Connect with us

Artificial Intelligence

IBM Reportedly Retreating from Healthcare with Watson  – AI Tendencies



IBM is reported to be contemplating a sale of Watson Well being, a sign of the challenges of making use of AI in healthcare. IBM continues to spend money on cloud companies utilizing Watson. (Photograph by Carson Masterson on Unsplash.)  

By John P. Desmond, AI Tendencies Editor  

Experiences surfaced final week that IBM is considering a sale of Watson Well being, representing a retreat from the market of AI utilized to healthcare that IBM had pursued below the course of its earlier CEO. 

The Wall Road Journal final week reported IBM was exploring the sale of Watson Well being; IBM didn’t affirm the report. Ten years in the past, when IBM Watson received on the Jeopardy! recreation present in opposition to two of the sport’s file winners, the Watson model in AI was established. 

As reported in AI Tendencies final February, the day after Watson defeated the 2 human champions on Jeopardy!, IBM introduced Watson was heading into the medical area. IBM would take its potential to know pure language that it confirmed off on tv, and apply it to drugs. The primary industrial choices can be accessible in 18 to 24 months, the corporate promised, based on an account in IEEE Spectrum from April 2019.  

It was a troublesome highway. IBM was the primary firm to make a serious push to deliver AI to drugs. The alarm was sounded by Robert Wachter, chair of the division of medication on the College of California, San Francisco, and creator of the 2015 ebook The Digital Physician: Hope, Hype, and Hurt on the Daybreak of Medication’s Laptop Age (McGraw-Hill). The Watson win on Jeopardy! Gave the IBM AI salesforce a launching pad.  

“They got here in with advertising first, product second, and obtained everyone excited,” said Wachter. “Then the rubber hit the highway. That is an extremely arduous set of issues, and IBM, by being first out, has demonstrated that for everybody else.”  

Then-IBM CEO Ginni Rometty Used Watson Victory to Launch AI in Healthcare  

Ginni Rometty, IBM’s CEO on the time, advised an viewers of well being IT professionals at a 2017 convention that “AI is mainstream, it’s right here, and it will possibly change nearly the whole lot about well being care.” She, like many, noticed the potential for AI to assist rework the healthcare trade. 

Watson had used advances in pure language processing to win at Jeopardy. The Watson group used machine studying on a coaching dataset of Jeopardy clues and responses. To enter the healthcare market, IBM tried utilizing textual content recognition on medical information to construct its information base. Unstructured knowledge resembling docs’ notes filled with jargon and shorthand could account for 80% of a affected person’s file. It was difficult.   

The hassle was to construct a diagnostic software. IBM shaped the Watson Well being division in 2015. The unit made $4 billion of acquisitions. The search continued for the medical enterprise case to justify the investments. Many initiatives have been launched round choice assist utilizing massive medical knowledge units. A give attention to oncology to personalize most cancers remedy for sufferers appeared promising.  

Physicians on the College of Texas MD Anderson Most cancers Heart in Houston, labored with IBM to create a software referred to as Oncology Skilled Advisor. MD Anderson obtained the software to take a look at stage within the leukemia division; it by no means turned a industrial product.   

The mission didn’t finish properly; it was cancelled in 2016. An audit by the College of Texas discovered the most cancers middle had spent $62 million on the mission. The IEEE Spectrum authors stated the mission revealed “a elementary mismatch between the promise of machine studying and the fact of medical care,” one thing that may be helpful to at the moment’s docs.  

IBM made a spherical of layoffs within the IBM Watson Well being unit in 2018, based on one other report on the time by IEEE Spectrum in June 2018. Engineers from one of many firms IBM had acquired, Phytel, reported a shrinking shopper base for its affected person analytics resolution from 150 to 80 because the acquisition. “Smaller firms are consuming us alive,” said the engineer. “They’re higher, sooner, cheaper. They’re successful our contracts, taking our clients, doing higher at AI.”  

Mismatch Seen Between Realities of Healthcare and Promise of AI  

Dr. Thomas J. Fuchs, Dean of AI and Human Well being, Mount Sinai Well being System

This notion of a mismatch between the promise of AI and realities of healthcare was seconded in final week’s Wall Road Journal report that tech firms could lack the deep experience in how healthcare works in affected person settings. “You really have to know the medical workflow within the trenches,” said Thomas J. Fuchs, Mount Sinai Well being System’s dean of synthetic intelligence and human well being. “It’s important to perceive the place you possibly can insert AI and the place it may be useful” with out slowing issues down within the clinic. 

Packaging AI advances in pc science right into a viable software program services or products has at all times been a elementary problem within the software program enterprise. “Watson could also be very emblematic of a broader concern at IBM of taking good science and discovering a strategy to make it commercially related,” said Toni Sacconaghi, an analyst at Bernstein Analysis.  

Toni Sacconaghi, analyst, Bernstein Analysis

New IBM CEO Arvind Krishna has stated AI together with hybrid cloud computing, can be pivotal for IBM going ahead. (See AI Tendencies, November 2020.) Krishna is shifting to exit struggling enterprise models and focus on these that may ship constant development. As a part of this effort, IBM is within the technique of spinning its managed IT companies division out into a brand new public firm; IT companies is seen as a declining margin enterprise by analysts. IBM had $100 billion in gross sales in 2010 and $73.6 billion final 12 months. 

One other problem for AI in healthcare is the dearth of data-collection requirements, which makes making use of fashions developed in a single healthcare setting and making use of it in others is tough. “The customization drawback is extreme in healthcare,” said Andrew Ng, an AI knowledgeable and CEO of startup Touchdown AI, based mostly in Palo Alto, Calif., to The Wall Road Journal. 

Healthcare markets the place AI has proven promise and achieved outcomes embrace radiology and pathology, the place picture recognition methods can be utilized to reply particular questions. Additionally, AI has made inroads in streamlining enterprise processes resembling billing and charting, which can assist get monetary savings and liberate workers to give attention to tougher areas. Administrative prices are stated to be 30 p.c of healthcare prices. 

In the meantime, funding for AI in healthcare continues, with spending projected to develop at an annualized price of 48% by 2023, based on a current report from Enterprise Insider. New gamers embrace giants resembling Google, which has outlined a Cloud Healthcare software programming interface (API), that may take knowledge from customers’ digital well being information through machine studying, with the goal of serving to physicians make extra knowledgeable medical choices. Google can be working with the College of California, Stanford College, and the College of Chicago on an AI system to foretell the outcomes of hospital visits 

AI can be being utilized to the transfer to personalised healthcare, for instance with wearable know-how resembling FitBits and smartwatches, which may alert customers and healthcare professionals to potential well being points and dangers.  

Whereas retreating from making use of Watson in healthcare, IBM is increasing the position of Watson in its cloud service choices. These embrace pure language processing, sentiment evaluation and digital assistants, based on entries on the IBM Watson weblog,  

Learn the supply articles and data in The Wall Road Journalin IEEE Spectrum from April 2019, in AI Tendencies February 2020, in IEEE Spectrum from June 2018, AI Tendencies, November 2020, from Enterprise Insider and on the IBM Watson weblog.  

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *