Data science and AI predictions for 2019
Data science became the highest-paid IT profession in 2018, and the field is set for further growth in 2019 as the tools and techniques become more accessible and AI moves from hype to practical use cases. By Tom Macaulay
A recent Deloitte survey estimated 57 percent of businesses are increasing spending in AI as organizations start to wake up to the potential business benefits.
“We are just at the beginning of the enterprise machine learning transformation. In 2019, we’ll see a new step in maturity, as companies advance from PoCs to production capabilities,” says Stephen Line, VP of EMEA at Cloudera.
“Enterprise machine learning adoption will continue as businesses look to automate pattern detection, prediction and decision making to drive transformational efficiency improvement, competitive differentiation and growth. As early adopters advance from proof-of-concepts to production deployment of multiple use-cases, we’ll continue to see an emergence of technologies and best practices aimed at helping operationalize, scale and ultimately industrialize these capabilities to achieve full transformational value,” he predicts.
Forrester principal analyst Michele Goetz believes that these developments will help use cases shift from unblocking bottlenecks in a process to uncovering new ways to execute the process.
“The AI capabilities that come pre-trained will still be popular, but they will become more embedded in broader solutions,” she says. “I can see acquisitions gaining steam. As firms become more adept at using AI to reengineer rather than tuning and automating tedious tasks, the value of AI will start to outshine existing analytic approaches that focus on narrow tasks and scoring.”
Forrester analyst Duncan Jones expects more vendors to embed AI in their software, reducing the need for IT departments to build it into their own tools.
Jones also believes that automation will shift the focus of business intelligence software “from drill down to alert up”, using automated checks to alert people about what warrants their attention.
“You end up managing more by exception than by rubber-stamping everything. That’s the prediction with business software,” he says. “The processes will become very much less manual. Everything will be stripped out and automated so that the human beings will be looking at stuff that they really need to look at, and the software will be dealing with everything else.”
The growing capabilities of data science could also lead to some rapid developments in emerging technologies. Gartner fellow David Cearley believes that the growth of autonomous things such as robots, drones, and vehicles to deliver advanced behaviors that interact more naturally with their surroundings and with people.
“As autonomous things proliferate, we expect a shift from stand-alone intelligent things to a swarm of collaborative intelligent things, with multiple devices working together, either independently of people or with human input,” he says.
“For example, if a drone examined a large field and found that it was ready for harvesting, it could dispatch an ‘autonomous harvester.’ Or in the delivery market, the most effective solution may be to use an autonomous vehicle to move packages to the target area. Robots and drones on board the vehicle could then ensure final delivery of the package.”
Forrester analyst Goetz also expects a growth in conversational experiences as natural language processing becomes more sophisticated.
“Virtual agents will come with more job expertise and the ability to engage across a broader set of conversational dimensions in a single engagement,” she says.
Exasol CTO Mathias Golombek shares the sentiment. “Amazon Echo, Google Home, and Apple Home pods have brought connected assistants to the home. For the first time, voice interaction has become a mainstream method of controlling devices to play music, get basic information, and administer smart home devices. However, these devices haven’t made much of an impact in business,” he says.
“My expectation is that in 2019, they will find their voices in niche business scenarios too, and that connected assistants will be interfacing to email, CRM systems, and diaries, to streamline processes and give a helping hand.”
Reuven Harrison, CTO at security policy company Tufin, believes this growth in voice applications will lead to an increasing number of data breaches of audio.
“These attacks will manipulate people into inadvertently giving voice commands or playing audio on their computer, prompting a sequence of events that leads to information on company performance or to further gather network information to ease an attack,” he says.
New methods of machine deception will further threaten consumer trust, as the emergence of deepfake video manipulation technique has demonstrated.
“At least for now, detection and forensic technologies have been able to ferret out fake video and images. But the tools for generating fake content are improving quickly so we must ensure that detection technologies are able to keep pace,” says Ben Lorica, Chief Data Scientist at O’Reilly Media.
“Machine deception does not just refer to machines deceiving humans, however. It also refers to machines deceiving machines – bots – and people deceiving machines – troll armies and click farms. Information propagation methods and click farms will continue to be used to fool ranking systems on content and retail platforms, and methods to detect and combat this will have to be developed as fast as new forms of machine deception are launched.”
GDPR and building trust
The introduction of GDPR could make any such breaches extremely costly. The ICO is yet to issue a big fine for a GDPR breach, but David Francis, head of security at IT services provider KCOM, believes 2019 will be the year it happens for the first time.
“If 2018 was the year of compliance, 2019 will be the year of retribution for everyone’s favorite data privacy regulation,” he says. “The period of grace is drawing to a close, and the new year will see the ICO taking its first high-profile scalp over treatment of personally identifiable information.
“That will set the precedent by which all further cases are judged – letting companies know along the way just how strictly enforced the rules are going to be, and how heavy the fines. Now is the time to check your compliance levels – don’t wait for the hammer to fall,” Francis said.
Despite the risks, companies are struggling to implement safe and effective data protection practices. A recent KPMG survey revealed that 61 percent of CEOs view building trust as a top three priority for their organization, but just 35 percent of IT decision-makers have a high level of trust in their organization’s analytics.
Forrester analyst Goetz believes explainable AI will need to evolve to build trust.
“Much of the emphasis on ML training has focused on the algorithm. However, the number one issue with AI is trusting the data,” she says. “This isn’t just a data science issue; it’s also a business issue. First, business experts need better ways to instruct data scientists what representative data should go into the model.
“Second, they need environments to understand and correct how the AI capability does its job. Scatter charts, histograms, etcetera aren’t business-expert friendly. Even traditional dashboards with charts and graphs can be incomplete. Business intelligence, data curation and management environments will help close this gap in trusting and onboarding AI.”
Mike Guggemos, CIO at Insight, expects that the fears will begin to subside, leading organizations to make a renewed push across all business functions to use as many AI-driven services as possible.
“2018 was the year of peak AI hype and anxiety. In 2019 we will move into the next stage of the hype cycle – nascent yet broad enterprise adoption,” he says. We’ll see an acceleration of real-world applications of AI and the technology will seep into the fabric of businesses and offices everywhere, changing business operations without most even noticing.
“AI services are becoming available to organizations from providers such as AWS and Microsoft as reference architectures which will emerge into – not as easy as it sounds – plug and play standard services. As AI becomes part of everyday business life, widespread and much hyped fears of job losses will subside as people realise that it mainly assists rather than replaces humans.”
The talent pipeline has struggled to keep up with developments in data science tools and techniques. ABBYY’s Global VP Neil Murphy expects that this will lead one-third of organizations adopting AI to hire more IT staff in the next six months.
“The need for specialized skills to work with AI and automation technologies will drive a huge hiring spree in 2019 across the world – in Europe and the US, one in three businesses will need to hire more employees in their IT departments to accelerate their tech offerings in 2019,” he says.
“Across industries from manufacturing and healthcare to non-profits, government, and financial services, the biggest challenge will be the same: upgrading their IT infrastructure and replacing legacy systems without failing on their digital transformation efforts. In order to achieve this, businesses will need to invest time and money into sourcing the best talent with the best skills for the job, or risk falling behind the competition.”
The AI skills gap will also cause universities to rethink their strategies. Dr Greg Benson, professor of computer science at the University of San Francisco and chief scientist at SnapLogic, predicts that machine learning will soon be required in computer science degrees.
“Until recently, machine learning and artificial intelligence were offered as elective courses in undergraduate computer science programs,” he says.
“In addition to the emergence of data science Bachelor’s and Master’s degrees, the core computer science curricula will require students to take machine learning as a compulsory course. In addition, the introductory computer science courses will include machine learning topics as early examples and projects.”
Deep learning and Python
Ben Lorica, Chief Data Scientist at O’Reilly Media, expects deep learning to gain early traction to supplement existing machine learning applications.
“Aside from new systems that use vision and speech technologies, we expect early forays into deep learning and reinforcement learning will be in areas where companies already have data and machine learning in place,” he says.
“For example, companies are infusing their systems for temporal and geospatial data with deep learning, resulting in scalable and more accurate hybrid systems -i.e., systems that combine deep learning with other machine learning methods.”
He adds that specialized hardware will be produced to make the technique more accessible, but predicts that hybrid models will be more commonly deployed.
“The resurgence in deep learning began around 2011 with record-setting models in speech and computer vision. Today, there is certainly enough scale to justify specialized hardware – Facebook alone makes trillions of predictions per day. Google has also had enough scale to justify producing its own specialized hardware. It has been using tensor processing units in its cloud since last year.
“Therefore, 2019 should see a broader selection of specialized hardware begin to appear. Numerous companies and startups in China and the US have been working on hardware that targets model building and inference, both in the data centre and on edge devices.”
Exosel CTO Golombek believes that Python will emerge as the leading data science language.
“In 2019, the variety of data science languages will continue to grow. But there is also a clear trend suggesting that Python will become the leading language for machine learning, and Python-based technology, such as the deep learning library TensorFlow, will continue to proliferate,” he says.
Advice for enterprises
Enterprises taking early steps into AI may find the biggest benefits in augmented analytics, which can be embedded in enterprise applications to provide automated insights and reduce the need for professional data scientists.
More advanced capabilities will require a bigger commitment.
“Think of AI as part of your human capital, not a replacement. The time, money and effort to train AI is as intensive as a human employee,” says Forrester analyst Goetz.
“AI is not something students will come out of school knowing what to do even if they have had AI coursework. Employees today are learning by trial and error by just jumping in. There are best practices and disciplines emerging in enterprise AI endeavours that demonstrate AI needs to be developed the way we develop our human talent in the workplace.
“Firms that look for the qualities in employees and potential employees are filling their AI workforce faster and putting the time in, training in, and career path development in to become a business that benefits from AI competitively.”