A study published on June 23 in collaboration with UT and Yale researchers developed an artificial intelligence tool capable of performing and analyzing the heart using echocardiography. 

The app, PanEcho, can analyze echocardiograms, or pictures of the heart, using ultrasounds. The tool was developed and trained on nearly one million echocardiographic videos. It can perform 39 echocardiographic tasks and accurately detect conditions such as systolic dysfunction and severe aortic stenosis.

“Our teammates helped identify a total of 39 key measurements and labels that are part of a complete echocardiographic report — basically what a cardiologist would be expected to report on when they’re interpreting an exam,” said Gregory Holste, an author of the study and a doctoral candidate in the Department of Electrical and Computer Engineering. “We train the model to predict those 39 labels. Once that model is trained, you need to evaluate how it performs across those 39 tasks, and we do that through this robust multi site validation.” 

Holste said out of the functions PanEcho has, one of the most impressive is its ability to measure left ventricular ejection fraction, or the proportion of blood the left ventricle of the heart pumps out, far more accurately than human experts. Additionally, Holste said PanEcho can analyze the heart as a whole, while humans are limited to looking at the heart from one view at a time. 

“What is most unique about PanEcho is that it can do this by synthesizing information across all available views, not just curated single ones,” Holste said. “PanEcho integrates information from the entire exam — from multiple views of the heart to make a more informed, holistic decision about measurements like ejection fraction.” 

PanEcho is available for open-source use to allow researchers to use and experiment with the tool for future studies. Holste said the team has already received emails from people trying to “fine-tune” the application for different uses. 

“We know that other researchers are working on adapting PanEcho to work on pediatric scans, and this is not something that PanEcho was trained to do out of the box,” Holste said. “But, because it has seen so much data, it can fine-tune and adapt to that domain very quickly. (There are) very exciting possibilities for future research.”

Google announced two new artificial intelligence initiatives on July 7, 2025, designed to support mental health organizations in scaling evidence-based interventions and advancing research into anxiety, depression, and psychosis treatments.

The first initiative involves a comprehensive field guide developed in partnership with Grand Challenges Canada and McKinsey Health Institute. According to the announcement from Dr. Megan Jones Bell, Clinical Director for Consumer and Mental Health at Google, “This guide offers foundational concepts, use cases and considerations for using AI responsibly in mental health treatment, including for enhancing clinician training, personalizing support, streamlining workflows and improving data collection.”

The field guide addresses the global shortage of mental health providers, particularly in low- and middle-income countries. According to analysis from the McKinsey Health Institute cited in the document, “closing this gap could result in more years of life for people around the world, as well as significant economic gains.”

Summary

Who: Google for Health, Google DeepMind, Grand Challenges Canada, McKinsey Health Institute, and Wellcome Trust, targeting mental health organizations and task-sharing programs globally.

What: Two AI initiatives including a practical field guide for scaling mental health interventions and a multi-year research investment for developing new treatments for anxiety, depression, and psychosis.

When: Announced July 7, 2025, with ongoing development and research partnerships extending multiple years.

Where: Global implementation with focus on low- and middle-income countries where mental health provider shortages are most acute.

Why: Address the global shortage of mental health providers and democratize access to quality, evidence-based mental health support through AI-powered scaling solutions and advanced research.

The 73-page guide outlines nine specific AI use cases for mental health task-sharing programs, including applicant screening tools, adaptive training interfaces, real-time guidance companions, and provider-client matching systems. These tools aim to address challenges such as supervisor shortages, inconsistent feedback, and protocol drift that limit the effectiveness of current mental health programs.

Task-sharing models allow trained non-mental health professionals to deliver evidence-based mental health services, expanding access in underserved communities. The guide demonstrates how AI can standardize training, reduce administrative burdens, and maintain quality while scaling these programs.

According to the field guide documentation, “By standardizing training and avoiding the need for a human to be involved at every phase of the process, AI can help mental health task-sharing programs effectively scale evidence-based interventions throughout communities, maintaining a high standard of psychological support.”

The second initiative represents a multi-year investment from Google for Health and Google DeepMind in partnership with Wellcome Trust. The funding, which includes research grant funding from the Wellcome Trust, will support research projects developing more precise, objective, and personalized measurement methods for anxiety, depression, and psychosis conditions.

The research partnership aims to explore new therapeutic interventions, potentially including novel medications. This represents an expansion beyond current AI applications into fundamental research for mental health treatment development.

The field guide acknowledges that “the application of AI in task-sharing models is new and only a few pilots have been conducted.” Many of the outlined use cases remain theoretical and require real-world validation across different cultural contexts and healthcare systems.

For the marketing community, these developments signal growing regulatory attention to AI applications in healthcare advertising. Recent California guidance on AI healthcare supervision and Google’s new certification requirements for pharmaceutical advertising demonstrate increased scrutiny of AI-powered health technologies.

The field guide emphasizes the importance of regulatory compliance for AI mental health tools. Several proposed use cases, including triage facilitators and provider-client matching systems, could face classification as medical devices requiring regulatory oversight from authorities like the FDA or EU Medical Device Regulation.

Organizations considering these AI tools must evaluate technical infrastructure requirements, including cloud versus edge computing approaches, data privacy compliance, and integration with existing healthcare systems. The guide recommends starting with pilot programs and establishing governance committees before full-scale implementation.

Technical implementation challenges include model selection between proprietary and open-source systems, data preparation costs ranging from $10,000 to $90,000, and ongoing maintenance expenses of 10 to 30 percent of initial development costs annually.

The initiatives build on growing evidence that task-sharing approaches can improve clinical outcomes while reducing costs. Research cited in the guide shows that mental health task-sharing programs are cost-effective and can increase the number of people treated while reducing mental health symptoms, particularly in low-resource settings.

Real-world implementations highlighted in the guide include The Trevor Project’s AI-powered crisis counselor training bot, which trained more than 1,000 crisis counselors in approximately one year, and Partnership to End Addiction’s embedded AI simulations for peer coach training.

These organizations report improved training efficiency and enhanced quality of coach conversations through AI implementation, suggesting practical benefits for established mental health programs.

The field guide warns that successful AI adoption requires comprehensive planning across technical, ethical, governance, and sustainability dimensions. Organizations must establish clear policies for responsible AI use, conduct risk assessments, and maintain human oversight throughout implementation.

According to the World Health Organization principles referenced in the guide, responsible AI in healthcare must protect autonomy, promote human well-being, ensure transparency, foster responsibility and accountability, ensure inclusiveness, and promote responsive and sustainable development.

Timeline

• July 7, 2025: Google announces two AI initiatives for mental health research and treatment
• January 2025: California issues guidance requiring physician supervision of healthcare AI systems
• May 2024: FDA reports 981 AI and machine learning software devices authorized for medical use
• Development ongoing: Field guide created through 10+ discovery interviews, expert summit with 20+ specialists, 5+ real-life case studies, and review of 100+ peer-reviewed articles

A new study shows that the language used to prompt AI chatbots can steer them toward different cultural mindsets, even when the question stays the same. Researchers at MIT and Tongji University found that large language models like OpenAI’s GPT and China’s ERNIE change their tone and reasoning depending on whether they’re responding in English or Chinese.

The results indicate that these systems translate language while also reflecting cultural patterns. These patterns appear in how the models provide advice, interpret logic, and handle questions related to social behavior.

Same Question, Different Outlook

The team tested both GPT and ERNIE by running identical tasks in English and Chinese. Across dozens of prompts, they found that when GPT answered in Chinese, it leaned more toward community-driven values and context-based reasoning. In English, its responses tilted toward individualism and sharper logic.

Take social orientation, for instance. In Chinese, GPT was more likely to favor group loyalty and shared goals. In English, it shifted toward personal independence and self-expression. These patterns matched well-documented cultural divides between East and West.

When it came to reasoning, the shift continued. The Chinese version of GPT gave answers that accounted for context, uncertainty, and change over time. It also offered more flexible interpretations, often responding with ranges or multiple options instead of just one answer. In contrast, the English version stuck to direct logic and clearly defined outcomes.

No Nudging Needed

What’s striking is that these shifts occurred without any cultural instructions. The researchers didn’t tell the models to act more “Western” or “Eastern.” They simply changed the input language. That alone was enough to flip the models’ behavior, almost like switching glasses and seeing the world in a new shade.

To check how strong this effect was, the researchers repeated each task more than 100 times. They tweaked prompt formats, varied the examples, and even changed gender pronouns. No matter what they adjusted, the cultural patterns held steady.

Real-World Impact

The study didn’t stop at lab tests. In a separate exercise, GPT was asked to choose between two ad slogans, one that stressed personal benefit, another that highlighted family values. When the prompt came in Chinese, GPT picked the group-centered slogan most of the time. In English, it leaned toward the one focused on the individual.

This might sound small, but it shows how language choice can guide the model’s output in ways that ripple into marketing, decision-making, and even education. People using AI tools in one language may get very different advice than someone asking the same question in another.

Can You Steer It?

The researchers also tested a workaround. They added cultural prompts, telling GPT to imagine itself as a person raised in a specific country. That small nudge helped the model shift its tone, even in English, suggesting that cultural context can be dialed up or down depending on how the prompt is framed.

Why It Matters

The findings concern how language affects the way AI models present information. Differences in response patterns suggest that the input language influences how content is structured and interpreted. As AI tools become more integrated into routine tasks and decision-making processes, language-based variations in output may influence user choices over time.

Notes: This post was edited/created using GenAI tools. Image: DIW-Aigen.

Read next: Jack Dorsey Builds Offline Messaging App That Uses Bluetooth Instead of the Internet