Healthcare AI Guy Weekly Newsletter | 6/20

Our Picks ✨

Highlights if you’ve only got 2 minutes…

McKinsey estimates AI could add up to $4.4 trillion to global economy

According to a report from McKinsey & Company, the productivity gains from generative AI could add $2.6T to $4.4T of value to the global economy annually.

The report suggests that by 2030 to 2060, half of all work could be automated — the power of generative AI tools has accelerated this forecast compared to previous predictions.

McKinsey also claimed that these changes “will challenge the attainment of multiyear degree credentials,” meaning that knowledge workers may be the most likely to take a career hit… Five key takeaways from the study:

1) ~75% of the value that generative AI use cases could deliver falls across four areas: Customer operations, marketing and sales, software engineering, and R&D.

2) Banking, high-tech, and life sciences are among the industries that could see the most significant impact as a % of revenue from generative AI. Healthcare is 6th on the list! (image below)

3) Current generative AI and other technologies have the potential to automate work activities that absorb 60-70% of employees’ time today.

4) The pace of workforce transformation is likely to accelerate, given increases in the potential for technical automation.

5) The era of generative AI is just beginning. (link)

EU leading AI regulation push

The European Parliament approved the EU AI Act, the world’s first comprehensive AI law which would ban or limit certain AI applications. It places AI systems into different 'risk' buckets with different rules for each. Systems with “unacceptable risk” (think manipulative AI or real-time face recognition) will be banned. “High-risk” systems (those in safety-critical areas or specific sectors like law enforcement) will need special registration and checks. And “limited risk” systems just have to be nice and clear about what they're doing. The legislation also requires companies to publish summaries of copyrighted data used to train their AI tools.

Some checks and balances are good (with great power comes great responsibility) such as labeling AI-generated content, but this Act likely limits the EU from becoming a potential AI innovation hub. The EU's aggressive stance towards AI also puts it on a collision course with American companies, with OpenAI warning that it may be forced to pull out of Europe if legislation is too restrictive.

Separately in the UK, OpenAI, Google DeepMind, and Anthropic just granted the UK government priority access to their AI models source code for research and safety purposes. UK Prime Minister Rishi Sunak emphasized the need to understand the opportunities and risks of these systems and build better evaluations and precautions. In addition, Google delayed its EU debut of Bard, pending an examination of the tool’s data privacy practices by regulators. (link)

AI discovers medicine that can combat ageing

AI has been used to identify three potential drugs that could combat ageing. The drugs work by safely removing defective cells linked to conditions like cancer and Alzheimers. The team developed a machine learning model using data from previous studies and then used the model to screen more than 4,000 chemicals, identifying 21 potential drug candidates for experimental testing.

Using AI was hundreds of times cheaper than standard screening methods the researchers said.

In the Mckinsey report (mentioned above), ‘pharmaceuticals and medical products’ was 3rd on the list in terms of industries that could see the most significant impact from AI as a % of revenue. This AI drug discovery announcement demonstrates how AI can revolutionize healthcare and life sciences by speeding up and reducing the cost of drug discovery. The combination of data scientists, chemists, and biologists can be a magical thing! (link)

Physicians using AI to improve empathetic communications with patients

A New York Times article reported that physicians are having success using chatbots “to find words to break bad news and express concerns about a patient’s suffering, or to just more clearly explain medical recommendations.”

For example, University of Texas at Austin Department of Internal Medicine Chairman Dr. Michael Pignone used ChatGPT to write a script for a program to improve “treatments for alcohol use disorder.” ChatGPT successfully produced a useful script written “at a fifth-grade reading level” and the project’s co-principal investigator, Dr. Christopher Moriates, said the “fifth-grade level script...feels more genuine.”

There’s definitely still a long way to go for AI to become true clinical decision support but examples like this highlight how the technology can help providers simplify and improve communications. Another example this week comes from UC San Diego Health: “ChatGPT will generate a draft response for physicians when they reply to a patient message through MyChart. In addition, all messages written by ChatGPT and sent to the health system’s patients will be marked as ‘automatically generated and reviewed by a physician.’” (link w/out paywall)

Miscellaneous 🔍

News, podcasts, blogs, tweets, resources, etc…

• WSJ piece on who makes the call: AI or care provider? (link w/out paywall)

• Found podcast ft Amy Brown from Authenticx: what it takes to mitigate AI biases in healthcare (link)

• Providers should use AI to automate admin support in health care (link)

• Centene CEO bullish on AI (link w/out paywall)

• Health system leaders discuss how to disclose AI use to patients (link)

• Can AI make older adults feel less lonely? (link)

• How to cite ChatGPT from American Psychological Association. (link)

• “Age of AI” lingo (link)

• Microsoft’s CEO Satya Nadella is betting everything on AI (link w/out paywall)

• Accenture to invest $3B in AI and have 80,000 staff working on AI, after laying off ~19,000 in March 2023 (link)

• Contra Marc Andreessen on AI (link)

• Charting a path to AI accountability (link)

• Paul Graham, famous entrepreneur / VC, tweets a thread on 10x productivity boost with AI (link)

• New book on AI from co-founder of DeepMind and InflectionAI (link)

• Very relevant take on cost of models:

Deal Desk 💸 

Spotlight on latest capital raises, M&A and investments…

📈 Apixio + ClaimLogiq: Apixio, an AI company that supports health plans and provider groups in their delivery of VBC, merged with ClaimLogiq, a company focused on reducing claim errors prior to payment. New Mountain Capital, an investor in both companies, is leading the follow-on investment to enable the merger, while ClaimLogic backer Eir Partners will provide additional capital to support the transaction and retain a minority stake. Together, the combination of infrastructure, AI, and proprietary healthcare data will accelerate the shift toward outcomes-based reimbursement models by reducing costs, minimizing claim adjudication errors, and optimizing the delivery of both payments and care. (link)

📈 Medivis: a developer of AR surgical tools based in New York, raised $30M in Series A funding. Thrive Capital led the round and was joined by Initialized Capital and Mayo Clinic. Medivis is looking to establish augmented reality as the new standard in surgical navigation and has partnered with the following providers: MD Anderson Cancer Center, Providence Health, University of Pittsburgh Medical Center, Northwell Health, Cincinnati Children's Hospital, and the Veterans Health Administration. (link)

Tool Box 🧰

Latest on business, consumer, and clinical healthcare AI tools…

🔧 Google: Google Lens, a vision-based tool that allows users to search what they see via their smartphones, added a new feature: the ability to search for skin conditions. Essentially, Lens compares an uploaded or captured image to other images throughout the internet and ranks the results based on similarities. The application now enables users to take a picture of their skin (e.g., a rash or discoloration) and matches it with the most relevant result, and provides insight into the condition. (link)

🔧 LUUM: LUUM, a beauty studio in Oakland, CA, is providing robotic eyelash extensions. This is one of the first AI robots to be used in the consumer beauty space. LUUM says its robotic eyelash extensions are safe, cheaper and quicker to apply than traditional eyelash extensions performed by humans, which typically take 90 minutes to an hour to perform. LUUM is now taking preorders to sell the machines for $60,000 to other studios. Ulta Beauty is one of its investors. Is this a preview of dental cleaning or medical services robots to come? (link w/out paywall)

AI Images of the Week 📸

Funny memes and pics from around the web…