Industry Watcher’s Digest

Buzzworthy developments of the past few days.

• Patients’ trust in generative AI for healthcare may already be fading. When Deloitte asked 1,000 adults who aren’t using the technology what’s keeping them away, 30% said they don’t trust the information. That’s a sizeable rise over the 23% who gave that answer last year, when the concept of gen AI was still new to consumers. Further, the 2024 results show consumer use of gen AI for reasons related to health and wellness at 37%. That’s down from 40% in 2023. On the other hand, the general public has remained “overall optimistic” about gen AI’s potential to help chip away at healthcare challenges like access and affordability. And two-thirds of 2024 respondents still hope it will help cut wait times for medical appointments and reduce out-of-pocket costs. More results here.
   
• Providers’ feelings are similarly mixed. McKinsey surveyed 200 executives of health systems and found 75% pressured to speed up digital transformation but stymied by shortages in resources and planning. More than half the field, 51%, named budget constraints among their top obstacles to investing at scale across all digital and AI categories of interest. And current investment priorities are frequently misaligned with areas they believe could have the most impact. All of which is unfortunate in light of separate research showing that AI, if implemented wisely, could cut global healthcare spending by $200 billion to $360 billion. More here.
   
• Three heavy hitters in healthcare technology are coming together over AI and quantum computing. Cleveland Clinic, IBM and the U.K.’s Hartree Centre announced the collaboration June 6, saying the work will kick off with two clinical research projects focused on care for epilepsy patients. From this will follow the development of “larger AI models that can integrate multiple types of data for analysis across different diseases.” The collaborators say their “ultimate aim” is to improve patient care and biomedical science. Announcement.
   
• Data suitable for training generative AI is a finite resource. The data mainly exists as words written by humans and shared online. Like California gold in the mid-1800s, it’s being mined too fast by too many. This is not news, but a new scientific study has the supply running out as soon as 2026 and no later than 2032. “At this point, the availability of public human text data may become a limiting factor in further scaling of [large] language models,” write lead author Pablo Villalobos and colleagues at the Epoch AI research institute. But wait. Don’t despair. The team is sanguine about the ability of innovative people to compensate for the coming depletion. “[A]fter accounting for steady improvements in data efficiency and the promise of techniques like transfer learning and synthetic data generation,” they write, “it is likely that we will be able to overcome this bottleneck in the availability of public human text data.” Study here (preprint), AP news coverage here, Time background coverage here.
   
• Apple-only loyalists can have their Apple experience and Microsoft-backed OpenAI too. That’s because Apple Intelligence, the new AI system imbuing iPhones and Macs with generative AI capabilities, is including access to ChatGPT among its selling points. Apple’s aim is to make AI a part of everyday life—not just for “Apple snobs” but also for hybrid users who split their time between digital ecosystems. If you’ve followed any tech news since the announcement June 10, you’re aware. But you probably haven’t heard it all yet. Apple announcement here, media coverage everywhere.
   
• European AI for European people. Should be an easy sell, no? Mark Zuckerberg’s Meta—he’s still the majority shareholder—has sent 2 billion notifications and emails to European users disclosing its desire to tap their public online postings for tailoring AI models. Meta wants to start the sweep June 26. The messaging tells recipients the company’s aim is to better serve Europeans. “Meta is not the first company to do this—we are following the example set by others, including Google and OpenAI, both of which have already used data from European users to train AI,” writes Meta global privacy director Stefano Fratta in a June 10 blog post. “Our approach is more transparent and offers easier controls than many of our industry counterparts already training their models on similar publicly available information.” The push (but not necessarily pushy) messages include an opt-out link. Europeans tend to be more privacy-minded than Americans, so all these efforts will be interesting to watch.
   
• How about Stateside? Many Americans are no more inclined than other cautious citizens of the digital world to have their online keystrokes sucked up and repurposed by AI giants. It’s for them that The New York Times has published a piece in the service journalism vein headlined “Can I Opt Out of Meta’s A.I. Scraping on Instagram and Facebook? Sort of.”
   
• May I call you Thunder Trunk? Dolphins are known to call out to other dolphins by aping other dolphins’ signature sounds. Parrots do this too. And now it turns out elephants take intra-species conversations a step further—“using individual names that they invent for their fellow pachyderms.” The behavior may have remained undiscovered if not for AI.
   
• Recent research roundup:
   
  • Johns Hopkins/SNMMI: New AI tool accurately detects six different cancer types on whole-body PET/CT scans
     
  • University of Maryland: What makes roads safer? New UMD study uses AI to find out
     
  • U of Michigan/Brown U: Novel AI method could improve tissue, tumor analysis and advance treatment of disease
     
• From AIin.Healthcare’s news partners:
   
  • Health Imaging: Large language models make radiology reports more patient-friendly
     
  • Cardiovascular Business: Embracing AI to enhance EP: Insights from cardiologist Jagmeet Singh
     
  • Health Exec: UW Health pilots AI for patient communications