Teachers are getting more exposure to AI, but educators say training still needs time and practical work. A national survey of 852 educators found about two thirds had attended at least one AI professional-development session, up 11 percentage points from last year. Teachers told Education Week that sessions are often optional or too brief. Dallas math teacher Lance Barasch says educators need to understand AI to guide responsible student use, while West Virginia science teacher Jaime Moss wants hands-on practice and time to share lessons with colleagues.
At Stanford, education professor Victor Lee urges teachers to make room for candid classroom discussions about AI. He says students may see a double standard when teachers use AI but restrict student use, damaging trust and making assignments seem pointless. Teachers should explain why some work has limits, that AI can produce errors because of poor training data or outdated information, and that the technology’s effect on future jobs remains uncertain.
In business, McKinsey partners say AI tools can review websites, company databases and confidential investment documents to generate value-creation ideas in minutes, potentially replacing the first four to five days of analyst work. David Pralong says detailed private data could support forecasts with useful predictability for about 36 months, but analysts still need to test proposals and financial plans. The partners warn that AI can be inaccurate or biased and requires oversight and fact-checking. Chase recommends starting with one AI agent in one workflow, noting that many organizations have begun AI efforts but few report major progress.
AI’s risks and uses also feature in public policy. Ahead of Victoria’s Nov. 28 state election, a guide recommends tracing where and when media first appeared, checking Content Credentials, comparing multiple detection tools, consulting trusted fact-checkers and checking the VEC misinformation register for voting claims. Missing credentials do not prove content is fake, since reposts and screenshots can strip them. At COP31, the presidency launched the Antalya Artificial Intelligence Pledge within a package addressing responsible AI, green industrialisation and climate technologies.
A separate warning urges the United States to move faster on AI security. AI models can already identify serious software weaknesses, while slow fixes and limited resources leave defenders exposed; the article warns that criminals, terrorists or hostile states could exploit insecure networks and that AI could raise biological risks. It calls for stronger defenses, advanced-model access for trusted security teams, better testing and tighter US export controls on advanced chipmaking equipment and high-end chips in China.
Workers may also lose some of AI’s apparent time savings to checking its output. Verification can take extra time and mental effort because errors may be hard to locate, and fixing them can trigger further review. The article says organizations should provide workers with time, training and authority to check AI-assisted work, and measure productivity across the entire process rather than just the first draft.
Technology’s longer history is the focus of Chelsea Follett’s book “The Grim Old Days.” The Cato Institute research fellow argues that people romanticize preindustrial life despite hardships including dark streets, poor sanitation and limited heating. Cato’s Human Progress project says economic output in the last two centuries exceeded that of all earlier centuries combined. Follett also notes that farmers now account for 2% of people in developed countries, compared with as many as 90% in many preindustrial economies, and warns that society could squander the AI revolution if it fails to learn from industrialization.
South Korea is widening its response to cheating with AI smart glasses. The Korea Exchange disqualified a candidate who used them in a June 27 written hiring test and barred the person from its tests for five years. The Korea Human Resources Corporation has found 12 smart-glasses cases in national qualification exams this year, twice the six previously known, and requested or plans to request police investigations in nine. The government will review all 528 national qualification tests held this year—491 technical and 37 professional—and may refer suspicious answer sheets to investigators.
AMD and Google Cloud are applying AI infrastructure to a different concern: data privacy. AMD says its 5th Gen EPYC processors will power Google Cloud’s TPU 8tx infrastructure, bringing hardware-enforced confidential computing to a cloud TPU system for the first time. AMD Secure Encrypted Virtualization is designed to keep customer prompts and model weights encrypted during processing for inference, fine-tuning and reinforcement learning. The companies highlight performance and privacy needs, while the article says adopters should also consider hardware costs, staff training and integration with existing systems.
Key Takeaways
- A survey of 852 educators found about two thirds had attended AI professional development, up 11 percentage points year over year; teachers want more required, hands-on training.
- Stanford’s Victor Lee says open discussion of AI limits, errors and future uncertainty can help maintain student trust.
- McKinsey partners say AI can compress the first four to five days of investment analysis into minutes, but outputs still require human testing, oversight and fact-checking.
- Before Victoria’s Nov. 28 election, voters are advised to verify media origins and use Content Credentials, multiple detection tools, trusted fact-checkers and the VEC misinformation register.
- The COP31 Presidency launched the Antalya Artificial Intelligence Pledge as part of efforts on responsible AI, green industrialisation and climate technologies.
- An AI security article calls for stronger defenses, trusted security-team access to advanced models, better testing and tighter US controls on advanced chipmaking equipment and high-end chips in China.
- Organizations should equip workers to verify AI-generated work and measure productivity across review as well as drafting.
- South Korea plans to review all 528 national qualification tests after the Korea Human Resources Corporation identified 12 smart-glasses cheating cases this year; nine cases have been or may be referred for police investigation.
- AMD 5th Gen EPYC processors will power Google Cloud TPU 8tx, using Secure Encrypted Virtualization to keep prompts and model weights encrypted during processing.
- Chelsea Follett’s “The Grim Old Days” argues that industrialization improved living conditions and warns against repeating past failures as society approaches the AI era.
Teachers say AI training needs more time and hands on practice
A national survey of 852 educators found that about two thirds had attended at least one professional development session on using AI, an increase of 11 percentage points from last year. But teachers interviewed by Education Week said training is often not required or given enough time, so it may not reach educators who are unsure about AI. Dallas math teacher Lance Barasch said teachers need to understand AI so they can help students use it responsibly. West Virginia science teacher Jaime Moss said educators want hands on practice and time to share what works, rather than one time demonstrations.
Stanford professor urges honest classroom talks about AI
Victor Lee, a professor at Stanford University's Graduate School of Education and faculty lead for the Stanford Accelerator for Learning's AI and Education initiative, discussed ethical questions about AI in schools. He said teachers should create a safe space for students to ask about AI and explain why some assignments have limits. Lee warned that students may see unfairness when teachers use AI but restrict student use, which can weaken trust and make assignments feel pointless. He also said teachers should explain that AI can make mistakes because of poor training data or outdated information, while discussing the uncertain ways technology may affect future jobs.
McKinsey says AI tools could sharpen investment reviews
McKinsey partners said AI tools can scan websites, company databases and confidential investment documents to produce ideas for creating value in minutes, replacing the first four to five days of analyst work. David Pralong said detailed private data could support forecasts with a useful level of predictability for about 36 months, while analysts still need to test ideas and financial plans. The speakers also warned that AI can produce inaccurate or biased results and needs oversight and fact checking. Chase said adoption remains a major challenge, noting that many organizations have started AI efforts but few report major progress, and advised starting with one agent in one workflow.
Five checks can help voters spot AI election misinformation
Ahead of Victoria's state election on Nov. 28, political campaigns are using AI, and AI made false or misleading content is drawing attention online. The article recommends checking when and where an image or video first appeared, and looking at Content Credentials, a digital record of who made or edited media. It also suggests using multiple AI detection tools, consulting trusted fact checking sites and checking the VEC misinformation register for claims about the voting process. The article cautions that missing Content Credentials do not prove media is fake because screenshots and social media posts can remove them.
A new book argues technology made modern life far better
Chelsea Follett, a research fellow at the Cato Institute's Center for Global Liberty and Prosperity, argues in her book “The Grim Old Days” that people often romanticize life before industrialization. She points to past hardships such as dark streets, poor sanitation and limited heating, and says modern technology has improved life for both rich and poor. According to Cato's Human Progress project, economic output in the last two centuries exceeded that of all earlier centuries combined. Follett says industrialization also reduced the need for farm labor, with farmers making up just 2% of people in developed countries today, compared with as many as 90% in many preindustrial economies. She warns that failing to learn from the Industrial Revolution could lead society to squander the AI revolution.
South Korea expands probe into smart glasses cheating
The Korea Exchange disqualified a candidate who used AI smart glasses during its June 27 written hiring test and barred the candidate from taking its tests for five years. The Korea Human Resources Corporation has found 12 smart glasses cheating cases in national qualification tests this year, twice the six cases previously known. It has requested or plans to request police investigations in nine cases, including incidents at exams in Seoul, Incheon, Busan and Ulsan. The government will review all 528 national qualification tests held this year, covering 491 technical and 37 professional qualifications. Experts will check suspicious answer sheets, and confirmed concerns may be sent to investigators as the government prepares broader anti-cheating measures by year’s end.
COP31 launches Antalya pledge on responsible AI and climate action
The COP31 Presidency launched the Antalya Artificial Intelligence Pledge as part of a package of actions. The package aims to advance responsible artificial intelligence, green industrialisation and the use of climate technologies.
US urged to act quickly against rising AI security risks
The article warns that AI models are already helping find serious weaknesses in computer code, while slow repairs and limited resources can leave defenders behind. It says criminals, terrorists or hostile states could use AI to exploit insecure networks, and that AI may also raise the risk of biological threats. The author argues that governments and companies need to strengthen defenses, give trusted security teams access to advanced models and improve AI testing. The article also calls for tighter US export controls on advanced chipmaking equipment and access to high-end chips in China. It says the coming years will test how quickly the United States can respond to AI-related risks.
Checking AI output can erase some productivity gains
AI can speed up drafting, but workers often spend extra time and mental effort checking its accuracy, relevance and bias. The article explains that verification is harder than creation because workers may not know whether an error exists or where to look, and fixing mistakes can lead to more rounds of review. Checks can also feel unrewarding, which may cause workers to become less alert over time. The author says AI is changing some workers’ skills from creating information to supervising AI output. Organizations should give workers enough time, training and authority to verify AI-assisted work, and measure productivity across the full process rather than just the first draft.
AMD EPYC chips bring confidential computing to Google Cloud TPUs
AMD says its 5th Gen EPYC processors will power Google Cloud’s TPU 8tx infrastructure, bringing hardware-enforced confidential computing to a cloud Tensor Processing Unit system for the first time. AMD Secure Encrypted Virtualization helps keep customer prompts and model weights encrypted while workloads are processed. The setup is intended to support AI tasks such as inference, fine-tuning and reinforcement learning while protecting sensitive data. AMD’s Dan McNamara and Google Cloud’s Mark Lohmeyer highlighted performance and privacy needs for AI systems. The article says businesses considering the technology should also weigh hardware costs, staff training and integration with existing systems.
Sources
- Teachers Pinpoint What’s Missing in Professional Development About AI
- Should Teachers Use AI Detectors? Learning About AI’s Ethical Traps
- McKinsey bets AI diligence can see 36 months out
- Five ways to spot AI misinformation before you vote
- Rising anti-AI movement forgets all the wonders, life-changing improvements technology has brought us
- Following the controversy over cheating using artificial intelligence (AI) by senior tax accountants..
- Cop31 launches Antalya Artificial Intelligence Pledge
- How the US Can Survive the AI Danger Zone
- Catching AI's mistakes is a mental burden—and a drag on productivity
- AMD EPYC Processors Enhance Google Cloud's AI with New Security Features
Comments
Please log in to post a comment.