Driving Research and Innovation in Decision AI

AI research is at an exciting stage, where decades of progress are now applied at an enterprise scale. We see Decision AI as the next major leap, going beyond predictions to prescribe real-world actions. Our research uses techniques like multi-agent systems, generative AI, deep learning, evolutionary computation, neuroevolution, surrogate modeling, and trustworthy AI to drive innovation. Our work leads to publications, open-source tools, AI-for-Good projects, and the development of the Cognizant Neuro® AI platform.

Spotlight: Papers at NeurIPS 2024

We presented two papers at the NeurIPS conference this year. Click on the image see a summary (a blog post and a video).

Unlocking the Potential of Human Expertise with AI

Evolutionary optimization can discover how to take advantage of good ideas in human-created solutions even when they are not perfect.

Semantic Density: Quantifying Uncertainty in LLMs

This new method allows measuring how likely each LLM-generated response is to be truthful vs. a hallucination.


Technologies

Evolutionary Computation

Evolutionary computation is an optimization method that mimics natural selection, helping discover innovative strategies for decision-making in complex environments.

Deep Learning

Deep learning trains neural networks to recognize patterns in data. Our work focuses on improving models with multitask learning and compression techniques for better scalability with large-scale data.

Neuroevolution

Neuroevolution combines evolutionary principles with deep learning to optimize neural networks. Our recent advancements focus on making AI more adaptive and efficient.

Generative AI

Generative AI, such as large language models (LLMs) encode general world knowledge that can be accessed through natural language. We use LLMs to make deployments of Decision AI systems general and robust.

Multi-agent AI

Multi-agent AI leverages LLMs to facilitate communication between autonomous agents, enhancing coordination and optimizing complex problem solving across diverse systems.

Surrogate Optimization

Surrogate Optimization uses models to simulate complex environments, enabling safe and efficient strategy optimization. This approach solves real-world problems, such as healthcare interventions.

Trustworthy AI

Trustworthy AI ensures AI models are safe, explainable, and built with accountability for confidence in every decision.


Research

Our research is reported in scientific papers, advancing new technologies that enhance decision-making.


Applications

We collaborate with experts from various fields to apply our AI research to real-world problems, allowing us to develop better AI methods while maximizing human potential.


Media

Our research and applications are being recognized in the media, exploring AI’s impact in transforming industries and addressing societal needs.