IBM Research Highlights Automation, Natural Language Processing, and Trust

In 2020, automation, natural language processing (NLP), and trust are expected to lead industry advancement in the AI sector.

Through automation, AI systems will operate swiftly and accurately, while natural language processing will play a central role in conversations, discussions, and problem-solving using everyday language. Meanwhile, supported by technologies enabling explainability and bias detection, the trend toward transparently and responsibly managing AI data is expected to expand.

IBM Research announced five major AI-related outlooks on January 2.


Enhanced Comprehension and Task Execution Capability
AI systems have the characteristic of advancing more rapidly as they acquire more data. In 2020, more AI systems are expected to begin relying on "neuro symbolic" technology, which combines learning and logic.

Neuro symbolic is an important indicator in the advancement of natural language processing technology, enabling computers to leverage common sense reasoning and domain-specific knowledge to help humans understand language and conversation more easily and quickly.

Enterprises can utilize this technology to automate customer management in conversational formats and use technical support tools, while also being able to train AI with less data.


Changes in Work Methods
AI will not replace human jobs but will transform the way people work through automation. Experts anticipate that AI will assist with tasks such as schedule coordination but will not directly impact specialized work such as design and strategy development.

Companies adopting AI are expected to reflect the trend of AI transforming work methods by adjusting employee tasks and investing in strengthening employee competencies in relevant areas.


Implementation of Trustworthy AI Systems
To trust AI, systems must be reliable and fair, technology must be secure, and one must be confident that conclusions or recommendations provided by AI are neither biased nor manipulated.

In 2020, components regulating trustworthiness are expected to be incorporated into the AI lifecycle, with trustworthiness being considered alongside performance in AI application development, testing, operation, monitoring, and certification processes.

Additionally, as AI controlling AI emerges, similar to Auto AI which uses AI to create AI, trustworthy AI workflows are expected to form across industries, including heavily regulated sectors.


Growing Demand for Eco-Friendly Technology
Data centers based on AI currently account for approximately 2% of global total energy consumption. In 2020, efforts to make AI technology more sustainable over longer periods are anticipated to increase.

This includes development of new materials such as transition-metal oxides that enable more flexible devices, new chip designs processing both analog and mixed signals, and novel software technologies based on approximate computing that operate with limited computing power while achieving results above certain performance levels.

All such technologies will be utilized as means to reduce carbon emissions while supporting growing AI-related workloads.


AI-Related New Materials Development
Over the past 200 years, organic molecular synthesis has been an important axis of research in chemistry and has been applied to pharmaceutical and synthetic fiber development.

Today, scientists worldwide are researching tens of thousands of chemical reactions to create new molecules, but the vast volume of information makes it impossible for any single expert to master all fields.

However, AI is expected to help overcome these limitations. As IBM developed RXN for Chemistry, an AI tool capable of synthesizing molecules on the cloud and predicting millions of chemical reactions before and after, 2020 is anticipated to see breakthrough progress in discovering and developing new materials utilizing AI and automation technologies.