AI Highlights — the whole picture — 2026-11-03 (Tue)

Funding and performance continue to drive expectations for AI. Tempus AI achieved its first profit in Q3 2026 and raised its full-year guidance. On the other hand, there are companies announcing layoffs due to AI, and companies simultaneously calling back employees to replace failed AI. There have been warnings of increasing isolation and reports of exam fraud using smart glasses. The yardstick for measuring the success or failure of implementation is shifting from investment amounts and announced plans to what actually happens to people. The following sections examine these changes in the following order: expectations, employment, human relationships, evaluation, and on-site accuracy.

Strength and fluctuation of enthusiasm shown by investment and performance

Tempus AI's financial results for the third quarter of 2026 are an example of expectations being backed up by the numbers. Revenue rose 23% to $411.5 million, giving the company its first profit and raising its full-year guidance to $1.6 billion to $1.605 billion (BigGo Finance). Sales were only slightly higher than expected (TradingView). Still, the stock price rose 10%, and the oncology field was reported to be experiencing "accelerating caseload growth" (TIKR.com). Additionally, Ark Invest, led by Cathie Wood, bought $12 million worth of the company's stock amid bullish Wall Street views (CoinCodex). The results did not exceed expectations by a large margin, but were praised for the combination of a slight upside, a return to profitability, and an increase in guidance. It can be seen that what the market is looking at is a direction towards profitability rather than the level of sales.

The flow of funds is also spreading to the surrounding areas. nPhase raises $30 million in Series D to expand its AI-native clinical data platform for life sciences globally (FinancialContent). On the infrastructure side, a $22 billion AI data center in Wisconsin is expected to create 1,500 jobs (tech-insider.org). From analyzing clinical data to developing computing resources, investments are moving toward both the application and infrastructure layers. However, the size of 1,500 jobs is small compared to the $22 billion. Readers should keep in mind when looking at a project that the size of the investment does not directly translate into the level of employment.

However, this movement is not uniform. ChinaTalk's Jordan Schneider discusses the $1 trillion drop in South Korean stocks and what it means for the global AI race (ChinaTalk | Jordan Schneider). Economists at BMO also noted that a slowdown in AI investment would ease inflationary pressure on the Fed (www.tokenpost.com). The latter view is based on the assumption that the momentum of investment was large enough to influence prices and monetary policy, and also acknowledges the possibility that that momentum may decline.

Tempus AI's return to profitability and increased guidance are evidence that individual companies are showing results. However, there are differences in the strength of expectations among countries and markets, and a downturn could occur. In that case, just because the amount of procurement or investment is large does not mean that the implementation was successful. The next thing to check is what the funds are bringing to the workers and the field.

Limits of replacement reflected in personnel reductions and recall

SUSE has announced plans to cut 10% of its workforce due to AI (The Register). This is an example of clearly stating AI as the basis for personnel planning, and shows that companies have begun to publicly state that the introduction of AI will be directly linked to reviewing personnel costs. Meanwhile, UkrMedia News reports that AI has failed the test and companies have started recalling staff. There are companies that announce cuts and companies that bring back the people they cut at the same time. Therefore, readers need to treat claims about how many jobs can be replaced by AI not as plans at the time of announcement, but as results that have been verified through operation.

The actions of banks and major corporations show a different side to the discussion of replacement. CNBC reported that fighting AI bots like Meta Muse may not be a winning strategy in the long term for JPMorgan and Amazon. As the headline suggests, both companies are under pressure to respond to AI bots. AI is not only a tool to improve the efficiency of your company's operations, but it is also a tool that can be used to improve the efficiency of your company's operations. If you only look at your company's implementation plan, you will overlook this external change. Hiring decisions also need to take into account not only the degree of automation within the company, but also the response to customers and competitors who use AI.

From the workplace side, Canadian HR Reporter asks, "Can AI make work more enjoyable?" Apart from the scale of reductions and the number of recall cases, the success or failure of the introduction will also be determined by how workers perceive the work that involves AI. If quality or burden issues arise after staff reductions lead to recall, the cost savings from the reductions may be offset by the costs of rehiring and retraining. However, it is unclear from the present data how SUSE reductions will change in the future.

Considering the examples above, whether employment decisions based on AI will continue will not be determined by the scale of the announcement. It is determined by the track record of passing the exam and the actual experience of the worker. Beyond the movement in the number of people through retrenchment and rehiring, the question remains as to how people are actually being affected.

Side effects of AI taking on loneliness and intimacy

The movement of AI entering social situations is starting to catch the attention of religious people and researchers. The Northwest Arkansas Democrat-Gazette reported that pastors are confronting AI lovers. It also covers the situation in which SNS is being offered as a ``digital solution'' to lonely young men. MindSite News reported that researchers have warned that AI chatbots could increase social isolation. The concern is that having someone always at hand to handle intimacy may reduce motivation to reconnect with other people. The situations where the conversation partner is an AI are not limited to romance and small talk. Walker Plus reported that the ``Omakase AI avatar'' interacts like a human and handles interviews and customer service.

On the other hand, the same technology can also be supportive. The Boca Raton Tribune profiles how AI and data are transforming child support in Palm Beach County. According to BigGo Finance, LG Uplus is testing its AI care infrastructure in nursing homes and is targeting $30 million to $37.5 million in senior business sales next year. Both child support and nursing care are situations in which the people being supported are people. In this case, AI is positioned as a tool to support human-to-human interactions, rather than a replacement for them.

When put side by side, it appears that the difference between warning against isolation and its use in the field is not so much the type of technology but the way it is used. What remains in the user's life will be different depending on how it is used to provide intimacy on behalf of another person and when it is used to help support the other person. Those considering introducing AI need to look not only at what AI can do, but also in what direction the human relationships of users will move. This is why the yardstick for measuring the success or failure of implementation is shifting from the amount invested to the actual impact on people.

매일경제 reported that exam fraud using smart glasses is occurring one after another. According to the newspaper, AI correctly answered more than 80% of questions on U.S. university entrance exams. Tests have been used to measure the test taker's ability, with the assumption that the test taker will be able to solve the test without any outside help. If an AI solves more than 80% of the questions and receives the answers through glasses, it will no longer be possible to tell whether the score indicates the individual's strength or the strength of the person using the tools. What readers should be looking at is not the number of fraud cases, but rather what the scores prove.

There is also an argument for shifting the object of measurement from ability to subjectivity. The New York Times published an article arguing that there is a good chance that AI could be sentient. The paper does not conclude that there is consciousness, but argues that the possibility cannot be ignored. The same paper also reported on an example of an art museum using AI to provide individualized viewing tours for each visitor. Scenes in which AI selects the display method according to the visitor are already in daily use. The task of confirming high performance is different from the task of deciding what kind of entity AI should be treated, and the measuring stick for the latter has not yet been determined.

The question of who decides the rules also appears in politics and regulation. According to the New York Post, the AI giant donated nearly 100 million yen to the California Democratic Party. The timing is just before the election. British authorities are investigating Meta under antitrust law. According to PYMNTS.com, the issue is the conditions under which a publisher's content can be used in AI. On the one hand, companies with money are lobbying politics, and on the other hand, authorities are encroaching on the terms of deals. Both indicate that the conditions for using AI are no longer determined solely by companies' independent decisions. Readers need to look not just at company announcements and performance comparisons, but also at whose funding and whose authority is shaping the terms.

Test scores, awareness, donations and research. What all three stories have in common is that it remains unclear who has the standards to measure the value of AI. If such standards are not determined, the next question becomes inevitable: what determines the success or failure of the introduction?

Improving accuracy and disparity in clinical and research settings

In the interpretation of ultrasound images, an AI agent that handles both images and language has been reported as a way to make interpretation faster and more accurate (Medical Xpress). Ultrasound is a test that is easily influenced by the technique of the person conducting the test and the experience of the person reading it. If the speed and accuracy of image interpretation improves, even facilities with a limited number of specialists will be able to quickly obtain information for making decisions. For those considering introduction, the questions shift from whether AI can read images to which steps of image interpretation should be delegated and which steps should be checked by humans.

In the field of oncology, an MRI-based deep learning model was reported to predict Ki-67 expression levels in intrahepatic cholangiocarcinoma (Abdominal Radiology). Ki-67 is an index that measures the growth rate of tumors, and if it can be predicted from MRI, it will increase the amount of information that can be used when considering treatment plans. Again, the value lies in the expanded range of information that can be obtained from image tests. However, it cannot be determined from this study alone whether the reported model maintains the same accuracy with images from other facilities or equipment. The actual effects will depend on whether the facility implementing the system has a procedure for verifying it using its own data.

Improvements in accuracy may not extend equally to all patients. A paper published in JCO Global Oncology points out that there are disparities that cannot be resolved by implantation alone, and uses the somatic cell evolution score as an example to argue that sequencing that is representative of the population is necessary. If the data used for learning is biased toward a specific group, the numbers produced by the model will also be biased. Even if the average accuracy is high, the results can be poor in groups that are not well included in the data. This point means that in addition to overall accuracy, accuracy for each group should be included in the criteria for deciding whether to introduce the system.

What all three materials have in common is that they simultaneously address the effects of increasing the speed and accuracy of diagnosis, as well as the conditions to which patients this effect will reach. If the era in which the value of AI is measured solely by investment amounts and performance numbers is coming to an end, the next question is how its effects will be manifested on the human side, and for whom they will not be manifested.

What each passage has in common is that the value of AI is beginning to be judged not by investment amounts or announced plans, but by its actual effects on humans. In the financial results, the direction toward profitability was evaluated more than the level of sales. In terms of employment, reductions were announced and recall calls were made, and claims about the number of people to be replaced were subject to operational confirmation. In social relationships, it has been pointed out that taking over intimacy can deepen isolation, and in exams, it is no longer clear whether scores reflect one's own strength. In clinical practice and research, improved accuracy was discussed along with confirmation that it would be maintained in other facilities and equipment. The success or failure of its introduction will depend not on the magnitude of what AI can do, but on whether we can continue to see what it brings to people and systems.

Q1: What indicators and over what period of time are you checking the results of the introduced AI, not as a plan at the time of announcement, but as a result of operation? Q2: What procedures and records can be used to confirm that the deliverables and evaluation results using AI demonstrate the skills and quality of the person in question? Q3: Apart from investment amount and sales growth, what criteria do you use to judge the actual effects of implementation on people and organizations, and how do you set the conditions for review?

Finally, three questions

- What indicators and over what period of time are you checking the results of the introduced AI, not as a plan at the time of announcement, but as a result of operation? Q2: What procedures and records can be used to confirm that the deliverables and evaluation results using AI demonstrate the skills and quality of the person in question? Q3: Apart from investment amount and sales growth, what criteria do you use to judge the actual effects of implementation on people and organizations, and how do you set the conditions for review? - What procedures and records can be used to confirm that the results of AI-based products and evaluations demonstrate the individual's ability and desired quality? Q3: Apart from investment amount and sales growth, what criteria do you use to judge the actual effects of implementation on people and organizations, and how do you set the conditions for review? -Aside from investment amounts and sales growth, what criteria do you use to judge the actual effects the implementation has had on people and the organization, and how do you set the conditions for review?