Experts at Johns Hopkins University recently debated whether generative AI helps or hurts college learning, with more than 100 students and community members in attendance. Gary Marcus and Cathy O'Neil argued that AI leads to cognitive surrender, while Ravi Pendse and Edward Maloney maintained that AI can support learning when used responsibly. Attendees voted before and after the debate, and the share favoring the enhance position rose from 26 percent to just under 40 percent.
A separate report from the University of Minnesota warns that casual student use of generative AI can disrupt memory, critical thinking, collaboration and social-emotional development. Lead author Panayiota Kendeou says primary school students face the greatest risk, and the report recommends introducing AI tools only after students have built strong foundational skills. Researchers stress that education should keep students and teachers at the center when adopting new technology.
Meanwhile, Eric Schmidt says China trails the United States by about six months in AI development. He had expected chip restrictions to set China back roughly 10 years, but noted that China has found ways to build powerful AI systems with far less energy. Schmidt described the U.S.-China AI competition as the defining race of the current era.
On the enterprise side, OpenAI and ServiceNow expanded their partnership to bring more AI models and agents into workplace workflows. New integrations connect ChatGPT and Codex with ServiceNow products for IT, security and application development. ServiceNow plans to give all employees access to ChatGPT Work and Codex, with enhanced capabilities expected by the end of 2026.
Cisco engineers rebuilt its PubHub platform using a custom AI-first development process. The team began by having AI document more than 15 legacy repositories, then used AI to create requirements, UI prototypes and full features with Claude Code. As the volume of generated code grew, human review became the main bottleneck in the pipeline.
In the United Kingdom, Geoffrey Vos, lord chief justice of England and Wales, said AI could help decide certain civil cases. He sees potential for AI to assist with document review, evidence analysis and decision-making in complex financial matters, while stopping short of replacing human judges. The UK government is already considering broader use of AI within the justice system.
A new study found that AI tools used for rare disease diagnosis suffer from serious data leakage problems. When strict data isolation was applied, recall dropped from 54.3 percent to 40.3 percent, and accuracy for supervised classifiers fell from 42.9 percent to 16.7 percent across external groups. Researchers say current evaluation methods measure leakage more than real medical reasoning and call for stricter assessment frameworks.
AI safety researcher Nate Soares, who once worked at Google, says he believes there is more than a 50 percent chance AI could wipe out humanity within ten years. He stopped contributing to his retirement savings over a decade ago and says he is certain about the danger despite working to prevent it. His views reflect deep concern within parts of the AI research community about long-term risks.
William and Mary held its annual &hacks competition with the theme AI and Imagination, drawing students from many disciplines who created 34 projects over the weekend. Organizers aimed to show that AI is more than a content generator. One standout project, Etymogle, combined linguistics and computer science into a word guessing game.
A separate analysis warns that universities relying on AI for administrative tasks may end up increasing burdens on faculty rather than reducing them. Instead of simplifying workflows, AI could add new systems, forms and compliance requirements. The author notes that existing setups in the U.K. already force teachers to juggle multiple disconnected platforms, and that schools risk using AI to preserve inefficient processes rather than fix them.
Key Takeaways
- Audience support for AI in education rose from 26 percent to just under 40 percent after a Johns Hopkins debate
- University of Minnesota report warns casual AI use threatens student memory, critical thinking and social-emotional development
- Eric Schmidt says China trails the US by only six months in AI development despite chip restrictions
- OpenAI and ServiceNow expanded their partnership to integrate ChatGPT and Codex into enterprise workflows by end of 2026
- Cisco rebuilt PubHub using Claude Code, but human review became the main bottleneck
- UK lord chief justice Geoffrey Vos says AI could help decide certain civil cases without replacing judges
- Rare disease diagnosis AI tools saw recall drop from 54.3 percent to 40.3 percent under strict data isolation
- AI researcher Nate Soares believes there is over a 50 percent chance of AI wiping out humanity within ten years
- William and Mary students built 34 projects at &hacks, including Etymogle, a linguistics and computer science word game
- Universities risk increasing faculty burdens by using AI to preserve inefficient administrative processes
Johns Hopkins debate leaves audience divided on AI in education
Johns Hopkins University hosted a debate on whether generative AI helps or hurts college learning. More than 100 students and community members attended. Experts argued both sides, with some warning about cognitive surrender and others promoting AI literacy. The audience vote shifted slightly toward the enhance side by the end.
Higher education experts split on AI's role in learning
At Johns Hopkins University, four experts debated generative AI in higher education. Gary Marcus and Cathy O'Neil argued AI causes cognitive surrender. Ravi Pendse and Edward Maloney said AI can help when used responsibly. Attendees voted before and after, with the enhance group growing from 26 percent to just under 40 percent.
Eric Schmidt says China trails US by only six months in AI
Eric Schmidt said China is now about six months behind the United States in AI development. He noted that chip restrictions were expected to slow China for about 10 years. China found ways to develop powerful AI with much less energy. Schmidt called the US-China AI race the defining competition of the age.
Cisco team rebuilds PubHub using AI-first development process
Cisco engineers rebuilt PubHub using a custom AI-first software development lifecycle. They started by using AI to document more than 15 legacy repositories. AI later helped create requirements, UI prototypes, and full features with Claude Code. Human review became the main bottleneck as generated code output grew.
William and Mary hackathon explores AI and imagination
William and Mary held its annual &hacks competition with the theme AI and Imagination. Students from many disciplines created 34 projects over the weekend. Organizers wanted to show AI is more than a content generator. One notable project, Etymogle, combined linguistics and computer science into a word guessing game.
AI researcher Nate Soares stops saving for retirement over extinction fears
Nate Soares, a top AI safety researcher, believes there is over a 50 percent chance AI will wipe out humanity within ten years. He says he stopped contributing to his retirement savings from Google a decade ago. Soares says he works hard to prevent this outcome but is certain about the danger. His views highlight deep concern in the AI research community about future risks.
Study finds AI rare disease diagnosis tools suffer from data leakage
A new study audited large language model systems used for rare disease diagnosis and found serious data leakage problems. When strict data isolation was applied, performance dropped sharply from 54.3 percent recall to 40.3 percent. Supervised classifiers saw accuracy fall from 42.9 percent to 16.7 percent across external groups. The researchers say current evaluation methods measure data leakage more than real medical reasoning. They propose stricter evaluation frameworks to fix these flaws.
University of Minnesota warns student AI use threatens learning
A University of Minnesota report warns that casual student use of generative AI can severely disrupt learning. Experts say it threatens memory, critical thinking, collaboration and social-emotional development. Lead author Panayiota Kendeou says primary school students are especially at risk. The report recommends AI tools only after students have built foundational skills. Researchers stress that education must center students and teachers when adopting AI.
OpenAI and ServiceNow expand partnership for enterprise AI workflows
OpenAI and ServiceNow expanded their partnership to bring more AI models and agents into enterprise workflows. New integrations connect ChatGPT and Codex with ServiceNow products for IT, security and application development. ServiceNow will give all employees access to ChatGPT Work and Codex. The partnership also covers cybersecurity tools, voice-powered experiences and application building. Enhanced capabilities are expected by the end of 2026.
UK judge says AI could help decide certain court cases
Geoffrey Vos, lord chief justice of England and Wales, says AI could help decide some civil cases. He believes AI could assist with document review, evidence analysis and decision-making in complex financial cases. Vos says AI would not replace human judges but could improve efficiency and accuracy. He acknowledges concerns about bias in AI systems but says careful design can address them. The UK government is already considering broader use of AI in the justice system.
AI could make university administration worse not better
The article warns that universities using AI to handle administrative tasks may actually increase burdens on faculty. Instead of simplifying workflows, AI could add more systems, forms, and compliance requirements. Faculty may spend more time on paperwork and less time on teaching and research. The author points out that existing systems in the U.K. already force teachers to manage multiple disconnected platforms. The risk is that schools will use AI to preserve inefficient processes rather than fix them.
Sources
- Debaters Tried to Change People’s Minds About AI. Here’s What Happened.
- Cognitive Surrender vs. AI Literacy: Higher-Ed Experts Debate AI
- Eric Schmidt says China is now just 6 months behind the U.S. in AI
- Rebuilding PubHub: AI-first SDLC, one bottleneck at a time
- &hacksXII turns imagination into innovation
- Top AI Safety Researcher Says He's Stopped Saving for Retirement Because He's So Sure We're All Going to Die
- Auditing evaluation artifacts and data leakage in large language model systems for rare disease diagnosis
- University of Minnesota report warns student AI use poses 'severe' threat to learning
- OpenAI and ServiceNow expand partnership across enterprise workflows
- Should AI Decide Your Court Case? This Judge Says Yes
- Universities should not be adopting AI to automate their own absurdities
Comments
Please log in to post a comment.