AI Highlights — the whole picture — 2026-10-12 (Mon)

AI investment in the US has reportedly reached $1 trillion. Just looking at the size of the funds, it seems that the speed at which human jobs are transferred to machines is increasing as well. But Apple has shelved plans to replace 5,000 AppleCare advisors with AI. AI is also making inroads in medical care, nursing care, classrooms, and farmland, but there is no rush to decide on its replacement. The amount of investment indicates how much the technology has become available, but it does not indicate what people will continue to do around the technology. The key to reading this half-day article is not the size of the amount, but rather what remains as human work in the workplace after AI is introduced, and what continues to be judged by humans.

Huge investment and workplaces that are not in a hurry to replace

U.S. AI investment has reached $1 trillion, according to The Hill. Just looking at the inflow of funds, it seems that the movement to replace human labor with machines is rapidly progressing. However, judgments on the ground do not follow that simple reading. According to Forkast News, Apple has shelved plans to replace 5,000 AppleCare advisors with AI. It's one thing to have a huge amount of money in play, but it's another to make a decision to reduce staff.

The same trend can be seen on the outsourcing base side. BusinessLine reports that when introducing AI, GCCs (Global Capability Centers) located in India and other countries have begun to focus on improving productivity rather than reducing headcount. The article also points out that office demand will depend on how flexible the office is. In other words, the introduction of AI does not necessarily directly lead to an increase or decrease in the number of employees, and the size of the workforce is treated as a variable that is adjusted according to changes in demand. The way we predict employment contraction based on the amount of investment needs to be reconsidered.

In individual implementation cases, AI is being incorporated as a tool to supervise and assist humans, rather than to replace them. According to Morningstar, IgniteUps.ai has announced NEXUS, a management tool for car dealerships that oversees AI human resources. Even though the object of management is AI, the design is such that the role of bundling and operating it remains with humans. In finance, Bank of America has expanded its treasury functions using AI and issued new corporate bonds, according to Yahoo Finance. It is notable that the decision to raise funds and strengthen functions is being talked about as an expansion of existing operations rather than a replacement of personnel.

What readers should look at is not the size of the investment, but rather which operations and in what form the funds are being used. Replacement plans have been shelved, the purpose of implementation has shifted to productivity, and new positions have been created to manage AI. The size of the investment reflects not speed, but the fact that deployments are being re-designed around the role of people. So, how exactly will the jobs that people perform change after the replacement of people has receded? We'll look at that in the next section.

Penetration into practical practices in medical care, nursing care, and agriculture

Pancreatic cancer is notoriously hard to detect, but Medical Economics featured Mayo's AI's early detection of pancreatic cancer alongside an article about Purdue Pharma heading for dissolution after $5 billion sentencing. In the nursing care setting, Skilled Nursing News reports that Trilogy and Signature are using AI for fall detection and auditing to improve facility operations. In agriculture, a lightweight AI model has been developed that uses smartphones to determine the severity of sugarcane diseases, according to Bioengineer.org. In both cases, AI is placed in areas of existing work that are likely to be overlooked or checked.

Specialized journals continue to report on stratification, which goes one step further than detection. A study using temporal radiomics to track liver metastases during MR-guided radiotherapy was published in Physics and imaging in radiation oncology. For lung cancer, a study was published in Frontiers in medicine that developed and internally verified a machine learning model that predicts death three years after resection, and turned it into an online risk calculator and nomogram. For pancreatic ductal adenocarcinoma, an attempt to stratify prognosis and treatment by phenotypic classification of tumor-stromal ratio using super-resolution MRI was published in MedComm, and a systematic review of MRI radiomics for surveillance therapy for prostate cancer was published in European journal of radiology.

There is also research that supports intraoperative and preoperative decisions. A feasibility study using intraoperative fluorescence spectroscopy for IDH classification of gliomas has been published in the British Journal of Neurosurgery, and a study has been published in Abdominal radiology that developed and externally validated a CT radiomics and topology-based intratumoral heterogeneity (ITH) model for preoperatively predicting nuclear β-catenin expression in hepatocellular carcinoma. The fact that the latter has undergone external verification indicates that it is getting closer to practical use.

What they all have in common is that AI provides detection and prediction candidates, and humans are responsible for confirming diagnoses, selecting treatments, and making facility management decisions. What readers should look at is the division of roles in which processes AI improves accuracy and who handles the results. How this division of labor is actually rearranged will be seen from the field side in the next section.

Reexamining ability and heart in the classroom and in daily life

In the classroom, the more work that is entrusted to AI, the more questions are asked about what people will learn. The Free Press Journal featured ``AI Transforming Classrooms'' and reported that educators are emphasizing critical thinking, emotional intelligence, and mental health. This means that the emphasis has shifted to the ability to question the answers given, face others, and maintain one's own state, rather than the speed at which answers are given. Meanwhile, in Nepal, an AI literacy pilot project was completed and The Rising Nepal advocated for inclusive digitization. Even if the content to be taught is decided, if some people will reach it and others will not, redefining ability will only be a matter for a few people. Readers looking at schools and training need to check not only what they teach, but also who they reach.

The same question exists on the work side. Yahoo! News reported that a manga artist who lost his job to AI turned to farming and learned the pain and joy of physical labor while harvesting peaches. In the field of music, the founder of CITY JAM said that although AI has made production faster, his true asset is his own sound (Morningstar). The former changes the job itself, and the latter places value on the unique sound rather than speed within the same job. Even in a situation where replacement of people is progressing, it is up to the individual to decide where they will find the part they can call their own work.

On the emotional level, concern and anxiety appear side by side. mos.ru has a page featuring videos and testimonials of AI girlfriends and AI partners, suggesting that people are increasingly seeking intimacy from AI in their lives. On the other hand, CNBC reported that Hollywood is making movies about Zuckerberg, Musk, and Altman, and concerns about AI are spreading. The more the story of the people who create AI is made into a movie, the more restless society becomes about its future.

What is common in classrooms, fields, studios, and love topics is not a list of what AI can do, but rather what people rely on to determine their own roles. The next point to consider is whether this reliance should be left to individual efforts, or whether it should be supported by systems and workplaces.

Governance surrounding bias, beliefs, and state involvement

Discussions surrounding AI are shifting from performance and investment costs to questions about how to approach the technology. Professor Claudia Flores argued at the United Nations that AI is not gender neutral (Yale Law School). If there is a bias in the learning data or design, that bias will appear as a judgment in hiring, evaluation, and service provision. For those implementing AI, in addition to comparing accuracy and cost, it is necessary to ascertain who may be disadvantaged.

It is not just the system that is in question. David Friedberg predicts that the belief that AI is sentient will give rise to a new religion (OfficeChai). This is just his prediction and is not a confirmed fact. Still, it can be seen that the difference between whether AI is treated as a tool or as an object of trust can influence users' judgment and degree of dependence. Both pointing out bias and predicting beliefs overlap in that users must decide for themselves how much to accept the output of AI.

Policies in each country are beginning to incorporate these questions into their systems. The developments surrounding AI safety since the Hugging Face attack were reported in chronological order (News4JAX). Vice President Vance attended the AI Summit in Paris on his first overseas trip (WNWO). This suggests that after the incident surrounding security, the leaders of various countries have begun to take a stance at international conferences. It is reported that China will include AI employment measures in its five-year plan and begin implementing them (FourWeekMBA). This means that ensuring safety and maintaining employment have begun to be treated within the same policy framework.

From these three movements, there is material that readers can use to make decisions. For companies, how safety and employment are incorporated into regulations in the countries and regions where they are introduced can become a prerequisite for their business. For working people, the role policies leave behind for people has a greater impact on their actual work than the speed of replacement. The three issues of being wary of bias, distancing from one's beliefs, and state involvement all converge on the question of who is responsible for AI's behavior. The next issue we need to examine is how each person on the ground will take on that responsibility.

The examples in each section point in the same direction. Although the scale of investment was large, the replacement of personnel was halted, and the areas where AI was added were limited to areas where there were many oversights and confirmation tasks. In education, the ability to question answers and face others is valued more than the speed with which answers can be given. The predictions about bias and belief in AI show that it is up to the users to decide how much of the output they want to accept. If we look at whether humans will be replaced by AI, we cannot grasp what is happening on the ground. The role of people is shifting from executing tasks to confirming, suspecting, and accepting tasks. What we should look at is not the increase or decrease in the amount of investment, but who is assigned the role and who is reaching them.

Q1: What part of my work should I check and ultimately take responsibility for even after I leave it to AI, and is this clearly stated? Q2: When bias or error is found in the output of AI, who has the authority to detect it and what criteria are used to stop its use? Q3: What indicators do you use to continually check changes in the roles played by people, rather than the amount of investment in AI?

Finally, three questions

- What parts of my work do people check and ultimately take responsibility for even after leaving it to AI, and is this clearly stated? Q2: When bias or error is found in the output of AI, who has the authority to detect it and what criteria are used to stop its use? Q3: What indicators do you use to continually check changes in the roles played by people, rather than the amount of investment in AI? - When a bias or error is found in the output of AI, who has the authority to detect it and what criteria are used to stop its use? Q3: What indicators do you use to continually check changes in the roles played by people, rather than the amount of investment in AI? - What indicators are you using to continually check changes in the role played by people, rather than the amount of investment in AI?