政治 ニュース: 4 9月 2025

October 21, 2025, 7:20 AM EDT. Cory Doctorow's term enshittification describes a pattern where platforms evolve from user-friendly gateways into extractive middlemen. The article explains the cycle: early-stage platforms are flush with investor cash, prize growth and network effects, and offer real value to users. In Stage 1 they lure users with cheap deals or privacy promises, exemplified by Facebook, Google, Amazon, and Uber. As these services scale, they shift toward monetization from advertisers and third-party merchants, often through predatory pricing to crush competitors. The core mechanism is locking in users via high switching costs and opaque algorithms, making user churn costly. The piece frames enshittification as a recurring trend across tech platforms, impacting the openness and usefulness of the internet.

October 21, 2025, 7:18 AM EDT. At Fortune's Most Powerful Women conference, Microsoft's Amy Coleman, Bloomberg Beta's Karin Klein, and Sola's Jessica Wu pushed back against the MIT-style belief that most AI pilots fail. They framed high failure rates as part of learning how transformative tech works, not a signal of doom. As Klein asked, has anybody ridden a bike on the first try? The trio urged embracing experimentation, culture, and vibe coding-using accessible AI tools to build without traditional coding. Wu noted that only about 5% of tools reach production, a gap similar to historical enterprise IT success rates, around 10% or lower. The conversation highlighted agentic process automation, the surge in AI experimentation, and the need to focus on joyful, toil-reducing outcomes and organizational readiness, not perfection at first rollout.

October 21, 2025, 7:16 AM EDT. Three leaders from Microsoft, Bloomberg Beta, and Sola argued at Fortune's Most Powerful Women conference that high failure rates aren't a bug in AI adoption but part of learning how transformative tech works. They pushed back on the MIT study's claim that 95% of enterprise AI pilots fail, noting that only a minority reach production while historical IT deployments have struggled too. Framing AI as an early innings journey, they stressed experimentation and the importance of cultivating culture over gadgetry. The panel urged embracing vibe coders who build with accessible AI tools, while Jessica Wu from Sola highlighted agentic process automation and the vast volume of experimentation underway. In summary, near-term wins will come in waves as organizations learn and adapt.

October 21, 2025, 7:14 AM EDT. Three leaders from Microsoft, Bloomberg Beta, and Sola push back on the MIT study that ~95% of enterprise AI pilots fail. They argue that failure is part of learning how transformative AI really works, not a fault in the tech. The panel frames this as an ongoing journey: expect an initial trough, then wins as teams experiment and evolve. Emphasis on experimentation and a culture that supports vibe coders-nontraditional users building with accessible AI tools-may matter more than the underlying algorithm. Jessica Wu notes that only about 5% of tested tools reach production, but historic IT deployments hovered around 10% or lower, suggesting broader challenges exist. Sola's agentic process automation reflects the scale of current experimentation and the inevitability of low short-term success.

October 21, 2025, 7:12 AM EDT. At Penn's CHIBE Roybal Retreat, Harvard's Elizabeth Linos urged researchers to use AI to turn evidence into actionable tools for policymakers. The two-day gathering gathered behavioral scientists from Penn and beyond to share latest findings. Linos argues policymakers care about cost, feasibility, political risk, and implementation, not just rigorous methods, so AI can guide decision-makers through what they don't know. She highlighted Policy Bot, a prototype at Harvard Kennedy School that uses AI to summarize studies and prompt considerations like cost-effectiveness and replicability. The aim: bridge the gap between research and real-world uptake, recognizing that incremental interventions often win larger-scale adoption. The field seeks robust AI-enabled decision-support systems that policymakers trust and use at scale.