Startups have developed tech that can accurately predict the future. While it’s easy to see the benefits, innovators still have to face a slew of both technical and ethical challenges
Don’t be alarmed but Facebook may know when you’re going to die. That’s one of the things you learn by looking through the thousands of patents the social media giant has filed since going public in 2012. For instance, one idea is to use your devices’ cameras to see if someone else is using your laptop or smartphone and then estimate how close your relationship is with that person. A second patent utilises the devices’ microphones to listen in and find out what shows you like to watch. A third one looks at things like your posts, messages, location and credit card transactions to determine whether or not you’re heading towards a life-changing event. You know, like dying.
While filing a patent doesn’t necessarily mean a company is using an innovation, this list points towards a bigger trend where more tech firms are developing technology to essentially predict the future. “Multiple companies are now using predictive analytics every day,” says Jean Ortiz-Perez, head of analytics for insurance and assistance at Collinson, the global loyalty and benefits company. For example, the British startup Black Swan has developed an algorithm that can estimate when the next flu epidemic will strike. In the past, it also accurately predicted the success of the movie Frozen, empowering Disney to up its production and match the demand.
However not all predictive technology solutions are that flashy – price-comparison sites use artificial intelligence (AI) to provide better quotes and your smartphone comes complete with autocorrect services. But no matter the industry, there are plenty of companies leveraging this new tech. “Machine learning and AI is now being used more frequently and I feel we are every day [getting] closer to enabling the customer to see the benefits of this [increased use of the] technology – not only reflected in price but also choices, recommendation and added benefits,” says Ortiz-Perez.
From estimating the value of stocks to predicting new trends, it’s clear this technology can have huge benefits for firms of all sizes. For instance, it can help boost companies’ interaction with clients. This is especially important given the former Mercedes-Benz CEO Steve Cannon’s assessment that “customer experience is the new marketing.” He has a point. “It’s not just a nice soundbite,” says Dan Somers, CEO of Warwick Analytics, the machine-learning company. Given how dissatisfied customers are more likely to swap suppliers and vent their critiques from the rooftops, businesses must consider their clients’ opinions more than ever. Predictive tech can improve this relationship in a number of ways. For one, it can help estimate when devices will break down and give customers a fair warning.
Another use could be to scan the online chatter about your devices and see whether people are likely to abandon your company for a competitor. Moreover, it also allows you to estimate how much you can boost your customer satisfaction by making small product changes. “You need to get inside the customer’s head, not your creator’s or your engineer’s head because, ultimately, the customer pays the wages,” says Somers. And that’s where predictive tech can be a huge help.
Given these benefits, it’s easy to see why investors are flocking to support predictive tech startups. “This space is very hot,” says Somers. For instance, Uptake, an American scaleup able to predict when machines will break down, raised a massive $117m series D round in November 2017, pushing its valuation above $2.3bn. In Britain, startups Black Swan and SwiftKey, the autocorrect firm, have both leveraged this interest to raise multi-million pound rounds. Not bad, considering the fact many founders start off working from their own dinner tables. “There are a lot of garages that have created millionaires and in this space I think there is room to make a few more,” says Somers.
But predictive technology is in itself nothing new. “Everybody is telling me that this is very new but I don’t think so,” says Daniel Wajngarten, partner at Data Reply, the data consultancy. He’s not wrong. Meteorologists have used predictive analytics for decades to estimate the weather and during the Second World War the scientists working on the Manhattan Project ran computer simulations to predict nuclear chain reactions. While these algorithms have clearly been around for a long time, the potential of them only recently came into fruition thanks to the leaps made in technology. Not only can modern computers process data faster but doing so has become a lot cheaper over the past 50 years. For instance, the price to store one gigabyte of data has gone down radically from $1m in 1967 to about two cents in 2017. “These factors in place [has] combined [made] this industry pop up and explode,” continues Wajngarten.
These giant leaps forward have been crucial to the launch of modern predictive tech startups. While none of them would exist without these storage and processing capabilities, these innovations still leaves them with the challenge of getting high-quality data. “You’d be amazed how hard it is to get access to the data,” says Wajngarten. Basically, entrepreneurs in this field have two options. The first is to collect all data themselves, which is a time-consuming and costly task. The second option is to use the data from clients or the public space. However, if founders opt for the latter, they still have to spend considerable efforts to mould it into something they can use. “The data is the hardest part of it,” says Wajngarten.
Another common misconception is that this technology is all-seeing and all-powerful and there are multiple examples that it isn’t. “Don’t believe the BS about predictive analytics,” says Somers. This has become apparent in a series of highly publicised cases where predictive tech has been used by police around the world. For instance, several American police departments have repeatedly been criticised for using algorithms based on data with a clear bias against people from ethnic minorities. In the UK, similar criticisms have been raised against police in Kent using historical data to classify suspects as a low, medium or high risk of breaking the law depending on things like where they lived. After facing extensive criticism of how the technology was used, Durham Constabulary began revising the data to rid itself from the biases within. Again, this points to the problem many startups in this sector face: the challenge of souring good data. “Garbage in equals garbage out,” sighs Somers.
And if you ask Gareth Edwards, senior consultant and data science specialist at Softwire, the bespoke software-development company, startups in this field have to make sure they pay close attention to how their tech is being used. “Although there are many benefits to machine learning it should be remembered that such systems cannot provide accountability for the choices made and may be biased in unacceptable or even illegal ways,” he says. Pointing at how the General Data Protection Regulation requires companies to be fair and transparent about how they process personal data, Edwards thinks it likely that there’s a trend of more legislation in this area. “Businesses should be looking towards machine-learning technology to help them enhance their business but all the while remembering that currently there’s still an ethical need for human-based decision-making,” he says. In other words, just because you’ve got the tech, it doesn’t remove your responsibility for how it’s being used.
Clearly there are huge risks with developing predictive tech but given there’s also great benefit and massive interest from investors, we forecast big rewards for innovative entrepreneurs that can walk this line with the right amount of care.