AI Daily Briefing — 2026-07-06

While much of the public excitement around artificial intelligence centers on chatbots and image generators, some of the most impactful applications are taking place in heavy industry. The energy sector, with its vast infrastructure and continuous streams of operational data, is emerging as a key proving ground. Companies like Woodside Energy are embedding AI into the core of their operations, moving beyond experimental tools to systems that enhance safety, efficiency, and decision-making in high-stakes environments.

Woodside’s journey with AI began years before the current generative AI boom. The company has long used predictive analytics and machine learning for exploration, drilling, and maintenance. According to Andrew Melouney, Woodside’s vice president for digital, the abundance of operational data from equipment and plants created clear, high-value use cases. This foundation in data governance and infrastructure now allows the company to adopt more advanced agentic AI systems that support complex industrial workflows.

A standout example is the “Startup Advisor,” an AI copilot that assists operators in the intricate process of starting liquefied natural gas plants. Rather than replacing human workers, these systems are designed to augment expertise. “We’re deeply thinking about how that work needs to be reimagined,” Melouney says. The company’s approach emphasizes scaling fast from small prototypes while maintaining a focus on human accountability and rethinking processes, not just bolting AI onto existing ones.

As AI systems become more autonomous, the companies that succeed will be those that built strong operational foundations long before the hype. Woodside’s ultimate goal is an autonomous enterprise where AI agents deeply interact with core workflows. This evolution from isolated experiments to enterprise-wide systems marks a significant shift, one that requires both technological and organizational change to realize the full potential of industrial AI.

As a parent, I see a stark contrast between my childhood and my daughters’. They are repulsed by smoking, a habit that was once everywhere in my youth—from my parents to cartoon characters. The UK’s new generational tobacco ban, part of the Tobacco and Vapes Act 2026, prohibits selling tobacco to anyone born after January 1, 2009, forever. This “endgame” approach aims to eliminate smoking entirely, not just reduce it. Yet no one knows if it will work. The Maldives tried it last year, and New Zealand’s similar law was repealed in 2024. Political support in the UK is mixed, with Nigel Farage vowing to undo it.

Supporters argue the ban protects future generations from addiction. Most smokers start as teens, regret it, and struggle to quit. Tobacco kills half its users and harms nonsmokers through secondhand smoke. Critics cite personal freedom, but advocates like Chris Bostic note that even health charities once dismissed the idea as impossible. The public health argument, says researcher Britta Matthes, is about freedom from addiction, not just freedom to smoke.

Experts agree that a multipronged strategy is best, including low nicotine limits and banning filters. But preventing teens from starting remains a powerful goal, backed even by many smokers. While the ban’s success is uncertain, it’s a step worth taking for the next generation.

The United Kingdom has passed a groundbreaking generational sales ban on tobacco products, aiming to eliminate smoking entirely rather than just reduce it. This 'endgame' approach represents a shift from traditional strategies like taxation and warning labels. However, experts remain uncertain whether the policy will achieve its ambitious goal of creating a smoke-free generation.

Jessica Hamzelou, a parent and journalist, reflects on how attitudes toward smoking have changed dramatically. Her children are repulsed by smoking, a stark contrast to her own childhood when smoking was culturally embedded. She supports the ban despite its uncertain effectiveness, noting that the policy is gaining traction globally and no longer seems as radical as it once did.

In other news, the latest issue of MIT Technology Review AI features a short story by award-winning author Elizabeth Bear. 'You Do Your Own Time' follows librarians who risk everything to preserve banned biographies of labor camp survivors. They send a solid-state drive into space to protect these erased histories from government surveillance and destruction, highlighting the power of narrative control and resistance.

Most large language models suffer from a groupthink issue, often producing the same predictable answers. For instance, asking a chatbot for a random number between 1 and 10 almost always yields 7. While this consistency works for coding or research, it stifles creativity during brainstorming or travel planning. Australian startup Springboards has developed Flint, an LLM trained to generate more diverse responses to open-ended questions like "Where should I go in Europe?" The company aims to push chatbots beyond obvious answers, offering a solution for users seeking variety. This innovation addresses a growing need for AI that can think outside the box, as reported by MIT Technology Review AI.

Springboards' Flint model stands out by prioritizing novelty over predictability. Unlike mainstream LLMs that default to common patterns, Flint uses specialized training to explore a wider range of possibilities. This could benefit tasks requiring creative input, such as event planning or idea generation. The startup's approach challenges the current trend of AI systems that often mirror each other's outputs, potentially unlocking new applications for users who value originality.

The broader implications of Flint's development highlight a shift in AI design. As LLMs become more integrated into daily life, their tendency toward uniformity can limit their usefulness. Springboards' method offers a path toward more dynamic interactions, though it remains to be seen how widely adopted such models will become. MIT Technology Review AI notes that this effort represents a step toward addressing AI's creativity gap, though challenges like maintaining accuracy while boosting diversity persist.

Ultimately, Flint's emergence signals a growing demand for AI that balances reliability with innovation. While mainstream models excel at structured tasks, Springboards' focus on variety could reshape how people use chatbots for exploratory purposes. As the field evolves, tools like Flint may help break the groupthink cycle, offering users a richer range of responses. MIT Technology Review AI continues to track these developments as they unfold.

Amazon has announced it will stop accepting new customers for Mechanical Turk, its crowdsourcing marketplace, effective July 30, 2026. The decision, attributed to “careful consideration” by Amazon Web Services, means existing users can continue operating normally, though no new features will be introduced. While the service remains active, the move signals a significant step toward its eventual phase-out.

Launched in 2005, Mechanical Turk paid workers small sums for tasks that resisted automation, such as completing CAPTCHAs or analyzing sentence sentiment. It later became a tool for data annotation in AI training via SageMaker. However, the platform also faced ethical scrutiny over labor practices and was linked to the Facebook-Cambridge Analytica scandal. Critics noted it enabled companies to pass off human-performed work as AI, echoing the original 18th-century hoax of a chess-playing machine.

The relationship between Mechanical Turk and AI grew increasingly ironic. A 2023 study found that 33% to 46% of workers were using large language models to complete tasks, undermining the platform’s purpose. Following Amazon’s announcement, online discussions suggested the service had already declined due to bots and fraud, with some predicting a full shutdown as Amazon deems server maintenance unworthy of continued investment.