OpenAI, Baidu, Baichuan Intelligence and Others Accelerate AI Integration in Healthcare: Filing IPO Prospectus

Led by OpenAI, large language model companies are making all-out efforts to expand into the healthcare market.

Not long ago, Sam Altman, the driving force behind OpenAI and its founder and CEO, established a medical AI company named Thrive AI Health (hereinafter referred to as “Thrive”), aiming to transform the behavioral habits of patients with chronic diseases through AI coaching. As expected, it will leverage OpenAI’s generative AI technology.

VCBeat has found, based on public information, that since 2024, OpenAI has been going all out in the healthcare market. Not only OpenAI, but domestic large model companies are also rapidly following suit. According to media statistics,Among 120 global financing deals in the large language model (LLM) sector in recent years, “LLM + Healthcare” accounted for the largest share, with up to 15% of LLM-related financings involving healthcare companies.

The medical sector is becoming the “crown jewel of large language models.”

According to reports, Thrive’s executive team boasts an all-star lineup. In addition to Sam Altman himself, co-founder Arianna Huffington is also the founder of The Huffington Post and the behavior-change technology company Thrive Global. Furthermore, former Google executive DeCarlos Love will join the new company as CEO.

The startup in preparation will also establish deep collaborative partnerships with several renowned academic institutions and medical centers, including the Stanford Medicine Institute, the Rockefeller Neuroscience Institute at West Virginia University, and the Walton College of Medicine.

It is aptly described as a “second-generation rich” entity born with a silver spoon in its mouth.

Thrive will develop an “AI Health Coach” platform, aiming to leverage generative AI technology to provide patients with chronic diseases with professional guidance on key health issues in daily life, such as sleep, nutrition, fitness, stress management, and social interaction, thereby reducing the global incidence of chronic diseases through changes in user behavior.

Taking the United States as an example, 129 million people suffered from at least one chronic disease in 2023, such as cardiovascular disease, depression, or diabetes. These chronic conditions can actually be improved through changes in daily behaviors.

According to foreign media reports, Thrive’s AI health coach will leverage Thrive Global’s accumulated expertise in healthcare and incorporate OpenAI’s latest advancements in artificial intelligence, including enhanced long-term memory capabilities and behavior-guidance models customized for specific domains. This generative AI will be trained on the latest peer-reviewed scientific literature, biostatistics, laboratory results, and other medical data, integrated with users’ personal preference data.

However, the two founders stated that Thrive will not venture into AI-assisted medical diagnosis but will instead focus on providing health advice. This may also be an effort to avoid sensitive issues related to medical data privacy and regulatory compliance.

Meanwhile, the specific form in which Thrive’s products will be implemented remains undetermined; they may be launched as standalone applications, offered through various models, or even integrated into workplace scenarios via platforms such as Microsoft Teams.

This is not the first time OpenAI has ventured into healthcare. A brief review of recent developments reveals that since the beginning of 2024, OpenAI has been aggressively expanding its presence in the medical sector.

The biggest news is none other than theIn April this year, renowned pharmaceutical company Moderna announced a partnership with OpenAI to deploy ChatGPT Enterprise across the entire organization, setting a goal for all employees with access to digital solutions to achieve full proficiency in generative AI within six months.。

In fact, as early as early 2023, Moderna built an internal AI chatbot, mChat, based on OpenAI’s API. It was adopted by 80% of employees and received positive feedback. After comprehensive evaluation, Moderna ultimately decided to implement ChatGPT Enterprise. Within just two months of its deployment, 40% of active users had created personalized chatbots, bringing the total number to 750, with each user averaging 120 conversations per week.

Currently, with the support of OpenAI, Moderna has introduced generative AI into many daily scenarios within the company.

For instance, Moderna has deployed generative AI as a data analysis assistant for its clinical research teams to evaluate and analyze the potential of clinical data, integrating and visualizing these large datasets to enhance the team’s clinical judgment and decision-making. For example, AI can facilitate detailed review processes, helping to prioritize safety and optimize scenarios such as vaccine development before advancing to later-stage clinical trials.

The regulatory affairs team is also highly enthusiastic about generative AI. With the assistance of generative AI, regulatory professionals can quickly obtain clear and readable contract summaries. Furthermore, generative AI enables employees to rapidly access internal policies, eliminating the need to sift through hundreds of documents as in the past, thereby significantly enhancing work efficiency.

Moderna’s PR team is also leveraging ChatGPT Enterprise. They have built a generative AI chatbot to assist in creating PowerPoint presentations for quarterly earnings conference calls, while another chatbot helps translate complex biotechnology terminology into accessible language for investor communications.

In addition to Moderna, a growing number of healthcare companies are partnering with OpenAI and integrating generative AI into numerous scenarios. Among them,AI Health Assistants Have Become One of the Most Proficient Areas for Generative AI in Healthcare EngagementIn addition to Thrive mentioned earlier, both Healthify and Whoop have partnered with OpenAI to build AI health assistants leveraging OpenAI’s generative AI capabilities. These assistants analyze user data and support human coaches in providing comprehensive guidance on various aspects, including diet and behavioral habits.

Meanwhile,Leveraging the significant advantages of generative AI in natural language processing, its application as an internal knowledge base or for assisting with documentation is currently one of the popular trends.。

For instance, in 2024, Oscar Health became the first healthcare company to sign a commercial partnership agreement with OpenAI. Leveraging OpenAI’s technology, Oscar Health developed an AI assistant that effectively tracks reimbursements and automatically answers patient inquiries regarding reimbursement claims. This innovation not only reduced the time required to process reimbursements by 50%, but also ensured accuracy comparable to or better than that of human agents.

Lifespa, a healthcare institution, is using generative AI to convert surgical consent forms from dense medical-legal jargon into text that is easier for all patients to read, thereby enabling more patients to understand the content of these forms. Meanwhile, Summer Health, a pediatric healthcare provider, leverages generative AI to assist physicians in organizing visit documentation, significantly reducing the average time doctors spend on paperwork per visit from 10 minutes to just 2 minutes.

In the burgeoning field of “AI + Clinical Practice,” the advantages of generative AI are also highly evident.。

Currently, the integration of AI into clinical practice has become a hot investment topic abroad. In 2023 alone, there were nearly 200 AI-designed clinical trials in the United States. Generative AI can read and analyze vast amounts of disease phenotype and genetic information to facilitate clinical indication selection, optimization of clinical endpoints, and identification of eligible subjects. Additionally, generative AI can analyze electronic medical records, thereby helping pharmaceutical companies identify hospitals with a higher concentration of patients meeting enrollment criteria, while also serving a role in clinical monitoring.

Historically, the primary challenge in clinical trial enrollment has been identifying and matching suitable patients. Manually reviewing vast volumes of patient medical records is impractical, resulting in most clinical trials being conducted by patients located near trial sites, which may introduce selection bias.

Paradigm leverages OpenAI’s GPT-4 model to extract and analyze medical case data, thereby identifying the most suitable patients for clinical trials. Practice has demonstrated that generative AI platforms can evaluate hundreds of patients per minute, a significant efficiency improvement over the manual process, which is limited to 50 patients per day. Moreover, accuracy has increased by 10% compared to manual screening, highlighting the broad prospects for the integration of generative AI into clinical practice.

Meanwhile,OpenAI’s generative AI has also demonstrated significant advantages in assisting physicians in delivering more personalized treatment plans for patients.。

For instance, in June, Color Health, which has long partnered with the American Cancer Society to help health plans and employers manage cancer care, announced that it had selected OpenAI as its AI solutions provider. The company plans to integrate medical data with clinical knowledge using GPT-4o to create a generative AI tool named “Copilot,” designed to develop personalized, comprehensive treatment plans for cancer patients.

Color Health’s Copilot has demonstrated its advantages in clinical trials. According to reports, Copilot can help healthcare institutions identify cases with missing laboratory, imaging, or biopsy and pathology results at a rate more than four times higher than before; furthermore, it significantly reduces the time clinicians spend analyzing patient records and identifying gaps from weeks to an average of just five minutes.

China is moving quickly to pioneer generative AI applications in healthcare. Take Baidu, which directly rivals OpenAI, as an example,In September 2023, Baidu Health released the first industry-level large medical model—the Lingyi Large Model.After nearly a year of development, the model has established a three-tier technical architecture comprising Mixture of Experts (MoE), terminal components, and intelligent agents. It has been widely applied in products related to science popularization content, internet hospitals, smart hospitals, intelligent diagnosis and treatment, and open platforms.

At the Baidu Health Industry Ecosystem Conference held recently,Baidu Health has launched four medical large language model applications and one open platform, addressing every stage of the patient care journey and meeting the needs of patients, healthcare providers, and pharmaceutical companies., significantly advancing the implementation of this innovative technology.

These four large medical AI model applications are: the vertical large-model application “AI Health Assistant,” which provides authoritative and convenient health education information and medical consultation services to general users; the “Online Medical Copilot,” which offers quality control, decision support, and other assistance at every stage of internet-based healthcare, thereby helping physicians comprehensively improve efficiency in their practice within online hospitals; the “AI Smart Outpatient Clinic,” designed for hospital-side use to address issues such as patients registering with the wrong department, inaccurate appointment scheduling, and strained medical resources; and the “CDSS+LLM” solution, which primarily targets common, time-consuming clinical scenarios for physicians, including case documentation generation, patient briefings, and intelligent question-answering.

Furthermore, Baidu Health has launched the Lingyi Open Platform, announcing that it will provide partners with medical service APIs and a free quota of 10 million tokens, as well as offer free co-development support for 20 standardized scenarios. Together, they will explore smart hospital services, improved enterprise operational efficiency, post-diagnosis patient management, high-quality science popularization content creation, and the upgrading of internet-based diagnosis and treatment, ultimately promoting the widespread adoption of AI and reshaping a smart-enabled healthcare service system.

Overall, the pioneering efforts by major generative AI players such as OpenAI and Baidu in healthcare scenarios are highly representative and will provide practical reference for other large model enterprises seeking to enter the medical field.

Besides tech giants, unicorns in the generative AI sector are also racing ahead on the path of “AI + Healthcare.” Just asRecently, BAICHUAN AI, a well-known domestic unicorn in the large model sector, announced the completion of its Series A financing round, raising up to RMB 5 billion. This may become the largest financing deal in China’s AI industry in 2024.。

Considering that BAICHUAN AI was founded only about a year and a half ago, this rapid pace of development is the best testament to the growing importance attached to generative AI.

Notably,BAICHUAN AI is a leading Chinese large language model company that has long focused on the healthcare sector, regarding it as its core mission.。

Wang Xiaochuan, the founder who also co-founded Sogou, chose to pursue interdisciplinary research in biology while studying in the Department of Computer Science and Technology at Tsinghua University, demonstrating a strong passion for life sciences. Since 2016, Wang Xiaochuan and his companies have successively invested in multiple health-tech enterprises, including Airdoc, Xiaolu Traditional Chinese Medicine Clinic, Baike Mingyi, Yuyi Ganlan, and Chunyu Doctors.

When Sogou was merged into Tencent in 2021, Wang Xiaochuan further stated in an open letter, “In the next 20 years, if I can contribute to the advancement of life sciences, health, and medicine, and make a modest contribution to public health, my life will be more meaningful.” Subsequently,Wang Xiaochuan founded BAICHUAN AI in 2023, firmly asserting that “healthcare is the crown jewel of large language models.”。

He believes that large models’ mastery of knowledge and experience, along with their multimodal capabilities, memory, and reasoning abilities—as well as their capacity to reduce hallucinations and demonstrate empathetic communication—can all be applied in the medical field. The better the model’s capabilities, the higher the level of medical practice, indicating an extremely high upper limit for this process.

Precisely because of this,Since its inception, BAICHUAN AI has embedded “health” into its corporate vision, thereby becoming the only large-model unicorn in China focused on the healthcare sector.。

This approach to building AI doctors using large language models aligns closely with the views of Geoffrey Hinton, one of the widely recognized “Godfathers of AI.” Hinton has long maintained that healthcare is a domain where AI can make significant contributions and generate substantial societal benefit. He believes that AI already rivals top experts in interpreting many types of medical images and excels at integrating vast amounts of patient data, signaling that the era of AI physicians is dawning.

A representative from BAICHUAN AI stated in an interview with VCBeat that healthcare is the industry with the highest data density, bar none. Clinical guidelines are a compression of medical records, and disease diagnosis is a prediction based on these guidelines.

“The fundamental principles of medical research and clinical services actually share the same paradigm as large language models,” he explained to VCBeat. “In physician training, internists broaden their knowledge through extensive exposure, while surgeons achieve mastery through repetitive practice—approaches that are highly consistent with the training methodologies for large language models and embodied AI.”

The executive further stated that BAICHUAN AI’s goal is to create “AI doctors.” This will trigger a revolution on the supply side of healthcare: “The pain point in the medical service industry is ‘unlimited demand versus limited supply.’ Internet-based healthcare has offered limited improvements in this regard, but large medical models are different. They can increase supply by hundreds or even thousands of times by developing AI doctors/assistants, thereby making medical services higher in quality, more accessible, and lower in cost.”

He depicted to usBAICHUAN AI’s Vision for the Implementation Path of AI in Healthcare: Analogizing AI Healthcare to Autonomous Driving Levels and Proposing an L0–L5 Classification。

From the perspective of BAICHUAN AI, current large model technology is already capable of supporting Level 3 (LLM) AI medical applications, wherein AI can assist in recommending treatment plans under specific conditions, subject to physician confirmation. In the near future, BAICHUAN AI will also take the lead in deploying its solutions in healthcare scenarios such as health consultations and assisted diagnosis and treatment, initially in the form of health advisors and physician assistants.

In the next phase, BAICHUAN AI will strive to achieve Level 4 (AGI). At this stage, the idealized AI physician will become a reality. Level 5 represents the ultimate large life model, capable of autonomously discovering new therapies based on real-world research and achieving comprehensive disease course management.

Furthermore, in the process of developing large medical models, BAICHUAN AI has summarized the essential qualities required for such models, namely “safeguarding the baseline and raising the ceiling.”

The so-called “safeguarding the baseline” refers to prioritizing medical risks and ensuring medical safety.

“The primary priority is not to improve accuracy, but to ensure that errors do not pose severe risks to life and health—this is the difference between 0 and 1. Only those who can better safeguard medical risk safety are qualified to stay in the game.” “We aim to build the Baichuan Medical Large Model into a sufficiently safe and controllable model that leads in China and even globally.”

Regarding “raising the ceiling,” BAICHUAN AI believes that there are many areas it aims to break through in developing its large medical language models, such as human-like emotional interaction models, long-term memory capabilities, robust multimodal abilities, and strong reasoning and planning skills.

“Healthcare is the crown jewel of large language models” is not an empty phrase; it takes a team of visionary scientists, billions of dollars or more in investment, and years of effort to have a chance at doing it right. “If healthcare is treated as just one among many industries, with only simple fine-tuning of an application, I believe such an approach is unlikely to succeed.”

According to reports, BAICHUAN AI has currently engaged senior chief physicians and attending physicians from top-tier tertiary hospitals, including Peking Union Medical College Hospital and Peking University Health Science Center-affiliated hospitals, as well as psychology experts, to conduct multi-dimensional human evaluations of its large language models. In the future, BAICHUAN AI will collaborate with more distinguished physicians and medical institutions to continuously enhance the capabilities of its models.

It is not difficult to observe that, both domestically and internationally, and across tech giants and unicorns alike, the integration of generative AI into healthcare is deepening and evolving rapidly.

With the integration of generative AI, healthcare business processes are transitioning from digitalization to intelligent digitalization, marking an essential first step for generative AI’s incorporation into traditional enterprise workflows. As large medical models become increasingly capable and able to perform independent tasks under human guidance, service supply will expand significantly, thereby enabling greater inclusivity within the industry.

VCBeat will continue to closely monitor the impact of generative AI on healthcare and provide relevant coverage. We believe that despite numerous challenges, the ever-evolving generative AI will ultimately bring benefits to all stakeholders in the healthcare ecosystem.

References:

Alyssa, Panken, Zhidx: “Godfather Hinton’s 46-Minute Interview: AI Can Replicate the Human Mind, Potentially Exacerbating the Global Wealth Gap”

Heather Landi，fiercehealthcare.com：OpenAI Startup Fund, Arianna Huffington back new AI health coach venture focused on chronic conditions

Li Shuiqing, Zhidx: “Surge of Hot Money into Large Models: Over 100 Billion-Yuan Financing Rounds in Half a Year, with the Highest Exceeding 50 Billion Yuan”