AI Highlights — the whole picture — 2026-10-28 (Wed) Daily News

Who will bear the cost of AI investment? According to Barron's, Microsoft aims to cut 60,000 jobs to pay for investments, and Reuters reports that Deutsche Bank has used AI to cut some corporate loan decisions to minutes. Decisions are being made quickly, and at the same time costs are being expressed in numbers for personnel and prices. On the other hand, people are redrawing the lines of how they should interact with AI, such as by establishing regulations that prohibit abuse, without recognizing that AI is a sentient being. From here, we will spend half a day looking at China's stance as a nation, the process of drug discovery and diagnosis, and the line drawn to AI, which plays a role in intimacy. What comes to mind throughout this is not how much the efficiency numbers have increased, but rather who is responsible for what behind the speed and automation. The examples in each section show that the answer is not yet determined.

Investment costs go towards personnel and prices

The cost of AI investment is first expressed in terms of personnel numbers. According to Barron's, Microsoft aims to cut 60,000 jobs to pay for this investment. In the games industry, Destructoid reports that EA employees are complaining that morale is "abysmal" due to a combination of layoffs and the push for generative AI. The costs aren't just within the company. According to Investing.com, Goldman expects the rise in AI investment to increase inflationary pressures. This is only the company's forecast and is not a confirmed result. However, it shows that the burden of investment can also be felt on people who are not directly involved in the investment, such as employee employment and morale, as well as prices.

On the other hand, the benefits of introducing it are manifested in the speed of decision-making in the field. According to Reuters, Deutsche Bank has used AI to shorten some corporate loan decisions to just a few minutes. The reduction in time is an outcome that is easily visible to both companies receiving loans and bank employees responsible for screening. However, it is not clear from these numbers who is responsible for the shortened decisions. The HR Director introduces the argument that AI agents also need annual personnel evaluations. The idea is to apply the same goal setting, performance confirmation, and correction that is applied to humans to AI. This can be read as an argument that if we are entrusted with making faster decisions, we need a system to check those decisions.

When we line up the five materials, we see that gains and burdens appear in different places. Gains are likely to be quantified as results of the department that introduced the system, such as a reduction in examination time. The burden will be felt in different ways: the scale of the 60,000 job cuts, the decline in morale that EA employees complain about, and the impact on prices that Goldman predicts. This allocation cannot be determined just by looking at the efficiency numbers on the implementation side. What readers should check is who is being addressed in the implementation design: those who will be subject to reductions, those who will bear the burden in the form of prices, and those who will review the decisions made by the AI.

If the question is how to allocate costs and benefits, the next thing to consider is how far the person receiving the AI's judgment in the field should leave that judgment to them, and where they should draw the line.

National stance regarding computational resources and data

Regarding China's AI strategy, orfonline.org argues that China is aiming to create international rules that go beyond competition in computing resources and performance (orfonline.org). On the other hand, if we look domestically, there are other issues. According to the South China Morning Post, data is essential for China's AI race, and the question is whether the newly established state-owned giants can play a leading role. The nation's stance needs to be viewed from two perspectives, as outwardly it aims to form international rules and inwardly it aims to operate a data infrastructure. What readers should be looking at is not so much the scale of China's computing resources, but rather who will take on the role of collecting data and organizing it into a usable form. Whether the newly established state-owned majors will actually play this role cannot be determined until we see how they operate going forward.

On-site norms and operations have also become more specific in various places. According to IT Brew, AI safety organizations are tracking when each AI model "cheats" during a task. A model's performance cannot be measured by scores alone, and safety evaluations are now beginning to include checking how the model accomplished its tasks. In the field of reporting, UNESCO is supporting Mongolian news organizations to use AI ethically (UNESCO). In agriculture, DAT-AI-LAB at the University of Pennsylvania School of Veterinary Medicine has launched two new initiatives in agricultural technology (Lancaster Farming).

The people responsible for this are not the nation, but their own respective places, including organizations that track the issue, international organizations that support media ethics, and university research laboratories close to agricultural fields. What each case has in common is that it is not a question of whether or not to use AI, but rather who decides on the confirmation procedures and boundaries when using AI. For those considering the introduction, in addition to choosing their company's model, it is important to first decide on a confirmation procedure to detect fraudulent achievements and a person in charge of determining the scope of use.

While nations vie for control over data and international rules, those on the ground are setting their own norms first. In the next section, we will look at how individuals redraw that line for themselves.

Research and drug discovery progress to the stage of automation

A virtual biotech company deploying 37,000 AI agents throughout the drug discovery process was reported on Bioengineer.org. The idea is that agents, rather than humans, will be responsible for everything from target search to design and verification, and the scale of AI is measured in terms of the number of processes rather than the number of people. On the molecular design side, the computational design and experimental validation of a photoresponsive PARP1 inhibitor was published in the Journal of the American Chemical Society. A single piece of research includes testing molecules designed through calculations through experiments. For readers involved in drug discovery, what should be considered is not the number of agents, but rather which processes should be left to machines to make decisions, and which processes should be verified by humans.

Even in the field of diagnosis, it is becoming more and more integrated into the process. The Journal of imaging informatics in medicine published a study on predicting VETC and prognosis of hepatocellular carcinoma using deep learning radiomics based on Gd-EOB-DTPA contrast-enhanced MRI. VETC is an index that indicates the characteristics of blood vessels within a tumor, and is unique in that it attempts to predict the prognosis in a continuous manner from images. Diagnostic Imaging summarizes the latest developments in the imaging field under the title ``Advances in AI October 2026.'' In diagnostic imaging, the position at which predictions are produced is moving from behind the image reading to the front, and the extent to which the output is used is a design issue that is determined by the user and each facility.

In terms of preparation for those who can master the technology, Forbes lists ``6 AI-related skills to hone now to increase income in 2027,'' expressing the view that AI-related skills will lead to higher incomes. The more AI enters the process of research and diagnosis, the more the results will be influenced by the skills of the people who check and use the AI output.

The more automated a field becomes, the more likely it is that it becomes unclear who is responsible for the results. In the next section, we will look at the movement in which people are re-determining their own distance and boundaries with AI.

How do people draw the line when it comes to AI that is responsible for intimacy and emotions?

In relation to AI, the lines that people draw are not aligned. According to theticker.org, which reported on the new research, humans do not accept that AI is conscious and draw the line there. Meanwhile, Anthropic will implement a provision prohibiting abusive behavior towards Claude on November 12th (BigGow Finance). Consciousness is not allowed, but abuse is prohibited. Although these two are side by side, they are not necessarily judgments based on the same premise. STAT also addressed the question of whether long-term use of AI will lead to isolation, and reported that researchers are calling for greater data disclosure. In other words, the use and regulations are moving ahead, even though there is still not enough material to measure the impact of intimate use. When readers decide how to use AI in their companies or homes, they don't have to wait for the evidence to be ready before drawing the line.

In the field of emotions, the role of AI is already appearing in concrete products and research. According to 36Kr, Banxing is a business that matches AI love practice partners with real people, and is currently seeking angel funding. NTT Docomo announced that it has established a "high EQ voice dialogue technology" that improves responses by sensing situations (NTT Docomo). Voice dialogue that shows empathy has become a practical technology. Meanwhile, a study by Canadian HR Reporter found that AI negotiation advice tended to encourage women to seek care-related benefits. The more empathetic the advice, the more the user needs to check where the direction is coming from.

Even in the details of daily life, AI is becoming a confidant. Apartment Therapy shared an experience where they asked an AI how to clean out a storage closet and it actually solved the problem. Here, the user can immediately check whether the advice was helpful or not. In contrast, in the conflict between a Canadian venue and AI art reported by Exclaim!, the issue was over who should make concert posters and how. Apart from whether or not it is useful, the question is whether there is an agreement on who should be entrusted with the creation of the work and under whose name it should be published.

The range of things that can be left to AI is different between situations where you can check the results yourself, such as advice on tidying up, and situations where it is difficult to see where the decision will be made or where responsibility lies, such as love affairs, benefits, and creation. Efficiency figures can be used as a guide for the former, but for the latter, a design is first needed to decide on the underwriter.

The examples in each section come to the same point. The costs of AI investments can also extend to people who are not parties to the investment, in the form of personnel and commodity prices. As the nation prepares for computing resources and data, the question is who will take on the role of organizing the data. In drug discovery and diagnostics, the deciding factor is which processes should be left to machines to make decisions, and which processes should be verified by humans. In areas that deal with intimacy and emotions, usage and regulations are in motion before the evidence is available. In all cases, efficiency numbers only tell part of the results. Even though decisions have become faster or the number of agents has increased, they alone do not tell us where the burden and responsibility lies. What readers should look at is not the numbers that result from the introduction of AI, but how they designed who would take on what. Its design affects not only those who decide to introduce it, but also those who use it, those who control it, and those who indirectly receive the costs.

Q1: In my work, is there currently a clear separation between the process of receiving AI output and making decisions, and the person responsible for the results of those decisions? Q2: Are you aware that the lines you have drawn regarding the use of AI were decided before the evidence was available, and have you set the conditions and timing for reviewing them? Q3: Of the efficiency figures expected from AI investment, who among employees, business partners, and customers is designed to bear the cost and responsibility?

Finally, three questions

- In your work, is there a clear separation between the process of receiving AI output and making decisions, and the person responsible for the results of those decisions? Q2: Are you aware that the lines you have drawn regarding the use of AI were decided before the evidence was available, and have you set the conditions and timing for reviewing them? Q3: Of the efficiency figures expected from AI investment, who among employees, business partners, and customers is designed to bear the cost and responsibility? - Are you aware that the lines you have drawn regarding the use of AI were decided before the evidence was gathered, and are you setting conditions and timing for reviewing them? Q3: Of the efficiency figures expected from AI investment, who among employees, business partners, and customers is designed to bear the cost and responsibility? - Of the efficiency figures expected from AI investment, who among employees, business partners, and customers is designed to bear the cost and responsibility?