AI adoption is advancing across business and public services, but the articles highlight limits and trade-offs. Australian universities are testing AI to mark student work, yet research by the University of Melbourne and Cardiff University found ChatGPT could not reliably assign or explain marks and often graded more generously than people. Universities are more receptive to AI-assisted feedback, but the article says humans must retain grading responsibility and warns that automated marking could weaken student-teacher relationships.
Companies are also rethinking how they organize product work. AI can help product managers build prototypes and help builders research users and assess feedback, but the article argues that combining the roles risks losing the different technical and product perspectives needed for sound decisions. Verint, by contrast, has deliberately put product and technology under one executive: Abhi Sharma, formerly CTO of Weave, takes the role after senior positions at Salesforce, Oracle, Amazon, Microsoft and Twilio. CEO Dave Rhodes says the move is intended to speed innovation and keep products aligned with customer needs; Verint says it serves more than 80 Fortune 100 companies.
In India, manufacturers are targeting the equipment needs of AI data centers, from optical transceivers and advanced packaging to cooling systems. Dixon Technologies is pursuing silicon photonics and optical transceivers through a joint venture with Taiwan’s Gemtek Technology, expected to start operations next quarter. FläktGroup has opened a Pune cooling-equipment factory targeting annual production of 6,500 units, while LG plans specialized HVAC production at its upcoming Sri City plant, expected to begin operating in December. Blue Star expects more data-center orders, and PG Electroplast is assessing the market. The government estimates investment in India’s data-center industry could reach $200 billion.
Research into AI models of brain activity is drawing both interest and skepticism. Teams including Meta, Stanford University and Neuralink are developing models trained on large volumes of brain data. Some researchers see accurate predictions as a way to test hypotheses, while others caution that matching recorded brain activity does not show that a model works like a brain or explains it. A mouse visual-cortex study published in Nature found a model could predict responses to new stimuli. Meta’s existing model uses more than 1,000 hours of brain scans, and the company announced a $15 million effort to build another using 15,000 hours of precision-scanning fMRI data.
Other developments show AI’s practical reach and its societal risks. Meta donated Ray-Ban Meta AI glasses to visually impaired U.S. veterans, with training for recipients at the University of Texas on Oct. 27, 2026; the initiative aims to provide 130,000 pairs. The glasses can help users recognize people, call loved ones and organize belongings. In Lafayette, the city council rejected a proposed downtown bar staffed by AI bartenders after concerns about downtown safety. At an OSCE and Republic of Korea conference, participants discussed AI’s potential to improve cyber threat detection and incident response alongside risks such as AI-enabled cybercrime, and emphasized resilience, digital trust, education and responsible governance.
Two articles raise broader questions about how people and organizations use AI. One argues that human working memory—estimated at roughly seven to nine chunks of information held briefly—can constrain AI because people provide information systems rely on when making decisions. Bilt, meanwhile, launched Neighborhood OS, an AI-powered platform combining property operations, resident services, rewards and nearby businesses. Its features include personalized recommendations, task automation and real-time data analysis intended to help property managers improve resident experiences and identify revenue opportunities.
Key Takeaways
- University of Melbourne and Cardiff University research found ChatGPT could not reliably grade or explain marks and often awarded higher grades than human assessors; universities still require human oversight.
- An article cautions that merging product management and product-building roles can sacrifice the distinct perspectives needed for good decisions.
- Verint appointed former Weave CTO Abhi Sharma as Chief Product and Technology Officer; his previous senior roles include Salesforce, Oracle, Amazon, Microsoft and Twilio.
- India’s government estimates data-center investment could reach $200 billion as manufacturers pursue optical transceivers, advanced packaging and cooling equipment.
- Dixon Technologies’ optical-transceiver joint venture with Taiwan’s Gemtek Technology is expected to begin operations next quarter; FläktGroup’s Pune factory targets 6,500 cooling units annually.
- Meta’s brain model uses more than 1,000 hours of scans; the company announced a $15 million project to build another model using 15,000 hours of precision-scanning fMRI data.
- Meta’s initiative aims to donate 130,000 Ray-Ban Meta AI glasses to blind and low-vision U.S. veterans; recipients received devices and training at UT on Oct. 27, 2026.
- Lafayette’s council rejected a proposed downtown bar staffed by AI bartenders after concerns about downtown safety.
- OSCE and Republic of Korea conference participants discussed AI’s uses in cyber defense, risks including AI-enabled crime, and the importance of responsible governance and digital trust.
- Bilt launched Neighborhood OS, combining property operations, resident services, rewards and nearby businesses with AI recommendations, task automation and real-time analysis.
Lafayette Council Rejects Downtown Bar With AI Bartenders
The Lafayette City Council voted against a proposed high-tech bar in downtown Lafayette after meeting late Tuesday. The decision followed concerns about safety in the downtown area. The owner planned to staff the club with artificial intelligence bartenders.
Lafayette Council Rejects Downtown Bar With AI Bartenders
The Lafayette City Council voted against a proposed high-tech bar in downtown Lafayette after meeting late Tuesday. The decision followed concerns about safety in the downtown area. The owner planned to staff the club with artificial intelligence bartenders.
AI Marking Still Needs Human Oversight at Australian Universities
Australian universities are exploring AI tools to mark student work, but the article says humans remain responsible for grading decisions. Research by the University of Melbourne and Cardiff University found that ChatGPT could not reliably assign or explain marks, especially for individual grading criteria, and often gave higher marks than human graders. Universities appear more open to using AI to help write feedback, though human oversight is still required. The article also warns that AI marking may harm student-teacher relationships and suggests universities consider assessment approaches that bring students and teachers closer together.
Combining AI Product Roles Could Be a Costly Mistake
Companies may use AI tools to combine product management and product building into fewer roles. AI can help product managers create prototypes without coding and help builders research and assess feedback. But builders and managers bring different perspectives: builders focus on technical issues such as architecture, implementation, dependencies, quality, and tradeoffs. The article argues that good decisions come from the meeting point between these perspectives, and combining the roles could be a mistake.
Visually Impaired Veterans Receive Meta AI Glasses at UT
Several visually impaired veterans received free Ray-Ban Meta AI glasses and hands-on training at the Dell Campus for Advanced Research on Oct. 27, 2026. The event was part of Meta’s initiative to donate 130,000 pairs of glasses to blind and low-vision U.S. veterans. Partners included East Texas Lighthouse for the Blind, the Texas Technology Access Program at UT, the Blinded Veterans Association, and Austin Lighthouse for the Blind. The glasses can help users recognize people, call loved ones, and organize belongings. David Manning, who has macular degeneration, received a pair, and his daughter hopes they will help him stay independent and connected with family.
Verint appoints Abhi Sharma to lead product and technology
Verint appointed Abhi Sharma as its Chief Product and Technology Officer, combining product and technology under one leader. He will guide product strategy and technology development as the company focuses on artificial intelligence and customer experience software. Sharma has held senior roles at Weave, Twilio, Salesforce, Oracle, Amazon and Microsoft, and most recently served as Weave’s Chief Technology Officer. Verint CEO Dave Rhodes said the change aims to speed up innovation and keep products aligned with customer needs. Verint says its customers include more than 80 Fortune 100 companies.
OSCE and Asian partners discuss AI risks and cyber security
The Republic of Korea’s Ministry of Foreign Affairs and the OSCE Secretariat organized a conference on AI and cyber and information security. Participants discussed using AI to improve threat detection and incident response, as well as risks from AI-enabled cybercrime, crime and terrorism. Speakers also stressed cyber resilience, digital trust, education and responsible AI governance. The event brought together OSCE participating States and Asian Partners for Co-operation to share experience and consider practical next steps. It continued a conference series started by the Republic of Korea in 2017.
Indian electronics makers target AI data center equipment
Electronics manufacturers in India are exploring products for AI data centers, including optical transceivers, advanced packaging and specialized cooling systems. Dixon Technologies is pursuing silicon-photonics and optical transceivers, including a joint venture with Taiwan’s Gemtek Technology that is expected to begin operations next quarter. Samsung group company FläktGroup opened a factory in Pune to make cooling equipment, with production expected to reach 6,500 units a year. LG plans to make specialized HVAC solutions at its upcoming Sri City plant, expected to start operating in December. Blue Star expects more data center orders, while PG Electroplast is evaluating entry into the market; the government estimates about $200 billion could be invested in India’s data center industry.
Researchers debate whether AI brain models can explain the brain
Researchers disagree about whether AI models that predict brain activity can also explain how the brain works. Teams including Meta, Stanford University and Neuralink are developing foundation models trained on large amounts of brain data. Some scientists say accurate models could help researchers test predictions and study brain activity, while others warn that matching recorded activity does not prove a model works like a brain or explains it. Research has shown that models can predict responses to new stimuli, including in a mouse visual cortex study published in Nature. Meta’s model uses more than 1,000 hours of brain scans, and the company announced a $15 million effort to build another model using 15,000 hours of precision-scanning fMRI data.
Article argues limited working memory may constrain AI
The article argues that human working memory may be a major constraint on AI because people supply information that AI systems use to make decisions. It says working memory supports learning, reasoning and decision-making but can hold only a limited amount of information. The article gives an estimate of around 7 to 9 chunks at a time, held for only a short period. It cites research by G. A. Miller, M. G. Cowan and A. D. Baddeley. The same points about working memory and AI appear several times in the article.
Bilt launches AI platform for neighborhood properties
Bilt unveiled Bilt Neighborhood OS, an artificial intelligence operating system and commerce network for neighborhoods. It brings property operations, resident services, rewards, and nearby businesses together in one platform. Property managers can use it to streamline tasks, improve residents’ experiences, and find new revenue opportunities through targeted marketing and commerce. Its AI features offer personalized recommendations, automate tasks, and analyze real-time data.
Sources
- Lafayette City Council rejects proposed new downtown AI bar
- Lafayette City Council rejects proposed new downtown AI bar
- AI marking means uni assessments could theoretically never be read by a human. What now?
- Companies Want Leaner AI-Powered Teams. They Could Be Making an Expensive Mistake
- Visually impaired veterans receive Meta AI glasses at UT Dell event
- Verint names Abhi Sharma as chief product & tech chief
- OSCE, Asian Partners Tackle AI Security Challenges
- Electronics Inc runs ops update to build AI infra
- Future of AI brain models sparks debate
- Why Human Working Memory Might Be AI’s Biggest Constraint
- Bilt Unveils AI-Powered Neighborhood OS
Comments
Please log in to post a comment.