Submitted applications should concentrate primarily in one of the IBM Focus Areas categories shown below.
- Optimization of open standards and open source code resulting in the enablement and
creation of a seamless hybrid cloud platform that can be deployed anywhere. Proposals can include
Platform Enablement and Optimization with an open source component; or use this category for
interdisciplinary entries.
- Flexibility and Scalability. Submittals could propose research that leads to the creation of
hybrid cloud platforms enables more flexible, scalable computing, unifying local environments
with a virtually limitless pool of computing power and capabilities, making bits, neurons, and
qubits available on-demand.
- Accelerating adoption. Making hybrid cloud adoption easier and safer by enhancing agility through
automation. Submittals could address leveraging AI for code to help automate essential tasks like
application modernization, vulnerability detection in code, troubleshooting of IT reliability
issues, and research focusing on cutting edge technologies in antivirus and other protection
mechanisms.
AI advancements in hardware or training models, characterizing and classifying unknown instances, and federated
learning. Security topics are highly encouraged: data encryption, storage advancements, unified endpoint
management, and firmware or chip-level proposals.
AI Hardware
-
Processing efficiency. The next improvements in devices, architectures, and algorithms Nominations could include
research that combines these topics and apply them to Deep Neural Networks far beyond the present architectures
of GPUs and CMOS Accelerators.
-
Digital AI Cores. New accelerators for existing semiconductor technologies that use reduced precision to speed
computation and decrease power consumption using reduced precision techniques.
-
Analog Cores. Memory-based technology to advance AI at VLSI, analog memory devices and hardware accelerators, mixed
precision in-memory computing, hybrid design for AI Software, and other 8-bit breakthroughs.
Hetergeneous Integration. AI applications drive the need for a system level optimization of AI Hardware through
Heterogeneous Integration of Accelerators, Memory and CPU to enable high-speed/high-bandwidth connectivity components.
Proposals can bridge these areas.
-
Machine Intelligence/Neural Networks. Machine Intelligence differs from machine learning.
Solving some of AI's greatest challenges using associative reasoning to mimic human intelligence.
with brain science.
AI Engineering
- Optimization. Tools for AI creators to reduce the time they spend training, maintaining, and updating their models.
New approaches, strategies, and research to explore advanced problems automatically. Best models for ML and data science
pipelines, best architectures for deep learning, and best hyperparameters for AI models and algorithms.
- Privacy and Security. IT Infrastructure consumption models, privacy assurance, hybrid cloud strategies, storage
infrastructures, privacy and security assurance held in the hardware, and progressive hybrid cloud infrastructure
including storage optimization.
Neuro Symbolic AI
- Deep Learning to combine the power of neural networks with symbolic methods to advance AI reasoning effectiveness.
-
Neuro-symbolic AI NLP and QA. Submittals that cover applied challenges posited by neural networks, like symbolic AI,
Q&A, probabilistic physics inference models, or new neuro-symbolic technique. New systems for knowledge-based question
answering.
Secure, Trusted AI
- Building evaluating, and monitoring for trust. AI is developing diverse approaches for how to
achieve fairness, robustness, explainability, accountability, value alignment, and how to integrate them throughout
the entire lifecycle of an AI application.
- Techniques to detect and mitigate bias in datasets and models. Addressing the need for understanding and removing
gender stereotypes, as well as citing and/or rating AI services for bias.
- Robustness, and Privacy. Evaluating and defending machine learning models and applications against adversarial
threats and/or conform to required privacy.
- Explain-ability, Accountability, and/or Transparency. Advancing an AI system to ‘explain itself.’ Exploring the
inner workings of an algorithm to provide stakeholders explanations for different purposes and objectives that are
tailored to their needs.
- Designing for security and compliance in niche areas or across the stack: from the hardware, encryption
technologies, the hybrid cloud platform, to the SecDevOps pipeline. Trusted service identity/ identity access management
across the stack or address niche areas such as high assurance through Encrypted Container Images. Research covering
any and all arrays of software and platforms are encouraged.
- Advanced foundational quantum information science.
Exploring and developing new quantum algorithms to reduce error rates and ensure more accurate and reliable results.
- Quantum hardware.
Specialized quantum hardware and systems to scale Quantum volume while also increasing qubit count.
- Quantum circuits and software
to explore and develop compelling use cases for this powerful form of computing.