Improving, evaluating and applying AI capabilities

Addressed by:

• Benchmarking of AI system capabilities  
  • Creating and establishing good standard benchmarks (...)
  • Organizing events for system benchmarking
  • Creating transparent platforms for publishing and comparing benchmark results of multiple systems
  • Incentivizing benchmarking (e.g., monetary rewards)
  • Creating metrics for benchmark coverage of certain problem domains
• AI for specific data modalities  
  • Working with symbolic sequence data (...)
  • Working with image data (...)
  • Working with signal data (...)
  • Working with tabular data
  • Working with multimodal data (...)
• Improving specific AI capabilities  
  • Improving generative modeling
  • Improving causal reasoning
  • Inducing useful representations of data (...)
  • Improving the ability of algorithms to work with unstructured data directly, decreasing the need for structured data
  • Interaction of AI systems with the physical world
  • Automated question answering
  • Improving generation of explanations of observations and inferences (...)
• Improving the ecosystem, workflow and tool chain for creating new AI systems  
  • Incentivizing open AI model sharing
  • Making existing open AI models easy to find, use and adapt
  • Creating more powerful and user-friendly AI/machine learning frameworks
  • Making existing AI models easy to access and use

Addresses:

• Grand challenges in utilizing AI to safely accelerate progress in science, technology and society

Categories

• Information technologies