Speaking at the DATAVERSITY® Smart Data Online 2016 Conference, Adrian Bowles, industry analyst, recovering academic, and founder of STORM Insights, and Steve Ardire, Merchant of Light, both talked about the future of Artificial Intelligence (AI). They addressed the question: “Will robots take my job?” The short answer could be summarized as: “Maybe.” Lower level jobs with easily improved productivity are likely candidates for automated replacements, says Bowles, but higher level positions with decision making capacity are more likely to have assistance from AI, rather than being replaced by it.

What Machines Can Do: Recent Developments

Bowles talked about recently going through the stack of books he used early in his AI studies and discovering that “nowhere in these books is the key phrase ‘Machine Learning,’ and yet today, that’s sort of at the center of everything we’re doing.”

Ardire adds, “Machine Learning [as opposed to Machine Intelligence], is about getting computers to learn how to program themselves, letting the data do the work with the training set.” Most learning now is supervised learning, using reinforcements for correct or incorrect answers, and using neural networks for Deep Learning, but he predicts the emphasis in the future will be on unsupervised learning, where machines can infer what they know or don’t know and are given no positive or negative reinforcement.

Although there is still a place for supervised learning:

“When you’re dealing with unsupervised learning, that’s a system using Machine Learning to discover patterns based on experience,” says Bowles. “And there are a lot of different kinds of applications – it’s not that one is going to replace the other, different applications are going to use supervised, and unsupervised learning.”

Ardire says most of the current uses of Artificial Intelligence now flooding the marketplace are based on the “What AI,” which includes machine vision, language pattern recognition, and “everything from IBM’s Watson to Google’s AlphaGo project,” used for assisted, augmented intelligence to help with automated tasks. It’s “where most of the players are,” he says.

The “Why AI,” the causal, Ambient Intelligence, goes beyond the “what.” It exceeds the capacity of the current reasoning. He says it’s much more granular:

“It’s personalized, it’s situational, it discerns contacts and makes inferences, it’s self-optimizing, and it can learn without supervision, so this is actually the means to an end that has the greatest potential to change the way industries and businesses operate and compete.”

He cited recent MIT work with the neural network, designed to anticipate human reactions correctly. At 43% success, compared to 71% success by humans, he says that researchers believe it could surpass the human ability to predict if given enough time.

Progress has also been made in the area of Natural Language Processing, although there’s been discussion in the community about “whether or not we’re really doing ‘real’ natural language processing,” says Bowles.

“Most of you have probably seen the Watson ads where Watson analyzes Bob Dylan’s lyrics and summarizes them, you know: ‘love lost’ and ‘time passes,’ gross simplifications, but the interesting thing for me is that if you look at it and you don’t have the historical context – you weren’t listening to these songs in the 60s – that’s pretty much what a college freshman today would get out of the same lyrics. So the natural English processing today is at a level of sophistication that I would say is fairly comparable to an early stage college student.”

Will Artificial Intelligence Take All the Jobs?

Ardire presented a slide with figures on job loss due to Machine Intelligence:

• $15 billion industry today, growing to $70 billion by 2020
• 80% of executives say AI boosts productivity and will replace more than16% of jobs by 2020
• By 2018, smart machines and digital assistants will be able to recognize customers by face and voice across channels and partners in 30% of interactions (Gartner)
• By 2018, 45% of the fastest-growing companies will “employ” more smart machines and virtual assistants than people
• By 2020, 85% of customer interactions will be managed without a human
• By 2020, smart agents will manage 40% of mobile interactions
  By 2020, most enterprise relationships with customers won’t require humans

In the enterprise, systems of record with primarily static roles are being replaced by systems of intelligence that can learn and predict influence based on real-time data. Spending on AI companies increased four-fold in the last six years, he says, and “AI teams are being acquired for an average of $2.5 million per employee, an employee value that often far exceeds business value.” Companies are learning how to deliver a better customer experience using Artificial Intelligence and it’s creating a feeding frenzy among startups. Ardire quoted Kevin Kelly, from The Three Breakthroughs That Have Finally Unleashed AI on the World, in Wired Magazine: “The business plans of the next 10,000 startups are easy to forecast: Take X and add AI. This is a big deal, and now it’s here.”

With 37% of consumer apps now being driven by virtual aids, Ardire says one of the fastest growing areas is chatbots, which is expanding from automating routine tasks to more sophisticated uses. He cited Robot Lawyer as an example of a successful app that has assisted users in overturning160,000 parking tickets in London and New York, and he expects more of this type of application on the horizon. Ardire quoted Yann LeCun,  “If we know how to build dialogue systems that have an idea what the person dialoguing wants or thinks, that means we can have chatbots that are actually useful and interact with you in a natural way.” Conversational dialog systems have been missing from the picture, Ardire says, so now we can “actually have meaningful interaction, rather than just the automating path.”

The ability to “hyper-align” a product is also here now, and it’s possible to determine user intent and behavior, he says. Machine Language algorithms are being used for decision making, which can be used in a variety of settings, and he cited potential uses in law, to minimize or prevent litigation, or to potentially dispense with judges. A study of 32 cases was used to compare machine decision making with judicial decisions, and the machine correctly predicted the decision with a 96% success rate. The White House is investigating the use of Artificial Intelligence and Machine Learning programs to “solve issues of mass incarceration because US taxpayers pay $39 billion/year to jail 2.3 million people, making US incarceration rates the highest in the world.” These programs,

“Can improve screening processes, scan body camera footage for police misconduct, and make sentencing more fair. When I saw this, [I thought] it’s a huge problem and kudos for the White House.”

AI is being used to brew beer in the UK, He says, as an example of one its more innovative uses, and one company has an online ordering system for pizza using chatbots, with to-your-door-delivery by automated trucks.

Will AI Take My Job?

Pointing out that there is an upside to job automation, Ardire says,

“Yes, if your job is routine manual, or routine cognitive, you have a higher chance of being replaced, based just on the trend lines of where this is going. The benefit here is [that] machines will do things humans don’t want to do anyway.”

Paraphrasing Peter Drucker, he says, “Effectiveness should be a human pursuit, while efficiency should be delegated to machines. The benefit is all smarter decisions.” Ardire says that it’s more about,

“Amplifying human intelligence; it’s the employees’ passive knowledge – that experience and wisdom – that brings the value in a way that machines can’t. AI may take away some tasks but it brings new ones. We still want humans to make life and death decisions in areas like health care,” and emotions will be crucial to the relationship.

Bowles also sees the emergence of Augmented Intelligence as a useful tool for health care, serving as a co-worker with doctors. Watson, for example, is able to capture case data and studies in medical journals, cross-referencing symptoms and causes across more cases that any one doctor possibly could. By including multiple possible diagnoses ranked by confidence level, he says, the doctor can easily justify further testing based on logic, which affords an an opportunity for a smarter, more efficient diagnostic process.

“The contention here is that machines will emerge more as top collaborators for the non-routine cognitive area by using their predictive potential to augment our capabilities,” adds Ardire.

“I want to distinguish between just mimicking some functionality and actually having the understanding,” says Bowles:

“But we still have this fear, so let’s talk about the business end of things. In the beginning, we were talking about replacement, and that’s still the fear. A lot of people are looking at this as ‘the robot overlords are coming.’ [Yet] most of the high-level stuff is reinforcement, and as Steve pointed out, a lot of low-level jobs are going to go away, and some new ones will be created. I think the net loss in jobs is very real, but the net gain in productivity is probably more interesting for those of us who are working in the field.”

Robots are learning to do complex tasks just by watching humans doing them, but it’s not yet what most would consider true Artificial Intelligence. Rather than replacing workers, new technology will require that people gain new skills, says Ardire, “Whether or not it’s going to replace your job, that’s really kind of up to you at this point in terms of finding [the] unique value that you’re adding that can’t be replaced just by faster access to the data.”

 

Here is the video of the Smart Data Online 2016 Presentation:

 

Register for the Smart Data 2017 Conference Today (in San Francisco Bay Area Jan 30 – Feb 1, 2017)