The huge HIMSS annual conference and exhibition was on in Orlando, Florida this week and the news was almost completely dominated by announcements about how the big guns were incorporating AI into their software, particularly generative AI.

One that caught the eye was academic publishing giant Elsevier, which announced late last year that it was partnering with OpenEvidence, a company specialising in AI for medicine, to build a conversational engine for Elsevier’s widely used ClinicalKey clinical decision support tool.

Elsevier launched ClinicalKey AI at the end of February and was showing it off at HIMSS, explaining the importance of the need for high-quality evidence-based clinical content, which Elsevier just happens to have in abundance, and the value of a retrieval augmentation generation (RAG) that reinforces clinical precision, which we don’t understand.

Also out and about was Epic, which announced its partnership with Microsoft to integrate large language model tools and AI into its EMR only just year. Now, it has 60 use cases under development and is also planning to integrate Nuance DAX Copilot into the software, with some of the largest healthcare system in the US planning to roll it out. There’s a comprehensive story here.

Salesforce was also boosting its Einstein Copilot, which is expected to be available later in the year and is mainly aimed at administrative and operational tasks; and Philips and AWS were talking about their work on generative AI for Philips’ HealthSuite PACS as well as digitising pathology slides in the cloud.

Meanwhile, Google Cloud launched its Vertex AI Search for Healthcare, which enables genAI search on a broad spectrum of data, including FHIR data and clinical notes, and which Australian company Magentus hopes to explore under its new agreement with Google Cloud.

There was also a speech by Mayo Clinic Platform’s John Halamka on the risks and benefits predictive and generative AI used in clinical settings. Dr Halamka’s team tests commercial and self-developed algorithms at its lab at Mayo and he is also a founding member of the Coalition for Health AI (CHAI), along with Amazon, Google, Microsoft and CVS Health, with a goal to develop “guidelines and guardrails” for health AI systems.

He also wrote a blog last month on whether the future of AI is in good hands, a title we’ve pinched for our poll question this week. It follows some other big AI news this week, with the EU adopted its landmark and much discussed AI Act, which will categorise AI systems into four tiers based on their potential risk to society.

As we wrote last year, the AI Act’s risk-based approach to AI regulation sets out harmonised rules for the development of AI systems and their placement on the market. The framework includes four risk tiers: unacceptable, high, limited and minimal. Medical devices are deemed high risk.

The AI Act was a number of years in the making, but as this AP story makes clear, “the astonishing rise of general purpose AI models, exemplified by OpenAI’s ChatGPT, sent EU policymakers scrambling to keep up”. It will be fascinating to see if the new act is able to keep up itself. The implications for healthcare are detailed here, and for consumers here.

AI is going to be a big topic at next week’s Australian Healthcare Week conference in Sydney, with a whole day and stage dedicated to it. A generative AI masterclass on what AI is and is not, how far can AI take us and how soon, and how to develop successful proof of concepts for AI use cases will be on show.

Speakers also include Healthdirect Australia’s chief medical officer Nirvana Luckraj on lessons from designing and scaling AI-powered clinical decision support for triage; and NSW Department of Customer Service chief data scientist Ian Oppermann on “AI: the elephant in the doctor’s room”. Pulse+IT will be out in force so look us up.

That brings us to our poll question for the week:

Is the future of AI in healthcare in good hands?

Vote here, and leave your comments here.

Last week we asked: Is the new funding for NHS IT likely to achieve the stated goals?

A big no from readers: 82 per cent were negative. We also asked If yes, could you explain how? If no, is it just another pipe dream? Here’s what you said.