Trusting Watson

Will IBM’s Artificial Intelligence Moves Repair Bad PR?

Named after IBM’s first CEO, industrialist Thomas J Watson, IBM’s artificial intelligence (AI) platform arrived with much fanfare in 2011 when it beat human contestants on the US TV show “Jeopardy.” But a recent string of bad publicity has hit Watson and cast a shadow over IBM’s efforts. It comes just as enterprise interest in AI has accelerated and competitors such as Microsoft, Google and Amazon Web Services have piled into the market.

Social Capital investor Chamath Palihapitiya famously stated in 2017 that “Watson is a joke”. In the same year, Forbes ran a much-circulated article highlighting MD Anderson Cancer Center’s decision to move away from the platform. Critical articles in the Economist and Wall Street Journal have surfaced in 2018, with the latter stating that “the diagnosis for Watson is gloomy“. IBM’s flat financial performance in its Cognitive Business Solutions unit hasn’t helped either.

All signs suggest IBM is floundering in the high-stakes battle for AI supremacy. However, over the past few weeks, the company has made some important announcements, building on a solid run of moves in 2018, that indicate the tide may be turning.

IBM Makes Headway in Artificial Intelligence

2018 is turning out to be an important year in the progress of IBM’s strategy for AI. At its Think event in March (which I covered here), IBM introduced several new products in data analytics, governance and management tools, data science services, application development, conversational agents and above all, as a result of its partnership with Apple. Later the same month it introduced Watson Data Kits, machine-readable data sets for firms in the travel, transport and food industries looking to build workflow applications with AI.

Today, IBM unveiled some of the most intriguing announcements in AI to date: new trust and transparency capabilities on the IBM Cloud aimed at helping AI administrators make the performance, health and explanations of machine learning models more transparent. They include runtime tools that help detect bias mitigation and fairness in machine learning models; tools to track model accuracy and improve the traceability and auditability of predictions made in applications, as well as features that help explain in business terms the outcomes and recommendations made by AI. IBM also said it will open source IBM Research’s bias detection tools.

Trust: The New Technology Battleground

With these moves, IBM is acting on the exhortation CEO Ginni Rometty made to the industry during Think to be better stewards of customer data, and her assurances that data trust and responsibility were important to the company. Trust has now become a major battleground in the technology industry as a whole, alongside traditional areas like product innovation and developer pull. Results from CCS Insight’s latest survey of employees’ attitudes to technology revealed trust in the tech giants is shifting rapidly following a year of security breaches and privacy scandals.

Trust, or a lack of it, is the biggest barrier to the adoption of machine learning in enterprises. We highlighted this trend in our survey of IT decision-makers in 2017. Watson has been criticized as being too much of a “black box” in the past: businesses don’t understand how it works owing to a lack of transparency and auditability in the outcomes it generated. IBM is now tackling these concerns head-on by providing tools that make it easier to explain AI in production and in ways that business users can understand.

“Climb the Ladder to AI”

It was against this backdrop that IBM held its “Change the Game: Winning with AI” event in New York City last week. With several customers on show including Experian, Geisinger and Mueller, the event was designed to showcase progress in AI and to help firms “climb the ladder to AI”, from data collection and organization, to insight and analytics, toward the final goal of automation with machine learning.

Not having an AI strategy today is the equivalent of not having an Internet strategy in 2000, according to IBM. Yet just one in 20 companies are getting real use of the technology because of poor data. At Think, IBM released Cloud Private for Data, a clumsily named but important product that centralizes and integrates data science, data engineering and application building processes into a single data platform to address these challenges. It aims to improve data collection, organization and analysis as well as cross-functional collaboration and governance for machine learning projects.

IBM announced several improvements to the platform at last week’s event, including IBM Cloud Private for Data Experiences, a free programme for those starting out, and a data-query service to help data scientists search for information on public cloud, on-premises and mobile device deployments.

The AI Race Heats Up

The announcements are timely reminders of IBM’s place among the leading AI platform suppliers as rivals host events to trumpet their credentials. After Google’s Cloud Next conference in July, Microsoft holds its Ignite later in September and Amazon stages re:Invent in November.

The announcements also reinforce some of IBM’s strengths against these competitors. It has a first-mover advantage in applied AI in industries and business applications such as health, legal, financial compliance, retail, travel and advertising. It has vital assets in its Apple partnership, which gives Watson preferential access to more than 3 million iOS developers globally.

It also leads in research, which is the source of many of its new AI products. In 2017, it invested $240 million in a partnership with Massachusetts Institute of Technology to create an AI lab. At an industry analyst event I attended in London this July, IBM demonstrated its latest research in computational argumentation from its Project Debater, showing how Watson can mine a corpus of 300 million articles and, using natural language processing, deep learning and text-to-speech reasoning, successfully debate with a human in the fields of genetic engineering and telemedicine. IBM believes this research will have practical applications in financial and legal professions.

These strengths translate to a broad portfolio of products — from systems to services — and maturity in AI. In addition to more than 16,000 Watson engagements in 20 industries and 80 countries, IBM has more than 250 AI projects deployed for internal use in areas like dashboards and sales.

Fighting the PR Battle

Negative publicity still often overshadows these strengths. IBM appears to be losing the perception battle in AI, arguably its biggest challenge at the moment. It therefore needs to communicate better and, above all, simplify its strategy. AI involves a complex web of products from its huge portfolio of data analytics, systems, security and cloud services — many of which also carry the Watson name. This complexity has rendered its messaging muddled and has diminished the visibility of its commercial success and the value of the Watson brand.

Understandably, IBM is becoming more modest about AI, having learned from the marketing mistakes of the past, when it hyped Watson and set customer expectations too high. But it will need to be more confident of its own customer successes and promote Watson’s unique differentiators more. It should do this by focusing on the singular business scenarios that matter most to the industries and customers it serves.

Above all, the announcements over the past week show us that IBM is moving in the right direction in 2018. Communication will be more important than ever to ensure it maintains its position as one of the leading AI players in the world.