A recent study published in Science tested OpenAI's o1-preview model against human doctors using data from 76 real emergency room visits. The AI provided an exact or very close diagnosis 82 percent of the time, slightly outperforming human physicians who scored 79 and 70 percent. Despite these results, experts warn that no agreed-upon standard exists for scoring such outcomes, and the technology should serve as a tool for second opinions rather than replacing doctors.
Adoption of AI tools in healthcare is already widespread, with about 65 percent of US physicians using OpenEvidence daily. In April alone, the service supported nearly 27 million clinical encounters and 90 million queries since 2024. While doctors praise its speed and ability to draft discharge notes, concerns remain about potential errors and the erosion of critical thinking skills.
Education sectors are reacting with stricter regulations. Orange County Public Schools plans to vote in June on a policy banning OpenAI programs for students and staff, requiring all tools to be vetted and prohibiting student data entry. Similarly, New York City public schools are finalizing rules to balance AI benefits with student data safety and human oversight.
Outside of healthcare and education, the energy demands of AI growth are drawing attention. Former Vice President Al Gore warns that massive data centers strain global energy supplies, noting a recent drop in corporate demand for renewables. Meanwhile, Microsoft released mimalloc, a memory management library designed to speed up large applications by minimizing memory allocation time.
Business and consumer adoption shows mixed results. A report indicates 99 percent of business development representatives now use AI tools, yet using them to write emails does not guarantee sales success; analyzing conversations yields better outcomes. In sales, feeling supported by one's job remains the strongest predictor of performance, outweighing technology alone.
Personal experimentation with AI is also a growing trend. Journalist Joanna Stern tested over 100 AI products in her new book, I AM NOT A ROBOT, evaluating everything from health diagnostics to digital companions against strict benchmarks. Additionally, researchers found that adding small delays to AI image generation can reduce unnecessary use and encourage more responsible behavior.
Global strategies are shifting as well. China is moving away from a catch-up mentality, focusing instead on open-source models and massive adoption driven by government support and domestic tech growth. In the mental health sector, more than half of Americans have used AI chatbots, with one in three teenagers using them daily, though experts caution against over-reliance without human integration.
Key Takeaways
["OpenAI's o1-preview model achieved an 82 percent diagnosis accuracy rate in a study of 76 emergency room visits, outperforming human doctors who scored 79 and 70 percent.", 'Approximately 65 percent of US physicians use the OpenEvidence AI tool, which handled nearly 27 million clinical encounters in April alone.', 'Orange County Public Schools plans to ban OpenAI programs for students and staff starting in June, requiring strict vetting of all AI tools.', 'Former Vice President Al Gore warns that AI data centers are straining global energy supplies, with corporate demand for renewables falling for the first time in nearly a decade.', 'Microsoft released mimalloc, a memory management library that uses thread-local heaps to improve application performance and reduce memory waste.', '99 percent of business development representatives now use AI tools, but using them for email writing does not guarantee better sales results compared to analyzing conversations.', 'Journalist Joanna Stern tested over 100 AI products in her book I AM NOT A ROBOT, benchmarking them against human performance and analyzing their costs.', 'Researchers found that adding small delays to AI image generation reduces unnecessary use and promotes more ecologically responsible behavior.', 'China is shifting its AI strategy to emphasize open-source models and massive adoption, moving away from a catch-up approach in the global race.', 'More than half of Americans have used AI chatbots for mental health support, with one in three teenagers using them daily, though experts warn against over-reliance.']Researchers Debate How to Measure AI in Medical Decisions
A new study published in Science has sparked a debate about how to evaluate artificial intelligence in clinical reasoning. Researchers tested OpenAI's o1-preview model against human doctors using data from 76 real emergency room visits. The AI provided an exact or very close diagnosis 82 percent of the time, slightly outperforming the human physicians who scored 79 and 70 percent. However, experts warn that there is no agreed-upon standard for scoring these results, and some studies using different methods have reached opposite conclusions. Coauthors of the study emphasized that AI should not replace doctors but could serve as a helpful tool for second opinions.
Most US Doctors Quietly Use OpenEvidence AI Tool Daily
About 65 percent of physicians in the United States are using an AI tool called OpenEvidence to help with clinical decisions and medical knowledge. The service was used in nearly 27 million clinical encounters in April alone, with 90 million queries submitted since 2024. Doctors praise the tool for its quick answers and ability to write patient discharge notes, though some experts worry about potential errors or a loss of critical thinking skills. The software is freely available and works well on mobile phones, making it easy for doctors in rural areas to access expert information quickly.
Orange County Schools Plan to Ban OpenAI Tools in June
Orange County Public Schools plans to vote on a new AI policy in June that will ban OpenAI programs for students and staff. The proposed rules require all AI tools to be vetted by the district and strictly prohibit student data from being entered into any AI program. Teachers will be required to use AI detectors to check student work, and students will not be allowed to use chatbots for emotional support. While some parents and students feel the rules are unfair, school leaders insist that keeping humans in charge is essential for safety and learning.
Al Gore Warns AI Data Centers Strain Global Energy Supply
Former Vice President Al Gore says the rapid growth of artificial intelligence is creating a major challenge for the global energy supply. He points to massive data centers like Project Marvel in Alabama as examples of the power and water demands caused by the AI boom. While some tech companies are buying clean energy, Gore notes that corporate demand for renewables recently fell for the first time in nearly a decade. He suggests that the best solution is to encourage tech developers to build renewable energy sources directly at data center sites.
Microsoft Releases mimalloc to Speed Up Large Apps
Microsoft has released a new memory management library called mimalloc to help applications handle large amounts of data more efficiently. The tool uses a unique design with thread-local heaps to minimize the time needed to allocate memory, which is crucial for apps running many threads. It was originally created for Microsoft's Lean and Koka programming languages but has since been used to improve response times in services like Bing. The library balances scalability and memory sharing to ensure fast performance without wasting resources.
NYC Public Schools Finalizing New Policy on AI Use
New York City public schools are finalizing a policy to regulate the use of artificial intelligence in classrooms. The district is addressing parental concerns about a technology that grows exponentially every day. Officials aim to balance the benefits of AI with the need to ensure student data remains safe and that human oversight is maintained. The final rules will likely determine which AI tools are allowed and how students and teachers can use them for schoolwork.
AI Adoption Soars in Sales But Does Not Guarantee Success
A new report shows that 99 percent of business development representatives now use AI tools, a jump from 53 percent in 2024. Despite this high adoption rate, using AI to write emails does not necessarily lead to better sales results. The data indicates that AI is most effective when used for analyzing conversations and training, rather than just generating content. Additionally, the report finds that feeling supported by their job is the strongest predictor of sales performance, outweighing the impact of technology alone.
Journalist Tests 100 AI Products in Her New Book
Joanna Stern, a former Wall Street Journal technology journalist, released a new book called I AM NOT A ROBOT about her year-long experiment with AI. She tested more than 100 AI products ranging from health diagnostics to digital companions and autonomous vehicles. Her testing followed three strict rules: ruthless testing, benchmarking against human performance, and analyzing costs. The book covers her experiences with AI doctors, therapists, and even digital lovers, offering a comprehensive look at the current state of generative AI.
Oregon Study Shows Pausing Can Reduce Unnecessary AI Use
Researchers at Oregon State University found that adding small delays to AI image generation can reduce unnecessary use. The study tested two methods: requiring users to search for existing images first and showing messages about environmental impact. The action-based method, which forced users to slow down, made them more ecologically responsible. The researchers note that high-performance computing could account for one-fifth of the world's energy consumption by 2030, making these changes important.
China Shifts Focus to Open Source Models in AI Race
China is no longer seen as playing catch-up in the global artificial intelligence race. The country has shifted its strategy to emphasize open-source models and massive adoption of AI technology. This change is driven by government support for research, a growing domestic tech industry, and increased access to data and computing resources. As a result, China is now a major player in the global AI landscape with applications in healthcare, finance, and transportation.
AI Chatbots Offer Mental Health Support With Mixed Results
More than half of Americans have used an AI chatbot, and one in three teenagers uses one daily for mental health support. These tools offer benefits like 24/7 availability, anonymity, and access for those who cannot afford traditional therapy. However, they also have downsides, including a lack of human empathy and the risk of users becoming too dependent on technology. Experts suggest that the best approach is to integrate AI chatbots with human therapy and to regulate their use to ensure safety.
Sources
- New Study Fuels Debate Over AI Clinical Decision Support
- OpenEvidence: Most physicians quietly use this medical AI tool
- How should Orange County’s schools allow AI usage? Board discussing updated policy
- 20 Years After ‘An Inconvenient Truth,’ Al Gore Grapples With the (Big) Wrinkle of Artificial Intelligence
- mimalloc: Microsoft's Speed Boost for Apps
- NYC Public Schools finalizing policy on AI
- 5 Findings From 6Sense's BDR Report That Every Sales Leader Should See
- An AI Immersion for 365 Days
- OSU study finds 'design friction' prompts can curb unnecessary AI image generation
- China’s A.I. race no longer looks like second place
- AI Chatbots For Mental Health – What Works, What Harms, and What’s Next
Comments
Please log in to post a comment.