It's time to implement your AI Governance

In this latest edition of our newsletter, I'd like to share my profound interest in Sam Altman's recent discussion with Lex Fridman . Altman, advocating for a cautious and iterative approach to AI development, highlights the importance of transparency and gradual progress. This perspective is not only refreshing but essential, ensuring that society, institutions, and governance structures have the necessary time to adapt to the rapidly evolving implications of AI technology. This echoes also the Gen-X pharmacist y who started his career at the dawn of the internet era, and who initially viewed AI with skepticism. However, witnessing its rapid growth, he now see its transformative potential in healthcare and beyond, far surpassing the internet's impact.

Where do those iterations lead and when are we going to witness Artificial General Intelligence (AGI)? Top scientist warns AI could surpass human intelligence by 202 7 - decades earlier than previously predicted! Equally as exhilarating is the recent passage of the Artificial Intelligence Act by the European Parliament on March 13th, 2024 . As a European, it fills me with immense pride to see our continent leading the way in establishing a framework that safeguards fundamental rights and fosters innovation. The EU AI Act's balanced approach to managing the risks and opportunities presented by AI, including its thoughtful regulations on AI applications, resonates deeply with the principles Altman discussed. This historic act is a significant step towards responsible AI development and deployment, aligning perfectly with the need for a thoughtful, incremental approach to technology that Altman advocates.

Generative AI and Regulation

In its new white paper, CapGemini emphasizes the revolutionary impact of #GenerativeAI and #LLM in Life Sciences, with concrete use cases in drug discovery and Life Sciences, while also highlighting the need for ethical guidelines, regulatory compliance, and human oversight. The EU officially passed the AI Act last month, comprising 459 pages and 88,087 words, “celebrated” as the first of its kind in AI regulations globally. In the US, the HHS is forming a task force to establish a regulatory framework for the use of artificial intelligence in healthcare , following an executive order from President Biden. So far so good, but where to start? What is certain is that innovations should focus on the needs and experiences of patients , ensuring that services are responsive to their needs. Assessing models could be a first step, but assessing AI model is hard and AI models tell us so little. TechCrunch discusses the challenges of reviewing rapidly evolving AI technologies, like ChatGPT, due to their broad capabilities and frequent updates . Despite these challenges, the publication emphasizes the importance of attempting reviews to provide a counterbalance to industry narratives and hype.

AI's Impact on Drug Discovery: A Future Shaped by Innovation

Novartis CEO Vasant Narasimhan emphasized the limited understanding of the human body and posed questions about AI's role in drug design . He urged caution and patience, stating that the true impact of AI may take 7-10 years to determine. At the same time, we have seen a number of progress that has shed light into what the future could be:

• Scientists at Johns Hopkins University have used AI to uncover the "dark genome " associated with tumors, potentially revolutionizing cancer detection. By analyzing "junk DNA," they found genetic code pieces linked to cancer that had changed during tumor formation. This research showcases the power of AI in understanding diseases and developing new therapies.
• In Denmark, researchers have developed FreeDTS, a software for modeling biological membranes. This development provides insight into cell behavior and disease diagnostics, including Parkinson's. This tool, made publicly available by the Niels Bohr Institute and the University of Southern Denmark, promotes collaborative research in biomolecular modeling.
• A new technology developed by Delft University of Technology, known as FRET X, offers a breakthrough in disease detection by identifying individual, full-length human proteins . This method, which relies on unique protein fingerprints compared against a database, enables the identification of proteins integral to diseases like cancer and COVID-19, promising earlier diagnoses and insights into disease mechanisms.

In my talks, I frequently discuss how the launch of the first AI-designed treatments, potentially by trailblazers like Insilico Medicine and BenevolentAI , is poised to validate the Insilico drug discovery approach, sparking a research evolution across leading pharmaceutical firms. And we are only 2 to 5 years from this event:

• Insilico Medicine has been named Top Biotech Company in Fast Company’s 2024 World’s Most Innovative Companies . Insilico Medicine has had a number of significant breakthroughs. These include the advancement of a new treatment for ALS designed using its platform, and developing drugs for “undruggable” cancer targets in partnership with the University of Toronto. Also on this breakthrough list is a new AI-designed treatment for inflammatory bowel disease being brought to clinical trials , as well as nominating a potentially best-in-class preclinical candidate for ovarian and breast cancer . The advancement of INS018_055 showcases generative AI's transformative potential in drug discovery, with Insilico Medicine's platform, Pharma.AI , utilizing AI models for target identification and molecule generation. ?These advancements have proved the skeptics wrong by establishing a new benchmark for AI-driven pharmaceutical research.
• BenevolentAI announces promising Phase Ia trial results for BEN-8744, showing it's safe and well-tolerated in healthy volunteers . This marks a significant step in developing a new treatment for ulcerative colitis with no serious adverse events or CNS-related side effects reported.

?As AI accelerates scientific research, it raises the question: Is there still a need for human participants? Utilizing AI in research as an adjunct to human participants can enhance efficiency while requiring ethical controls, global norms, and collaborative efforts to maintain research quality. Andrew Hunt, who studies deep learning and robotics at Carnegie Mellon University , notes that AI can initially assist in testing research queries, but for social sciences, real human complexities are irreplaceable. AI's aggregated data may not capture these nuances, underscoring its role as a supportive tool rather than a replacement in understanding human intricacies.

Some recent Partnerships aiming to embrace AI in an effort to revolutionize drug development:

• ???Sanofi and Owkin have announced the expansion of their collaboration in leveraging artificial intelligence (AI) to include drug positioning in immunology .
• ??? 拜耳 partners with German AI company Aignostics to enhance precision oncology drug research and development , using AI to connect pathology and clinical data for improved patient outcomes and faster clinical development of oncology medicines.

Nvidia's Healthcare Revolution: AI Transformations from Diabetic Retinopathy Exams to Drug Discovery Collaborations

Nvidia's strategic pivot to healthcare, exemplified by AI eye exams for diabetic retinopathy and significant partnerships like those with Johnson & Johnson, GE Healthcare and Microsoft , underscores its ambition in medical AI. This ambition extends to collaborations with companies like Recursion Pharmaceuticals , leveraging Nvidia's computing prowess for drug discovery, indicating Nvidia's holistic approach to revolutionizing healthcare diagnostics, treatment, and research. See more on Nvidia Tool launches at the 2024 GTC AI conference.See more on Nvidia Tool launches at the 2024 GTC AI conference.

AI Revolutionizes Clinical Trials: Enhancing Designs, Accelerating New Treatments

As described in Nature, AI has a transformative potential in clinical trials . By optimizing study designs, enhancing patient recruitment, and improving outcome analysis, AI could substantially reduce development times and costs. This advancement addresses the growing challenges of drug development, demonstrating AI’s role in increasing probability of success and duration of clinical trials, and accelerating the delivery of new treatments. For example, eClinical is a platform that integrates AI to enhance data review, anomaly detection, and risk-based quality management , mitigating risks as trials progress.

An emerging trend is improving clinical trial design through simulation. The Critical Path Institute has launched a Clinical Trial Simulator for Duchenne muscular dystrophy research , which optimizes trial design and accelerates the development of effective therapies.

As stated in a previous edition, "Data is the food of AI ". Thus, it's not surprising to see different types of data used more frequently in clinical trials:

• EHR: A study shows that using Electronic Health Records in clinical trials is beneficial for various purposes , like improving patient recruitment and data collection efficiency. Future research should focus on enhancing EHR interoperability and standardizing methodologies for broader application in clinical research.
• Real World Data: Curated Real-World Data (RWD) from clinical data registries can streamline patient recruitment and optimize clinical trial designs, especially for rare diseases. An ocular drug trial case study shows how the American Academy of Ophthalmology's IRIS Registry was used to efficiently identify and enroll suitable patients . The study demonstrated RWD's crucial role in advancing clinical research and enhancing collaboration between trial sponsors and medical practices for improved outcomes.
• Wearable technology is transforming data collection in clinical trials, allowing for continuous, real-time monitoring of patients , and enabling proactive health management. However, integrating wearables into clinical research presents challenges. This includes, data privacy issues, the need for sophisticated data analysis methodologies, participant compliance, and ensuring device accuracy and reliability.

AI Revolutionizes Medical Diagnostics with Breakthrough Technologies

Advancements in AI are transforming diagnostics across various medical fields, offering novel solutions such as:

• ?Unlocking Early Detection of Lung Cancer With MIT’s Inhalable Nanosensors : MIT researchers have developed a new diagnostic tool for lung cancer that involves inhaling nanoparticle sensors and taking a urine test.
• AI eye exams for diabetic retinopathy are emerging as a successful application of AI diagnostics in clinical settings . Offering significant benefits like immediate results and cost savings, this highlights the future direction for AI in healthcare.
• Scientists to study real-world eating behaviors using wearable sensors and artificial intelligence

• Carnegie Mellon University: Machine learning, quantum computing can transform healthcare, including diagnosing pneumonia
• Johns Hopkins: AI can now detect COVID-19 in lung ultrasound images
• Link?ping University: AI-based app can help physicians find skin melanoma
• Mass General Brigham: Researchers develop AI foundation models to advance pathology
• Meharry Medical College/IADR: ChatGPT to help dental students retain knowledge and enhance performance
• Rutgers: Predicting loneliness through online digital footprints
• Developed by Oxford University, the new AI tool DrugGPT aims to minimize the 237 million medication errors annually in England by providing clinicians with a reliable prescribing safety net and informative guidance
• Recent studies reveal ChatGPT 4's enhanced accuracy in healthcare, successfully identified drug interactions in 100% of cases , marking a significant improvement over previous versions and highlights its potential in the medical field.
• The AI tool Foresight, using NHS electronic records, shows promise in predicting future health conditions for patients . Developed by researchers from King’s College London and University College London alongside NHS Trusts, this tool provides the potential to aid doctors in diagnosis and patient monitoring,

Massive Cyber Attack on Change Healthcare Exposes Critical Cybersecurity Vulnerabilities in U.S. Healthcare Industry

Healthcare data breaches have emerged as a significant challenge during the ongoing digital transformation of the healthcare industry. The recent cyber attack on Change Healthcare by BlackCat/ALPHV, has highlighted critical vulnerabilities within the U.S. healthcare industry, disrupting services for numerous healthcare organizations by cutting off access to essential patient claims submission and payment systems. With a $22 million ransom reportedly paid to restore services , this incident underscores the critical need for robust cybersecurity measures in protecting patient data and healthcare infrastructure.

Companies are grappling with the challenges of managing sensitive data amidst rising breaches. This creates a demand for synthetic data as a privacy-compliant solution for AI training and analytics across various industries, promising to enhance data privacy and mitigate biases in AI applications. However, a study on synthetic data generation for health data privacy emphasizes the importance of ensuring that synthetic datasets can reliably replicate real data analysis results and support valid population inferences . The finding revealed that sequential synthesis with boosted decision trees performs better than GANs in terms of replicability metrics and privacy risk.

Generative AI in Data Engineering

Generative AI, empowered by advancements in large language models like GPT, is revolutionizing the data engineering lifecycle by enhancing every phase from data generation to serving. This is transforming the traditional challenges of data management into innovative opportunities. It offers smart solutions for creating synthetic datasets, improving data accuracy, optimizing storage efficiency, and streamlining data transformation and serving. This marks a significant shift towards a more efficient, data-driven future.

Integrating Generative AI into data engineering workflows can significantly enhance data management and analysis. It allows for summarizing vast amounts of structured and unstructured data, thereby expanding data accessibility for teams. However, its application requires careful consideration to ensure it drives meaningful growth within organizations. This article explores practical uses of Generative AI in data pipelines, including feature engineering with unstructured data, utilizing new data sources, webscraping for external data collection, and optimizing business processes. The article also emphasizes the importance of a strategic approach and collaboration between business stakeholders and data teams.

Upcoming 2024 Conferences

• Bio-ITWorld , Boston, USA, April 15th, 17th 2024
• Boom Summit by Day One , Basel, April 10th and 11th
• Swiss Biotech Day 2024 , , at the Congress Center Basel in Switzerland, April 22-23, 2024
• Ai-Med Global Summit 2024: Florida, USA, May 29 - 31, 2024
• BiotechX USA 2024 , Philadelphia, September 10 - 11, 2024
• Intelligent Health AI , Basel, September 11-12, 2024
• BiotechX Europe 2024 , Basel, Switzerland, October 9-10, 2024
• Med Tech World, MALTA ,November 6th, 8th 2024

Other Great Newsletters on Healthcare and AI

Stay up-to-date with the latest insights and industry trends by subscribing to these informative newsletters from my esteemed colleagues and professionals in the field.

• Healthcare in the Metaverse by Philippe GERWILL
• DigitalhealthReadingList by Wolfgang Schleifer
• Future Tech Trends by Bernard Marr
• All about Blockchain Adoption by Kamlesh Nagware
• AI Frontier by Steve Nouri
• The Atlas by Generative AI

?