AI is reshaping industries from healthcare to hiring, but the results so far are mixed. In asset management, only 8 percent of fund managers report that AI has actually improved investment returns, according to a September 2026 BluePaper report by Morgan Stanley and Oliver Wyman. While 55 percent of asset managers have integrated AI into their workflows, most gains are limited to operational efficiency and research, with core decisions like portfolio construction and trade execution seeing just 13 percent and 7 percent adoption respectively.
On the coding front, costs are coming down. Jellyfish found that the cost of a merged pull request fell 26 percent even as AI spending surged 60 times. Heavy AI users reported 67 percent more throughput and a 32 percent drop in cost per PR. However, revert rates barely changed, leaving quality questions wide open. The data also does not control for team size, model mix, or pricing shifts, so it is still unclear at what point tokens stop buying cheaper delivery.
Chinese tech companies are pushing AI hardware into the real world. Astrall Dynamics won an IFA 2026 award for its Hypertron-T01 quadruped robot, which can carry up to 80 kilograms and handle rough terrain. The robot has already been tested in power substations, and Astrall plans to open SDK interfaces to European partners to speed up industry-specific solutions. More than 1,000 exhibitors from China and Hong Kong took part in IFA 2026.
In healthcare, the human element remains a point of contention. Retired doctor James Isbister, who spent four decades in medicine, says Evidence-Based Medicine started with good goals but became a regulatory framework that pushed away the art of understanding each patient. He worries AI will outperform humans in every way EBM defines success and calls for keeping the human side of medicine alive. Australia is also tightening rules: the Therapeutic Goods Administration is reviewing 12 AI medical scribe products and found eight likely need to be classified as medical devices, raising safety concerns.
Cybersecurity experts urge stronger safeguards as AI risks grow. Etay Maor of Cato Networks says enterprises should apply zero-trust principles to AI systems, with restricted access and continuous evaluation. He warns that a kill switch is not reliable because harm can happen before it triggers. Maor calls for mandatory incident reporting and independent evaluations. Fortanix released CoCo, a tool that protects AI models and applications on Kubernetes from infrastructure operators, using confidential VMs and integrated attestation. The project has been CNCF incubating since July and can run as a cloud service or in fully air-gapped environments.
AI is also moving deeper into hiring. Companies now use AI to screen resumes, book interviews, and ask initial questions before a human gets involved. The shift marks a significant change in recruitment, though the article notes that the question of when human judgment remains necessary is still open. Experts overall say they are concerned but not afraid of AI's future, warning that it can power scams, attacks on power grids, and even biological weapons, while also believing it could help cure cancer and solve hard problems.
Key Takeaways
- Only 8 percent of fund managers report improved investment returns from AI, despite 55 percent integrating it into their processes
- AI coding costs fell 26 percent per merged pull request, but revert rates barely changed, raising quality concerns
- Astrall Dynamics won IFA 2026 award for Hypertron-T01 quadruped robot capable of carrying 80 kilograms
- Australia's TGA found 8 of 12 AI medical scribe products likely need medical device classification
- Etay Maor of Cato Networks urges zero-trust principles for AI systems and mandatory incident reporting
- Fortanix launched CoCo to protect AI models on Kubernetes using confidential VMs and integrated attestation
- Just 13 percent and 7 percent of asset managers use AI for portfolio construction and trade execution respectively
- AI is increasingly handling early hiring steps including resume screening and interview booking
- Retired doctor James Isbister warns AI could outperform humans in every metric Evidence-Based Medicine defines
- Experts warn AI can power scams, grid attacks, and biological weapons while also potentially helping cure cancer
Retired doctor reflects on Evidence-Based Medicine and AI
James Isbister spent four decades in medicine and watched the field change. He says Evidence-Based Medicine started with good goals but became a regulatory framework. This shift pushed away the art of understanding each patient as a person. He worries that AI will outperform humans in every way EBM defines success. He calls for keeping the human side of medicine alive instead of letting AI take over completely.
AI coding costs fall but quality questions remain
Jellyfish found that the cost of a merged pull request fell 26% even as AI spending rose 60 times. Heavy AI users saw 67% more throughput and a 32% drop in cost per PR. Revert rates barely changed, so quality is still unclear. The data does not control for team size, model mix or pricing shifts. The key question is at what point tokens stop buying cheaper delivery.
AI drives global growth for Chinese tech companies
Chinese AI hardware companies are showing real-world strength beyond labs and gadgets. Astrall Dynamics won an IFA 2026 award for its Hypertron-T01 quadruped robot built for heavy loads and rough terrain. The robot can carry up to 80 kilograms and has been tested in power substations. The company plans to open SDK interfaces to European partners to speed up industry-specific solutions. More than 1,000 exhibitors from China and Hong Kong took part in IFA 2026.
Experts weigh risks and promise of artificial intelligence
AI and cybersecurity experts say they are concerned but not afraid of AI's future. They warn that AI can power scams, attacks on power grids and water systems, and even biological weapons. AI agents that act on their own raise extra worries about unintended harm. At the same time, experts believe AI could help cure cancer and solve hard problems. They urge safeguards to keep pace with fast-growing AI capabilities.
Enforceable AI accountability and zero-trust security urged
Etay Maor of Cato Networks says AI-driven cyber risks need stronger rules. He urges enterprises to treat AI systems with zero-trust principles like restricted access and continuous evaluation. He warns that a kill switch is not reliable because harm can happen before it is triggered. Maor calls for mandatory incident reporting, independent evaluations and federal action. He says organizations must stop treating AI like ordinary software and plan for it to act against its creators.
Fortanix launches tool to hide AI data from admins
Fortanix released CoCo, a new project that protects AI models and applications on Kubernetes from infrastructure operators. Since July, this CNCF incubating project has focused on securing data at rest and in transit during CPU and GPU workloads. The solution uses confidential VMs and integrated attestation to ensure trusted key management without extra work for teams. Companies can run this system as a cloud service or in fully air-gapped environments to keep intellectual property safe.
Only 8 percent of fund managers see AI boost returns
A new report by Morgan Stanley and Oliver Wyman shows that only 8 percent of fund managers report improved investment returns from using AI. While 55 percent of asset managers have integrated AI into their processes, most gains are limited to operational efficiency and research capabilities. The September 2026 BluePaper report found that 69 percent saw better efficiency and 55 percent gained better insights, but only 24 percent reduced costs. Core investment decisions like portfolio construction and trade execution still use AI very little, with just 13 percent and 7 percent adoption rates respectively.
Australia tightens rules for AI medical scribe tools
The Therapeutic Goods Administration in Australia is reviewing 12 AI medical scribe products and found eight likely need to be classified as medical devices. These products were not previously registered as such, which raises safety concerns for their use in healthcare. The TGA plans to work closely with manufacturers to ensure these tools meet strict safety and efficacy standards. No specific timeline has been set for when this review will be completed.
Artificial intelligence takes over early hiring steps
Artificial intelligence is increasingly handling the first stages of the hiring process, including screening resumes and booking interviews. Machines are now asking initial questions to candidates before humans get involved in the selection. The video explores how much of the hiring workflow is automated and when human judgment remains necessary. This shift marks a significant change in how companies manage recruitment today.
Sources
- A retired physician's reflections on Evidence-Based Medicine and AI
- Tokens Are Cheap, But Is the Math Real?
- AI brewing fresh global growth for Chinese tech
- What The Tech: What could go right when it comes to Artificial Intelligence
- Need for enforceable AI accountability, zero-trust security as AI-driven cyber risks threaten critical infrastructure
- Fortanix Confidential AI isolates workloads from infra operators
- Only 8% of fund managers report improved investment returns from AI adoption
- newsGP - AI medical scribes face stricter TGA oversight
- AI is hiring now
Comments
Please log in to post a comment.