Billionaire tech mogul Elon Musk took to his X social media platform this past week to ask users of his Grok artificial intelligence chatbot to upload their medical images and generate a diagnosis.
“Try sending X-rays, PET images, MRI images or other medical images to Grok for analysis,” says Musk suggested on October 29 the X platform, where subscribers get access to the AI.
Uploading medical data via a social media platform raises questions about privacy.
Some in healthcare have already tested the large language model — which is controversial for collecting social media data as part of the training — and are giving mixed reviews about its accuracy.
Meanwhile, Grok had already called on European privacy regulators to question possible violations of the General Data Protection Regulation.
Mixed reviews on accuracy
“This is still in the early stages, but it is already quite accurate and will be extremely good,” Musk said in an interview after on X.
Dr. Derya Unutmaz, who researches the mechanisms of human T cells at The Jackson Laboratory, a nonprofit scientific research institute in Connecticut, was one of many who tested Grok with an X-ray last week and noted that he could not answer the question correctly. to get:
“When I used this prompt, Grok got it right this time: ‘You are an emergency physician, this patient arrived at the emergency room with shortness of breath. No apparent trauma. What is your diagnosis?’ for the next X-ray,” he said answer to Musk.
“Apparently it needed some context,” he added. “However, the response has been very thorough and impressive!”
Dr. Laura Heacock, a breast radiologist, deep learning researcher and associate professor at NYU Imaging, said she was less than impressed. answer.
Heacock said she used the same breast mammograms, ultrasounds and MRIs she previously used to benchmark GPT4, and in a series of 12 social media posts, she detailed Grok’s answers and her thoughts on those answers.
“So how did Grok do in chest radiology? A little better than GPT4v, but none of the diagnoses were correct,” she said, noting that she expected much better performance given Musk’s statement about accuracy.
Grok was able to identify an image as a mammogram or an ultrasound, but responded to Heacock’s brief questions with incorrect body parts and incorrect possible findings. On a mammogram of known malignancy, Grok suggested that what Heacock called an “obvious large left cleft cancer” may have been calcifications.
“There aren’t any,” she said.
Heacock said she would try Grok again in a few months, but “for now, non-generative AI methods continue to outperform in the field of medical imaging.”
Social data in model training
xAI began laying the foundation for its “anti-woke” Grok-1 AI model in November 2023.
“Grok is designed to answer questions with a bit of humor and has a rebellious streak, so please don’t use it if you hate humor,” xAI said in the announcement last year.
“A unique and fundamental advantage of Grok is that it has real-time knowledge of the world through the X platform. It will also answer tough questions rejected by most other AI systems,” the company added.
On March 11, Musk announced on his social media platform that xAI would make the Grok algorithm, a direct competitor to ChatGPT, accessible as open-source software.
The company then said in May that it had raised $6 billion following the release of the Grok-1.5 model, which it said offered improved reasoning and problem-solving capabilities, and the release of Grok-1.5V, which added the ability to perform visual process information – including documents, diagrams, graphs, screenshots and photos.
Dr. Recognizing the potential to disrupt healthcare, Sai Balasubramanian, who was named director of the Public Health Informatics Data and Exchange Unit at the Texas Department of State Health Services earlier this year, opined for Forbes that Grok would become a foundational model in everyday healthcare. can be embedded. workflows, such as in radiology practices.
Like other AI models, Grok could be used to increase task automation by improving clinical workflows and productivity, but would likely face competition from industry veterans in the enterprise space, Balasubramanian said.
However, the director of public health informatics noted in the article that there is a significant opportunity for Musk to bring together xAI and his medical device company Neuralink.
In 2020, Musk demonstrated his Neuralink brain computer implant in pigs. In September, the neurological tech startup announced that it had cleared the U.S. Food and Drug Administration groundbreaking device designation for its Blindsight implant, which aims to restore vision.
The CEO of Tesla and SpaceX was once on a crusade to stop what he called “the AI apocalypse” and exchanged words about its potentially destructive power with Facebook CEO Mark Zuckerberg.
The latter criticized Musk’s views on the demise of AI and was optimistic about AI’s potential for improved diagnostic capabilities and innovation in drug discovery. Musk, who had previously said AI has the potential to reduce humanity to “a pet or a house cat” in a future governed by it, responded that Zuckerberg’s “understanding of the subject is limited.”
Stand up against privacy rules
Currently, X premium and premium+ users have access to two new models: Grok-2, which integrates real-time information from the X social media platform, and Grok-2 mini.
With the beta version of Grok-2 in August, users were given the ability to create images – many of which were highly inappropriate due to a lack of filters, as researchers were able to show this summer – from text message prompts and publish them on the social platform. And by connecting real-time data from X, Grok could respond to clues about current events as they unfold.
Government observers responded, citing a looseness and general disregard for secure AI and social media standards that permeate Grok. In addition to claims from the European Union over possible violations of the Digital Safety Act, nine states sent letters alleging disinformation about the 2024 election created using the platform.
“X has a responsibility to ensure that all voters who use your platform have access to guidance that reflects truthful and accurate information about their constitutional right to vote,” Minnesota Secretary of State Steve Simon said in an Aug. 5 message . letter.
“We urge X to immediately implement a policy directing Grok users to CanIVote.org when asked about US elections”
Data Collection and When HIPAA Applies
At least one privacy expert has expressed concern that HIPAA covered entities participating in efforts to improve Grok’s accuracy in healthcare data through social media could jeopardize patient privacy in the US
And European government leaders previously questioned developer xAI over suspected GDPR violations. On August 6, the Irish Data Protection Commission asked the High Court of Ireland to shut down the social media network collect account holder data for training de Grok LLM.
However, when users ask Grok something, something won’t automatically be posted to the Healthcare IT news.
“At this point you can turn it on or off – shows you messages, as well as your interactions, inputs and results with Grok, so you can use them for training and tuning,” Kim said on Monday, sharing a screenshot. “You can also choose to delete your conversation history.”
While the company called the DPC’s attempt to ban Grok “unwarranted, overbroad and excludes X without any justification” on its social media platform, xAI vowed to “continue to protect people’s privacy.”
“Unlike the rest of the AI industry, we have chosen to provide all the company in August.
While users can make a number of choices based on xAI’s privacy policy“They need to be aware of how their data is used and processed,” says Kim. HIPAA may apply to data shared on social media platforms if a covered entity discloses protected patient information, she explained.
“It all depends on who makes the information public,” Kim said. “If it is a covered entity, business associate, or organized health care arrangement, HIPAA applies.”
Andrea Fox is editor-in-chief of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.