By Jon Pienaar
In an interview with Bloomberg two years ago, Elon Musk, South African-born Tesla CEO and billionaire founder of SpaceX, likened building artificial intelligence to “summoning the demon”. And globally acclaimed physicist Stephen Hawking has warned that it could “spell the end of the human race”.
But futurist, inventor and computer scientist Ray Kurzweilsays the potential benefits of AI far outstrip the perils. In Time magazine, Kurzweil writes: “We have a moral imperative to realize this promise while controlling the peril. It won’t be the first time we’ve succeeded in doing this.”
AI isn’t just the stuff of tomorrow, or science fiction anymore – it is likely you’ve already interacted with smart machines. If you’ve used Siri on iPhone, or OK Google on Android, then you’ve interacted with a learning machine. Similarly, spam filters are intelligent software that sorts through mail to figure out what is junk. These programs learn from user behaviour all the time.
It is a fundamental revolution for humankind. As Andrew Ng, chief scientist at Chinese web giant Baidu puts it, “AI is the new electricity. Just as electricity, about a hundred years ago, transformed industry after industry, I think that AI is now in a position to have a similarly large impact on society.”
ICT giants like Microsoft, Twitter, Intel, Salesforce and Apple are investing heavily in AI technology. Since 2011, according to research firm CB Insights, nearly 140 companies that have developed AI technology have been acquired as part of a race between these global corporations to own this space – 40 of these acquisitions took place in 2016.
AI is hard at work across most sectors. Financial institutions use software to monitor market movements, while security forces use machine learning in facial recognition systems that fight terrorism. In the medical field, AI examines vast datasets to factor in genomic, phenotypic and social aspects to assist doctors with making diagnoses.
Not a robot
“When you tell people you are working on AI, they assume you are building a robot that is going to take your kids to school,” Dale Humby, chief technology officer at Nomanini, says with a chuckle. Humby is at the coalface of AI development in the field of embedded systems and micro-payments.
He explains that commercial machine learning works with algorithms to perform specific tasks. “Like predicting what songs you might want to listen to next on Spotify, or which movie you might next find interesting
on YouTube.”
Nomanini’s clients are all prepaid services vendors, who sell airtime and electricity tokens directly to the public. These vendors have to purchase stock in advance, so the Nomanini system can advise them how much to purchase at certain times of the month.
“We also advise them where the hotspots are, or where they should stand to sell the most airtime or electricity, based on learning where they live,” says Humby.
AI enables Nomanini to make personalised decisions for these vendors based on large sets of data, explains Humby. “It’s only through using computers and machine learning that we can do this in practice. There is not enough time or money to have a whole team of analysts do this manually.”
Computers deciding for you
Humby contrasts social-media-mediated purchasing to the typical supermarket experience, where one is presented with a specific set of items to purchase, reinforced by advertising and product placement. “When you go on your Facebook, it is incredibly personalised,” he says, explaining that the Facebook back-end computers are making decisions “based on things you are going to find interesting or useful”.
“For business this helps with client engagement – particularly if you are a consumer-based web store like takealot.com.”
Humby says that if the AI engine shows you things that are specific to you, there is a higher likelihood that you are going to buy it.
“It knows your demographic – for example, if you are a new mom it presents you with baby merchandise and information, rather than chainsaws.” The more data an AI system has about customers, the better it can predict their purchasing needs. “Generally you are predicting what someone’s thinking, and suggesting what they might need before they remember that they need it,” he explains.
In the mathematical world of investing, where fast, accurate analysis is key, AI is making great progress. Magda Wierzycka is the CEO of Sygnia Asset Management, which launched a “RoboAdvisor” in May this year. The RoboAdvisor is a computer program that was installed as a back-end to a website that asks clients to input a range of information about themselves. It then provides investment advice based on an ever-fluctuating financial environment.
“Financial planning is quantitative in nature, so there shouldn’t be any emotive judgment applied to the information given to you,” says Wierzycka. “It should be a quantitative mathematical overlay that gets supplied to the information provider that comes up with an optimal financial plan for an individual.
“Based on the client’s inputs, and the financial planning model that underpins the website, the system calculates your net worth and projects what you will require in order to be able to retire without taking a cut in living standards,” Wierzycka explains.
Can a RoboAdvisor do better than a human financial planner? On a basic level, Wierzycka says, it does a great job. But when it comes to issues like death and disability, the human touch is still needed. Incremental improvements are taking place all the time, and Wierzycka says that in 10 to 15 years, AI-based financial advisers will be the norm.
Wierzycka says the financial services industry is an environment most conducive to the use of AI because machine learning allows for business functions to be done faster and cheaper.
In the UK, the Financial Conduct Authority has encouraged people to make use of AI tools to save costs.
Intelligent adviser
Ryan Falkenberg, the co-founder and joint CEO of Stellenbosch-based AI start-up Clevva, and his team have been working on an AI platform that is an intelligent “virtual adviser”. The machine learning being developed by Clevva aims to assist people where human error or boredom is a factor that could impact on the quality of service provision.
“Our view is, how do we use AI to make people stronger or more effective, so that they still retain relevance – as opposed to excluding them entirely? How do we help companies leverage AI, but in a way that gets more out of people?” asks Falkenberg.
“Most training and knowledge management is aimed at capturing process logic or sales logic,” he says. “In contact centres, this becomes formulaic.” The net effect is that these systems require people to be trained to behave like robots.
But, he points out, the human brain is not good at memorising formulae and vast tracts of information, especially over the long term. Trainees use their short-term memory to pass the course, but generally only 20% of what is learnt is retained.
The solution Clevva presents enables human beings to do what they are good at – engaging with the client, and putting emotional intelligence to work.
“Innovation, creativity, human engagement – these are the types of capabilities that differentiate human beings from technology. We are on a mission to make sure human beings stay strong and relevant by enabling them with decision logic. Together humans and computers add a value that AI and computers alone would struggle to add,” he explains.
AI is often seen as a threat to unskilled or undereducated people, but Falkenberg says an unskilled person who has good emotional intelligence, a great attitude and the right aptitude, can be teamed up with a decision-making machine that helps to solve customer queries or sell products, or even deliver services.
He has a high-road vision in which SA’s education deficits are augmented by AI, and where people get a better shot at university and careers through supplemental machine learning.
Let’s hope AI can help address SA’s failings and be an engine for economic growth
© Copyright 2017. Fin24