Cognizant knows that AI is a story of many vendors and models. In the critical life sciences environment, the company is working on the implementation of both classical and generative artificial intelligence. What works for the pharmaceutical industry can immediately provide useful guidance for other organizations looking for a way to future-proof AI.

Dr. Pierre Marchand started his career in AI before it was cool. “I have worked in data analysis and data science my entire life, including in the banking world, manufacturing and the chemical industry,” he says. “When you’ve been in business for thirty years, you start to see the big picture.” Six months ago, the expert took the leap to Cognizant, where he now works as Chief Data Strategist.

Clear vision in a crucial sector

This step was no coincidence. With the advent of generative AI, Marchand wanted to be back at the center of the action. “Cognizant has already talked about Gen AI for competition,” he says. “The company is taking a leadership role and pursuing the right plan of action.”

We speak to Cognizant in the context of AI in the life sciences and pharmaceutical industries. Business organizations operate in this sector where business operations are based on a scientific approach and competition is fierce. On the one hand, AI has enormous potential to accelerate development and production in the pharmaceutical industry, but on the other hand, trust in AI and the protection of sensitive data is almost nowhere more important. What’s good enough for life sciences will almost certainly be good enough for other industries, so we’re excited to see what role AI plays there today.

Classic or generative?

Marchand emphasizes that AI in life sciences and pharmaceuticals is nothing new. “Classic AI is already being adopted by all major players in the industry in many processes.” Traditional or classic AI uses clearer rules created by experts. Algorithms are already optimizing all kinds of tasks and have the great advantage that they are not mysterious. The results delivered are the result of understood programming.

Today’s revolution and hype is about generative AI. It differs from classic AI in that the intelligence does not come from clear algorithms, but from a neural network that is trained on the basis of large amounts of data. During training, the network evolves into an intelligent and useful AI model, but what happens under the hood is less transparent. Such neural networks have the potential to revolutionize processes, but have significant limitations.

Data for the black box

“To train a generative AI model, you need enormous amounts of qualitative data,” says Marchand. “These are not always available. For example, there are limited data available on some rare diseases, causing predictions to lose accuracy. When data is scarce or imbalanced, the reliability of generative AI fails.”

If the data is sparse or unbalanced, the reliability of generative AI fails.

Dr. Pierre Marchand, Chief Data Strategist at Cognizant

This reliability is another tricky point. Marchand points this out flight recorder-Dilemma. The neural networks behind generative AI work, but it is often not clear why certain inputs lead to certain outputs. “This lack of transparency hinders the acceptance of AI-generated results,” he knows. “Especially in the scientific world, where decisions impact people’s lives.”

From suspicion to trust

“Without integrity there is no trust,” says the AI ​​guru. “Nobody wants to hire a corrupt accountant. Integrity comes from traceability. As a human, you want to be able to figure out where an answer comes from. Especially for AI, it is helpful to be able to check why a model gives a certain answer. That creates trust.”

“In most cases, we see that people fully check the answers once or twice and then trust slowly builds up. No one wants to risk their job by simply accepting an AI prediction, but when you get the chance to test yourself and research, your confidence grows.”

First test, then use

In the pharmaceutical sector, it is not possible to simply accept a suggestion, no matter how good it is. “When generative AI suggests a new molecule for a particular treatment, it is almost like suggesting a new musical note that has never been heard before,” explains Marchand. “The proposal is exciting, but still needs to be thoroughly tested to ensure both safety and effectiveness. The path from AI insights to actionable applications is paved with extensive validation, slowing the speed of implementation.”

The path from AI insights to actionable applications is paved with comprehensive validation.

Dr. Pierre Marchand, Chief Data Strategist at Cognizant

But generative AI is already playing a role in the industry today. “Atomwise uses generative AI to predict molecular activity,” says Marchand. “The company uses deep learning models to potentially discover drugs. Insilico Medicine does something similar. Tempus uses generative AI to suggest tailored treatments based on patient data. Using deep learning models, deep genomics attempts to predict how genetic mutations can lead to disease.

The possibilities are practically limitless. In the long term, Cognizant expects generative AI to be used across the entire pharmaceutical industry value chain. The technology can support experts in the discovery of new active molecules and other research and development, the design and execution of clinical trials, but also more traditional areas such as content creation and the sale and marketing of finished drugs. The ultimate goal remains clear: to achieve ever better results for patients.

Leaflets with fruit juice

Cognizant is working with these and other companies to integrate AI in a structurally sound way. This requires a well-thought-out approach because the field is evolving quickly. Marchand: “To keep up, you have to eat academic papers for breakfast. Every week there is a tsunami of progress. I have never experienced such rapid development in my entire career. It’s almost too much, but if you want to keep up, you have to suffer.”

Marchand and his colleagues take on the role of sufferers (pun intended) so they can help customers. In practice, he sees that most companies start their generative AI journey with what he Employee GPT Mentions. An internal version of GPT-4 or a similar model is presented with lots of data and can now answer questions and support employees. “That’s a good thing conceptual proof“but not where the true value lies,” warns the expert.

Strong together

“The real value comes when you can use AI to develop better internal processes.” This requires the right approach to the rapidly evolving AI field. “AI is not a one-vendor story. In the past, you could invest using Gartner’s magic quadrant. You selected what fits your budget in the top right and waited for your promotion. This is not possible with generative AI because the competitive conditions change every day.”

According to Marchand and Cognizant, the solution lies in a broader, flexible approach. “For optimal results, you don’t need one AI model, but several,” he knows. “With a chatbot, for example, you can start the conversation with OpenAI technology. If the conversation focuses on a medical topic, MedPalm is more suitable.”

Different models can make different contributions. “Orchestration is important. AI works best within a model where different models can work together as a whole. To achieve this, the models need to be able to use shared memory: you don’t want to have to explain your problem again when OpenAI sends you to MedPalm.”

Follow the brain

Cognizant therefore developed Neuro R AI. This is a model independent framework. It allows companies to combine models into a whole, but also to exchange them for new and better versions. The framework ensures that the most appropriate model is used in the right context, while a combination of long-term and short-term memory ensures consistency. When Marchand explains it, we are reminded of our own brain and how it is made up of specialized parts that yet work together as a whole.

“Neuro-AI is not just limited to generative AI either,” notes Marchand. “The framework can perfectly combine new generative AI with classic AI and analytics.” For Marchand, Neuro AI was one of the reasons to join Cognizant. The field is evolving rapidly, but based on his experience he believes the agnostic orchestration approach is the right one.

We are living through turbulent times. Generative AI is shaking the world and will have a major impact in almost every field, not just life sciences. However, everyone faces similar challenges. The currently almost unstoppable further development of AI is just as important as the issue of trust. There are still some unanswered questions, but Marchand and Cognizant already see a clear solution for implementing AI. To achieve the best results, multiple models, algorithms, and types of AI need to work together, and that requires an agnostic framework.