Response to - Request for Information
Preparing for the Future of Artificial Intelligence


IBM has been researching, developing and investing in AI technology for more than 50 years. The public became aware of a major advance in 2011, when IBM Watson won the historic Jeopardy! exhibition on prime time television. Since that time, the company has advanced and scaled the Watson platform, and applied it to various industries, including healthcare, finance, commerce, education, security, and the Internet of Things. We are deeply committed to this technology, and believe strongly in its potential to benefit society, as well as transform our personal and professional lives.

To this end, we have engaged thousands of scientists and engineers from IBM Research and Development, and partnered with our clients, academics, external experts, and even our competitors to explore all topics around AI. And we have developed a unique point-of-view, informed by decades of research and commercial application of AI.

At IBM, we are guided by the term “augmented intelligence” rather than “artificial intelligence.” It is the critical difference between systems that enhance and scale human expertise rather than those that attempt to replicate all of human intelligence. We focus on building practical AI applications that assist people with well-defined tasks, and in the process, expose a range of generalized AI services on a platform to support a wide range of new applications.

We call our particular approach to augmented intelligence “cognitive computing.” Cognitive computing is a comprehensive set of capabilities based on technologies such as machine learning, reasoning and decision technologies; language, speech and vision technologies; human interface technologies; distributed and high-performance computing; and new computing architectures and devices. When purposefully integrated, these capabilities are designed to solve a wide range of practical problems, boost productivity, and foster new discoveries across many industries. This is what we bring to market today in the form of IBM Watson.

The following are brief responses to the questions in the RFI (re-ordered and slightly re-factored), with links to more detailed information.

A. The use of AI for public good
(RFI question 2)

For decades, we have been stockpiling digital information. We have digitized the history of the world’s literature and all of its medical journals. We track and store the movements of automobiles, trains, planes and mobile phones. And we are privy to the real-time sentiments of billions of people through social media. It is not unreasonable to expect that within this rapidly growing body of digital information lies the secrets to defeating cancer, reversing climate change, or managing the complexity of the global economy. We believe that many of the ambiguities and inefficiencies of the critical systems that facilitate life on this planet can be eliminated. And we believe that AI systems are the tools that will help us accomplish these ambitious goals.

We are already doing much of this work:

  • For healthcare, AI systems can advance precision medicine by ingesting patients’ electronic medical history and relevant medical literature, performing cohort analysis, identifying micro-segments of similar patients, evaluating standard-of-care practices and available treatment options, ranking by relevance, risk and preference, and ultimately recommending the most effective treatments for their patients.
  • For social services, AI systems can provide timely and relevant answers to citizens in need, assist citizens with insurance, tax, and social programs, predict the needs of individuals and population groups, and develop plans for efficient deployment of resources.
  • For education, AI systems can assist teachers in developing personalized educational programs for individuals or groups of students, assist students using a range of learning styles and methods, and develop effective early education, primary, secondary, and higher education programs.
  • For financial services, AI systems can expand financial inclusion by qualifying applicants, assist in providing the best insurance coverage at the right cost, ensure compliance with federal, state and local regulations, and reduce fraud and waste in tax and other financial programs.
  • For transportation, AI systems can improve the efficiency of public transportation systems, support public vehicles with driver assistance using semi-automated features, manage incidents, optimize the use of fuel and support maintenance of infrastructure and rolling stock.
  • For public safety, AI systems can support safety personnel with anomaly detection using machine vision, build predictive models for crime, and help investigators find associations in massive amounts of information.
  • For the environment, AI systems can understand complex relationships and help construct environmental models for accurate prediction and management of pollutants and carbon footprints.
  • For infrastructure, AI systems can assist with prediction of demand, supply, and use of infrastructure, planning and execution of projects, and maintenance of built infrastructure.

See more here.

B. Social and economic implications of AI
(RFI question 4)

AI systems are already changing the way work gets done. But history suggests that new technologies like AI result in higher productivity, higher earnings, and overall job growth. In particular, we believe that new companies, new jobs, and entirely new markets will be built on the shoulders of this technology. And we believe that AI systems will improve access to critical services for underserved populations. Overall, we anticipate widespread improvements in quality of life.

In order to be fully accepted into society, AI systems need to have significant social capabilities, because their presence in our lives has a profound impact on our emotions and on our decision making capabilities (e.g., elder care). AI systems also need to understand how to learn and comply with specific behavioral principles for aligning with human values.

See more here.

C. Education for harnessing AI technologies
(RFI question 7)

The potential for AI solutions for public and private uses has created a fast growing demand for AI skills. To meet this demand, top universities are crafting new AI curricula. Leading firms offer faculty and students access to cloud platforms with AI-based services, from image recognition to machine learning. However, most courses and platforms require programming skills and advanced mathematics as prerequisites. Government agencies, research institutions, universities, and foundations can work together to make learning to build, understand, and work with AI systems more accessible to a broader range of students and professionals retooling their careers.

See more here.

D. Fundamental questions in AI research, and the most important research gaps (RFI questions 5 and 6)

In order for AI systems to enhance quality of life, both personally and professionally, they must acquire broad and deep knowledge from multiple domains, learn continuously from interactions with people and environments, and support reasoned decisions. Broadly, the AI fields’ long-term progress depend upon many advances:

  • Machine learning and reasoning: Most current AI systems use supervised learning, using massive amounts of labeled data for training. Fundamental research is needed for AI systems that learn as humans do: through instruction, interaction (by discussing, debating, watching other people learn), by doing things (utilizing motor skills), generalizing from very little data, and by transferring skills across many tasks.
  • Decision techniques: For AI-based systems to succeed broadly, new techniques must be developed for modeling systemic risks, analyzing tradeoffs, detecting anomalies in context, analyzing data while preserving privacy, and making decisions under uncertainty.
  • Domain-specific AI systems: Deeply understanding the domains of human expertise, such as medicine, engineering, law and thousands more, poses particularly difficult issues of knowledge acquisition, representation, and reasoning. AI systems must ultimately perform professional-level tasks, such as managing contradictions, designing experiments, and negotiating.
  • Data assurance and trust: Training and test data can be biased, incomplete, or maliciously compromised. Significant effort should be devoted to techniques for measuring entropy of datasets, validating the quality and integrity of data, and for making AI systems more objective, resilient, and accurate. People will trust AI systems when systems know users’ intents and priorities, explain their reasoning, learn from mistakes, and can be independently certified.
  • Radically efficient computing infrastructure: When deployed at scale, AI systems will need to handle unprecedented workloads that will require the development of high-performance distributed cloud systems, new computing architectures such as neuromorphic and approximate computing, and new devices such as quantum and new types of memory devices.

See more here.

E. Data sets that can accelerate AI research
(RFI question 9)

A major bottleneck in developing and validating AI systems is public access to sufficiently large, openly curated, public training data sets. Machine learning, supervised and unsupervised, requires large, unbiased data sets to train accurate models. Deep learning is advancing speech transcription, language translation, image captioning, and question and answering capabilities. Each new AI advance, e.g., video comprehension, requires the creation of new data sets. Deep domain tasks, such as cancer radiology, or insurance adjustment, requires specialized and often hard-to-get datasets. Incentives must be created for greater sharing of both input datasets and trained models through mechanisms like model zoos.

See more here.

F. Multi-disciplinary research
(RFI question 8)

Most of the research areas in Section D cannot be achieved by AI researchers alone. Collaboration with experts in multiple disciplines -- such as psychology, philosophy, sociology, art, regulation, and law -- will be crucial. In addition, there is an important role for professional associations with industry-specific knowledge to play in informing AI applications. To this end, IBM is in the process of creating a network of several academic centers to jumpstart the scientific ecosystem.

See more here.

G. Role of incentives and prizes
(RFI question 10)

As the fundamental building blocks of AI improve, so too should the incentives that inspire next-generation, people-centered systems design. As an example, IBM established a $5 million AI XPrize for the best use of AI system to empower teams of people to tackle the world’s grand challenges. IBM is developing additional scientific challenges for the AI research community.

See more here.

H. Safety and control issues for AI
(RFI question 3)

To reap the societal benefits of artificial intelligence, we will first need to trust it. That trust will be earned through experience, of course, in the same way we learn to trust that an ATM will register a deposit, or that an automobile will stop when the brake is applied. Put simply, we trust things that behave as we expect them to.

But trust will also require a system of best practices that can guide the safe and ethical management of AI; a system that includes alignment with social norms and values; algorithmic accountability; compliance with existing legislation and policy; and protection of privacy and personal information. IBM is in the process of developing this system in collaboration with our partners, university researchers, and competitors.

See more here.

I. Legal and governance implications of AI
(RFI question 1)

Responsibility must be the foundation for AI policymaking. Inclusive dialogues can explore relevant topics, going beyond the headlines and hype, promoting deeper understanding and a new skills focus. Every transformative tool that people have created – from the steam engine to the microprocessor – augment human capabilities and enable people to dream bigger and do more. People with these tools will solve whole new classes of big data problems. Our responsibility as members of the global community is to ensure, to the best of our ability, that AI is developed the right way and for the right reasons.

See more here.

J. Other issues: Business models
(RFI question 11)

In market-driven economies, progress also crucially depends upon the creation of new business models that rewards more effective outcomes and overall benefits to society.

See more here.

Concluding Remarks

AI systems are augmenting human intelligence and will ultimately transform our personal and professional lives. Its benefits far outweigh its risks. And with the right policies and support, those benefits can be realized sooner.

Policy makers should focus on:

  • Facilitating a fact-based dialogue on the capabilities and limitations of AI technologies
  • Developing progressive social and economic policies to deploy AI systems for broad public good
  • Developing progressive education and workforce programs for future generations
  • Investing in a long-range interdisciplinary research program for advancing the science and design of AI systems

See more here.