48% support the idea of using generative AI tools to support clinical decision-making, underscoring the desire of many clinicians to spend more time on patient care and less on burdensome admin work. view article here.

Yet issues with these tools fabricating or ‘hallucinating’ information means that generative AI’s suitability for clinical applications is far from certain.

Some evidence of why that uncertainty exists, might be based on evidence to date. Another article this past week explained how clinicians should not be too worried for their jobs just, as LLMs are eager to please but soon admit fallibilities when pressed:

The potential of AI approaches to assist in neurological research, education, diagnosis and care is enormous. Even so, AI lacks judgment based on real-world experience and the intangible information derived directly from patients and the clinical situation.

AI has a lot of potential but still requires oversight. For now our jobs appear safe.

But of course AI does not simply live in LLMs and 飞利浦 were in the news receiving sizeable funding from the Bill & Melinda Gates Foundation to advance their AI-powered ultrasound service which provides very detailed information in a far more portable and accessible format - particularly for developing countries.

Philips developed its AI to help frontline health workers, such as midwives, identify possible complications during pregnancy.?

The tool, which uses informatics, digitization and AI to supply operators with an interpretation of image results, was utilized during a trial period in Kenya?with positive results.?

The Kenya pilot underlines the power of the solution we are developing. Within hours the healthcare professionals can conduct ultrasound scans to triage pregnant women on six critical parameters. Normally, this would require weeks of training.

Just to illustrate how the current AI boom is really no more than a continuation, this particular effort started some years ago and is only now starting to realise its potential. However it also illustrates the potential timescale and funding required to deliver such advances to market.

Dwelling on Africa for another moment and it's good to see early signs that developers in the continent might not get left behind. Maybe this time, the ubiquity and availability of cloud platforms should act as a leveller for people anywhere to get involved early.

Primary schools should integrate AI into their curriculum, addressing societal challenges through technological solutions. We must develop curricula that adapt to the present and anticipate the future.”

Elsewhere... If you're male and my age, it's quite likely your social media might have had companies like Hims and Numan offering treatments that bring your hairline down and, er, perk up other things. It's clearly big digital business, as Hims reported 57% year-on-year growth in the last quarter at a run rate of nearly $1 billion a year.

It would therefore come as no surprise that they too are looking at their backlog for AI-driven opportunities, and a new service called MedMatch which pairs anonymised data collected from customers to likely treatments that may suit them:

Currently in its beta testing phase, MedMatch utilizes artificial intelligence and machine learning, drawing on a large, anonymized dataset from the Hims & Hers customer base. This technology aims to quickly identify treatments that best match individual patient needs. Its initial application is focused on assisting customers seeking help for anxiety and depression on the Hims & Hers platform.

Hims and Hers, despite the high revenue, still posts sizeable (although narrowing) losses each quarter and according to Crunchbase, has raised $233 million to get itself this far. No doubt AI is as much an expectation for investors as a product necessity.

But again as mentioned in #5, #7, #9, #12 and #14, AI and investment seem to go hand-in-hand at the moment. Again in Digital Health a report noted a tectonic shift from investments in the pandemic period, mostly around telemedicine, wellness and remote workplace solutions, to more advanced health, and particularly AI companies. And notes crucially that although there may be slightly less money on the table before, the fewer companies asking for it are better-organised and more investable.

In the digital pharma sector, similarly, tech bio and synthetic bio, of which 86% of the latter are UK-founded companies, have generally outperformed investments in patient recruitment for clinical trials and digital biomarkers over the same period. In the case of digital biomarkers, a number of companies in this area have moved across to the digital care space.

Meanwhile, some of the best-funded at the height of the pandemic – including teleconsultation, digital therapeutics and workplace wellness – have seen a decline in investment. Both teleconsultation and workplace wellness have struggled to expand outside of Europe. ?

Speaking to people in the UK health space, there is clearly money out there for the right founders and businesses - particularly in AI-driven health - but no doubt it'll be watched carefully as the hype cycle unfolds.

A(nother) word from our sponsor, ie. the company that pays me through the week so I can waste Sundays on this ??

There's another fascinating article subtitled "The app will see you now", penned by my colleague Niaz Rizwani , which is well worth a read as a flavour of what we do and how we think about health at Waracle . Have a read!

Small print: This newsletter goes out to subscribers and across LinkedIn most Sunday nights around 7:30 pm. Feel free to contact me if you've seen or are creating something interesting in digital health. I work for?Waracle, but all opinions and content selections are my own. Anything in which I have a work or personal interest will be declared.

Cover photo was generated by Playground AI using the simple prompt "the app will see you now".

2000 words again this week after saying to Kevin Stewart there wasn't much to write this week ?? - well done if you got this far.