What’s Next in AI is fluid intelligence
What’s Next in AI is fluid intelligence
Our vision
Today's AI is narrow. While many AI models deliver value in specific, well-defined situations, applying those same models to new challenges requires an immense amount of new data and training.
Enterprises need AI that is fluid and adaptable, capable of applying knowledge acquired for one purpose to new domains and challenges. They need AI that can combine different forms of knowledge, unpack causal relationships, and learn new things on its own.
In short, enterprises need AI with fluid intelligence — and that's exactly what we're building.
Strategy
Four workstreams will achieve our vision


To achieve our vision, we'll have to make significant strides in algorithm development while improving our AI engineering tools and hardware.
Neurosymbolic AI ↓There is no AI without algorithms. In our pursuit of fluid intelligence, we are continuing to advance our work in deep learning, and are putting significant investment into neurosymbolic AI, which combines the power of neural networks with symbolic methods to help AI reason more effectively.
Secure, Trusted AI ↓In parallel, we're delivering secure, trusted AI, with work focused on explainability, fairness, and bias reduction.
AI Engineering ↓AI engineering, tools and capabilities that simplify and automate key tasks like data preparation, training and lifecycle management.
AI Hardware ↓Finally, we're developing a new class of AI hardware, that is inherently energy- efficient, increasing compute power by orders of magnitude without the demand or increased energy.
Strategic teams
Neurosymbolic AI
We're integrating neural and symbolic techniques to build AI that can perform more complex tasks by understanding and reasoning more like we do.
More on Neurosymbolic AI research
Featured progress
Blog
Getting AI to Reason: Using Neuro-Symbolic AI for Knowledge-Based Question Answering
4-Dec-2020

Blog

Neurosymbolic AI to Give Us Machines With True Common Sense
8-July-2020

Open Source

CLEVRER: The first video dataset for neuro-symbolic reasoning
18-Sept-2020

Blog

Logical Neural Networks: From Deep Learning to Deep Thinking
3-Dec-2020

Secure, Trusted AI
Trust and security should be baked into the core of any AI we put out into the world. We're building tools to help you ensure that it is.
More on Secure, Trusted AI research
Featured progress
Blog

Unleashing Cyber Reasoning Potential in The Era of AI Security
1-May-2020

Open Source

AI 360 Toolkits: Fairness, Explainability, and Robustness
18-Sept-2020

AI Engineering
We're building tools to help AI creators reduce the time they spend training, maintaining, and updating their models.
Featured progress
Blog

AI for AI set to make it easy to create machine learning algorithms
3-Feb-2020

Blog

Automating machine learning with a joint selection framework
1-May-2019

AI Hardware
Our digital and analog accelerators are driving massive improvements in computational power while remaining energy-efficient.
Featured progress
Blog

Fulfilling Brain-inspired Hyperdimensional Computing with In-memory Computing
1-June-2020

Blog

IBM Systems Research: AI and Hybrid Cloud Can Advance Only as Far as Hardware Can Take Them
16-June-2020
