AI In Education
Microsoft AI in Education Report: A 2025 Microsoft report shows education has the highest rate of AI adoption across all industries surveyed, with 86% reporting its use.
- High Usage, Low Confidence: While most education leaders (99%), students (93%), and educators (87%) have used AI for school-related purposes, fewer than half of students (41%) and educators (44%) feel they “know a lot about it.”
- Student Applications and Concerns: Students frequently leverage AI as a conversation partner for brainstorming (37%), summarizing information (33%), getting quick answers (33%), and receiving initial feedback (32%). While general concerns about plagiarism and cheating have slightly decreased since last year, they remain a top worry for educators. Students are primarily concerned about being accused of plagiarism or cheating (33%) and the potential for becoming overly dependent on AI tools (30%).
- Training Gap: Despite high usage, there’s a clear deficit in formal AI training. 76% of leaders report that at least half of their institution’s AI users have received training, while 45% of educators and 52% of US students state they have received no training. Per Microsoft’s report: “this gap points to both a need for more training and a need for leaders to focus on the gap between what they feel they have provided and what educators and students feel they have received.”
EDUCAUSE Teaching with AI Program: EDUCAUSE is offering a 2-week online program for Teaching with AI. The program explores AI’s implications in higher education, examining academic integrity concerns, evaluating course redesign examples, experimenting with AI tools, and facilitating live peer discussions.
AI In Research
Medical Diagnosis Breakthrough: Microsoft has developed the Microsoft AI Diagnostic Orchestrator (MAI-DxO), which correctly diagnosed up to 85.5% of complex New England Journal of Medicine cases, compared to a group of 21 experienced physicians with 20% accuracy. The orchestrator also demonstrated cost-effectiveness with lower overall testing costs; however, the technology is not yet approved for clinical use.
Antibody Design Advance: Chai Discovery introduced a new series of models achieving double-digit hit rates in de novo antibody design, representing a major breakthrough in molecular design and drug discovery.
Google’s DNA Analysis Tool AlphaGenome: Google DeepMind developed AlphaGenome, an AI tool that can predict how single variants in human DNA sequences impact gene regulation. AlphaGenome achieves state-of-the-art performance, outperforming the vast majority of external models.
Independent AI Research Initiative: Andy Konwinski, co-founder of Databricks and Perplexity, has launched a nonprofit with a $100M pledge to foster independent AI research. The nonprofit positions itself as an alternative to increasingly commercial AI research labs, with its first major grant set to build a new AI Systems Lab at UC Berkeley.
AI Current Events
Senate Removes Proposed AI Regulation Ban: The US Senate has removed the AI regulation provision from the One Big, Beautiful Bill, which would have banned state and local AI regulation for 10 years. Proponents of the provision argued that a patchwork of state laws would hinder progress, while opponents contended the ban would weaken state protections and allow a lack of accountability in the pursuit of innovation.
Consumer Usage Report: A consumer report by Menlo Ventures shows that over 61% of American adults have used AI in the past six months, with one in five relying on it daily. The top uses are in writing (51%), coding (47%), and academic assignments (43%). General AI chatbots dominate with 91% of users, with specialized tools only gaining traction for very specific tasks.
Copyright Court Victory for AI: AI companies continue to rack up wins in copyright court as a federal judge sided with Meta in a lawsuit claiming copyright infringement on copyrighted books, ruling the company’s use as “fair use.” This follows a similar ruling last week with Anthropic, although multiple lawsuits are still active in various courts.
Weekly AI Tip
Are you still struggling to understand what makes something an AI “Agent”? If so, you are not the only one! 2025 was predicted to be “The Year of the AI Agent,” and so far it seems to be living up to that potential. But “Agent” would also win the award for “most poorly described word” as well.
Custom GPTs and Gemini Gems are often called agents because you can embed instructions and even grant them built-in tools (e.g., image generation, Deep Research); however, they remain largely user-driven, responding only when prompted.
In contrast, AI agents plan, use external tools, remember intermediate results, and act autonomously, sometimes coordinating as specialist teammates, so they can pursue an overarching goal with minimal human steering. True agents operate in loops or cycles: they act, observe results, replan, and act again, engaging in iterative, self-directed problem-solving rather than simply responding to prompts.
Resources: Tina Huang has created an “AI Agents Fundamentals in 21 Minutes” video that dives into AI Agents in more depth and serves as a great introduction to agentic workflows. Interested in going into more depth and building some agents yourself? Check out her video “Building AI Agents in 44 Minutes.”
Upcoming AI Events
Building Custom GPTs & Gems Workshop – August 8, 10:00AM-2:00PM in DMC 101
- The Building Custom GPTs and Google Gems Workshop focuses on creating personalized and functional AI tools using ChatGPT and Google Gemini. Participants will learn to develop AI tutors, AI generators for assessment content, and AI writing/editing assistants. The workshop is designed so each attendee leaves with their own functional custom AI and the foundational knowledge to build additional tailored solutions. RSVP
