AI Highlights — the whole picture — 2026-10-24 (Sat) Daily News

Warp Sciences has raised $215 million to build a lab that uses AI and robots to run chemical experiments. During the same period, American Express was reported to be cutting 5% of its workforce, and Colliers warned that a lack of AI talent could hinder growth in Europe. In a place where funding and expectations are high, a reshuffle of personnel and a shortage of personnel are occurring at the same time. What the articles we will read in this half-day have in common is the view that the difference between success and failure is less about how much money is invested in AI and more about how the investment goes through on-site verification and by whose hands it is run. Research in diagnosis, speech classification, and prediction also confronts readers with the differences between results under appropriate conditions and results in real environments.

Partnership with huge funds gathered in drug discovery AI

Warp Sciences raises $215 million to build an AI-powered drug discovery lab (citybiz). According to BRIDGE, the company is a company that uses AI and robots to conduct chemical experiments for drug discovery, and the company raised funding through a Series A led by Lightspeed. At the Series A stage, more than $200 million was raised. The destination of the funds is also a point of interest. The aim is not only to develop software and models, but also to build labs to actually perform experiments and automate chemical experiments.

Another company, Pathos, an AI drug discovery company, raised $237 million at a valuation of $3 billion (STAT). Reuters reports that Pathos will partner with Bristol Myers and Tempus AI to advance drug development (Reuters). Here, the procurement amount and valuation value include the practical ties of a partnership with a major pharmaceutical company. The reason why funds are being raised is not just the promise of AI alone. It appears that evaluation will also include how it will be incorporated into the development process of pharmaceutical companies.

The total amount raised by the two companies comes to $452 million. This scale shows that expectations for drug discovery AI go beyond the superiority of algorithms and extend to investments in the entire business, including experimental equipment and partners. What readers should take away from these numbers is not the high valuation itself. The question is whether the funds are being used in the form of automated experimental setups or collaborations with big pharmaceutical companies.

However, all of the materials presented this time are announcements of procurement and collaboration, and are not reports on drug discovery results. While the funds will be collected first, it is necessary to confirm separately whether the human resources to run the lab and the verification at the actual development site will proceed at the same speed.

Personnel reductions and human resource shortages progress simultaneously

American Express will cut 5% of its workforce, Demokratis.co.id reports. The reason the headline indicates is that it is AI compatible. This reduction indicates that the introduction of AI has not only improved operational efficiency, but has also entered the stage of reorganizing the composition of employment itself. 5% is a percentage of all employees, not a monetary amount. A better indicator of how serious a company is than whether or not it has invested a large amount of money in AI is whether that investment manifests itself in changes in staffing. When readers look at their own companies and business partners, determining which jobs are decreasing and which jobs are in demand is a more accurate indicator of the actual situation than the amount of AI-related budgets.

On the other hand, it is not enough to just talk about having a surplus of people. According to Data Center Solutions, property services giant Colliers has warned that a lack of AI talent could hinder Europe's AI growth. American Express's 5% cut and Colliers' warning are separate reports that came out around the same time. However, when you line them up, you can see a single composition. Existing jobs are subject to downsizing and reorganization, and there are not enough people to design, operate, and maintain AI in the field. It has been pointed out that the supply of human resources may be the upper limit for growth in the European region, which means that simply raising funds does not guarantee growth. When readers are planning an introduction, they need to first check where they will procure the personnel, in addition to securing equipment and contracts.

Movements to change the relationship between tools and their bearers are also occurring outside of industry. Frieze featured painter David Salle inviting AI into his studio. The atelier is a place where creators complete their works with their own hands. The introduction of AI into this field can itself be seen as a sign that the division of roles between those in charge of work and the tools that support them is being reconsidered. Although the scale of these three issues is different: corporate downsizing, local human resource shortages, and individual production sites, they all have one thing in common: who is in charge of what, and to what extent AI will be in charge. Based on these facts, it seems that we are moving to a stage where the success or failure of introduction is measured not by the size of the amount invested, but by whether the results can be verified through actual use in the field, and whether the people who can carry it out can be secured. In the next section, we will look at how far this proof has progressed.

Verifications that surpass humans in diagnosis and prediction begin to appear

Microsoft presented a study showing that an AI modeled on a group of doctors outperformed doctors in diagnostic accuracy (Medical Economics). It is said that the design uses AI to reproduce the way medical treatment is carried out, in which multiple experts discuss and arrive at a conclusion. Until now, AI diagnostic support has often been talked about as a tool to assist single doctors. This time, the subject of comparison is the judgment of individual doctors, and the axis of evaluation has shifted from the convenience of assistance to the diagnostic accuracy itself. However, this result was obtained within the framework of research. Whether the same accuracy is maintained in actual outpatient and emergency settings will need to wait for further verification.

Testing of predictions has also begun in areas close to treatment selection. BostonGene will collaborate on research using its AI to predict efficacy and toxicity of CAR-T therapy (RegMedNet). With CAR-T therapy, treatment decisions center on both identifying patients who can be expected to benefit and estimating the risk of severe side effects in advance. If the scope of prediction expands from diagnosis to treatment efficacy and toxicity, the output of AI will go beyond being a consideration for doctors and will become the basis for decisions about whether to start treatment and how to manage it. The form of joint research indicates that the project does not end with the presentation of a method, but has entered the stage of incorporating and verifying data and expertise from the clinical side.

There are also examples outside of medicine where AI is responsible for on-site observation. EdenCore demonstrated its AI-based fruit monitoring technology at FIRA USA (Fruit Growers News). Up until now, the task of continuously observing and understanding the condition of the fruit has been carried out by people visiting the fields. If AI is to take over this observation, the value of its introduction will depend on whether it can continuously collect information that can be used to make decisions. Since the technology has only been demonstrated at an exhibition, it is necessary to see how it is used to see if it will produce results in the field.

Although the fields are different, such as diagnosis, treatment prediction, and fruit observation, what they have in common is that AI has begun to take on the role of observation and prediction, which is the gateway to human judgment. In this case, the question is not about performance in research or exhibition settings, but whether the same results can be reproduced in a real environment, and whether there will be enough people to accept and use the results.

AI that is close to daily life and confirmation in real environment

An AI voice model that classifies the severity of dysarthria using real-world recordings is featured on Bioengineer.org. Dysarthria is a condition in which the intelligibility of speech is impaired. What is notable about this report is that it covers recordings from real environments, rather than audio collected under set conditions. At home or in a nursing care setting, you cannot expect a quiet room or a dedicated microphone. Whether it can be used in this situation depends on whether it can be classified based on speech that contains noise or oscillations from the speaker. Those considering introducing it need to not only look at the accuracy numbers, but also check under what recording conditions the accuracy was achieved.

On the daily side, AI is expanding into relationships and how we use our time. The Hokuriku Regional Development Bureau's page includes a price comparison of AI girlfriend apps, the free version of AI Lover, and precautions. The fact that there are articles comparing the pricing structure and the scope of the free version shows that this type of app has moved from trial use to a stage where some people are choosing it based on cost considerations. Users should consider separately what they can get started for free and what they will have to pay to continue. Meanwhile, Unite.AI reports that three high school students launched MyDaye AI. It's a tool to incorporate busy schedules into daily plans, and the creators are from a generation close to the users.

In areas related to emotions, we can also talk about cars. Yahoo! News featured Mazda's Vision Cross Compact, which carries expectations for the next generation Mazda 2. This car is said to be attracting attention as a "best friend" equipped with "empathetic AI." However, these are expectations for a car that is still in the conceptual stage, and it has not yet been confirmed in actual driving how empathy works while driving.

The four cases so far deal with different situations: medical support, love, schoolwork and schedules, and transportation. However, what they all have in common is that AI has begun to be used close to the details and emotions of people's lives. In that case, the next question is not so much about its high performance, but rather who will continue to use it in the field and who will be responsible for it.

At first glance, the amount of funding for drug discovery AI, the rate of staff reduction, comparison of diagnostic accuracy, and classification using recordings of real environments are different topics. However, what was common in each section was that the criteria for measuring the success or failure of implementation was moving away from investment amounts and expectations, and toward on-the-ground demonstrations and securing the people who would take the lead. Even if the funding is large, whether the lab will be able to carry out the experiment remains another question. Whether the accuracy achieved within the framework of research is maintained in outpatient settings or at home requires separate verification. The same situation is reflected in the fact that some companies are reducing their workforce while others are complaining of a shortage of human resources. Rather than looking at what has been invested in AI, looking at what can be verified under what conditions and whether there are enough people to handle it will provide clues for deciding whether to introduce it in the future.

Q1: When using AI in the work at hand, under what conditions was its accuracy and effectiveness confirmed, and can you explain it in light of the conditions at your own workplace? Q2: When a decision based on AI output is incorrect, how far back can we check where the responsibility lies and the records of verification? Q3: Apart from the AI-related budget, how many resources are allocated to securing personnel and conducting on-site demonstrations? Can you explain that allocation as a management decision?

Finally, three questions

- When using AI in the work at hand, under what conditions was its accuracy and effectiveness confirmed, and can you explain it in light of your own workplace conditions? Q2: When a decision based on AI output is incorrect, how far back can we check where the responsibility lies and the records of verification? Q3: Apart from the AI-related budget, how many resources are allocated to securing personnel and conducting on-site demonstrations? Can you explain that allocation as a management decision? - When a decision based on AI output is incorrect, how far back can we check where the responsibility lies and the records of verification? Q3: Apart from the AI-related budget, how many resources are allocated to securing personnel and conducting on-site demonstrations? Can you explain that allocation as a management decision? - Apart from the AI-related budget, how many resources are allocated to securing personnel and conducting on-site demonstrations, and can this allocation be explained as a management decision?