The White House’s Office of Management and Budget recently released final guidance for the use of AI by executive-branch departments and agencies. OMB director Shalanda Young fleshed out the particulars of the guidance in a 34-page memo sent to heads of the affected bodies March 28 (summarized and linked here).

Less than two months later, a former IT administrator for the OMB is taking issue with the execution of the instructions. “[I]t is very clear that the new OMB memo has a major shortfall in guidance for using AI to improve government results,” writes Mark Forman in commentary published by the Federal News Network.

Today Forman consults governmental clients on technology with the Washington-area firm Dynamic Integrated Services. In the early 2000s he was appointed by President George W. Bush to be the first federal CIO (which was then called the OMB administrator for e-government and IT).

In his May 22 opinion piece, Forman acknowledges that Young’s OMB faced an uphill battle trying to launch fresh AI strategy in the fourth year of a presidential administration.

Addressing whomever takes up the mantle in the next administration—which could well be Young again, should Biden win reelection—Forman offers several tips and perspectives. Among the arguable best are these three:

1. The public will best benefit from government AI initiatives tied to strategic needs and opportunities.

“Based on what we know about the agency AI initiative submission to the White House last year, about a thousand AI projects are underway in government,” Forman writes. “Those projects range from good old-fashioned algorithmic AI to robotic process automation to neural networked generative AI tools and chatbots.” More:

‘We should not spend billions of dollars on AI projects that are compliant with OMB guidance and are merely interesting from a research perspective. Rather, with so much effort on AI projects, we need a return on investment tied to the biggest strategic needs and chronic problems of government.’

2. The government could improve efficiency, effectiveness and responsiveness from using a ‘know your customer’ approach.

Agencies might adopt means and methods from the private business sector, which uses AI as a tool in customer relationship management (CRM) efforts, Forman suggests. “Take, for example, companies that provide myriad products and services, such as Amazon and Facebook Marketplace,” he adds. “They know you as one customer who has unique needs, and they customize offers for those needs.” More:

‘By analyzing patterns in participant feedback and complaints, AI algorithms can pinpoint areas where programs are falling short and suggest targeted interventions to improve outcomes.’

3. Successful use of AI in government requires more than just a focus on compliance with controls.

Achieving strategic value requires articulating a vision and motivating people to establish and sustain it, Forman suggests. This is “the only way to assure government performance gains as well as that the guidance does not get reduced to a reporting exercise,” he argues. More:

‘[I]t is essential to have clearly stated goals and shared vision across agencies to ensure that the benefits are realized without compromising individual rights and societal values.’

Read the whole thing.

Buzzworthy developments of the past few days.

• Healthcare orgs looking to test AI models have a new open-source validation toolkit to work with. And just because the code is free to download doesn’t mean it’s second-rate in action. The data specialists behind the offering are with the EHR giant Epic, no slouch in healthcare software. Available now on GitHub, the kit automates data collection and mapping in users’ EHR systems to provide fast metrics and analysis on AI models, Epic leaders tell Fierce Healthcare. The automation element “creates consistency and eliminates the need for healthcare organization data scientists to do their own data mapping—the most time-consuming aspect of validation.” Fierce coverage here, Epic’s GitHub “seismometer” here.
   
• If it’s a statistic, it relates to AI and it’s more or less current, it’s probably in AIPRM’s ‘AI Statistics 2024.’ The online report has a table of contents to navigate the gobs of information gathered on one long scrollable page, but you might be better off using your browser’s wordsearch tool for, say, healthcare and such. There’s just that much stuff. Some nice infographics too. Check it out.
   
• Lunit has finished acquiring Volpara. The deal had been in the works for some time before an announcement came May 22 stating it was finally done. Lunit is based in South Korea and specializes in AI interpretation of medical images for cancer detection. Volpara’s bailiwick is AI for breast imaging specifically. Lunit says the acquisition “marks a significant milestone in Lunit’s mission to conquer cancer through AI.” The announcement doesn’t name the terms, but last December the purchase price was said to be not far below the $200 million mark.
   
• Nvidia’s ascent continues. This week The Wall Street Journal is reporting the chipmaker’s sales have tripled over the past year, to $26 billion. The performance is not only beating analysts’ expectations but also signaling that “the AI boom is still going strong,” WSJ semiconductor watcher Asa Fitch writes. Fitch adds that Big Cloud companies—Google, Microsoft, Amazon, those types—account for almost half of Nvidia’s data-center revenue stream. That’s more than $10 billion right there.
   
• A healthcare software and services supplier has opened a patient safety organization. RLDatix of Chicago says it’s moving into the space to find or refine best practices for reducing safety risks in care settings. The company plans to use AI tools in the effort, letting algorithms help identify risk factors. The new operation, RLDatix Safety Institute, is headed by Ann Louise Puopolo, RN, a former VP of enterprise patient safety with CVS Health and the present chair of RLDatix’s advisory board. Announcement here, institute website here.
   
• They’re popping the champagne at GE HealthCare. What’s the occasion? Confirmation that they’re still leading the league in AI-equipped medical devices cleared by the FDA. Of the 850-plus devices listed by the agency in a May 13 update, some 72 are GE brainchildren—“more than any other medical technology company,” GE HealthCare points out. If that sounds at all immodest, the company’s chief science and technology officer, Taha Kass-Hout, MD, puts the quote in context. The company, he suggests, is doing what any other would do after paying so many dues for so many decades. “We are leveraging our unique capabilities,” Kass-Hout says, “honed through offering leading medical devices for more than 125 years.” Fair enough. Announcement here.
   
• Creepy is as creepy does? Microsoft’s new Windows-loaded PCs, the ones with Copilot+ built in or soon to be, have an AI feature people seem to love to hate. It’s called Recall. Its intent is to “remember” everything that ever comes up on your screen so that nothing is ever irretrievably lost. As Inc. tech writer Kit Eaton puts it, the feature “sounds useful but has undeniably creepy overtones.” Evidently officials in the U.K. are troubled enough by the capability that they’re asking Microsoft to explain how Recall won’t compromise users’ privacy, safety and security. Read more about it from many others.
   
• Regardless of Recall, Copilot+ has brought back the Mac vs. PC wars of words. For the first time in decades, Microsoft consumer chief marketing officer Yusuf Mehdi tells Yahoo Finance, Microsoft believes it’s better than Apple not just on pure performance but also on the capacity to do “completely unique things that you can’t do anywhere else.” Article here.
   
• And then there’s Scarlett Johansson of Hollywood vs. Sam Altman of OpenAI. It’s probably no accident that an iteration of OpenAI’s new bot called “Sky” sounds suspiciously like Johansson. Or, more specifically, like Johansson’s voice as given to the emotionally alluring AI assistant Samantha in the movie Her. Johansson is threatening to sue. Altman is saying Sky’s voice is from another actress. The labor union SAG-AFTRA is using the dispute to lobby for more protection against AI for artists. What a fine but fun techno-mess.
   
• Funding news of note:
   
  • KKR invests $150 million in AI-powered healthcare revenue solutions provider Infinx (from MoneyControl.com)
     
  • Health tech company Tempus AI files for $100M IPO (Fierce Healthcare)
     
  • Canadian healthcare ventures nab $26M for AI innovations (BIV)
     
  • Amino Health secures $10 million in funding round led by Transformation Capital to advance product investments in AI care journeys
     
  • Videra Health secures $5.6M in funding to advance mental health AI solutions led by Peterson Ventures
     
• From AIin.Healthcare’s news partners:
   
  • Health Imaging: ‘Too many cooks in the kitchen’: How AI is both helpful and a hindrance to radiologists
     
  • Health Imaging: Machine learning helps experts identify ADHD biomarkers on MRI scans