It’s been hard to miss Bill Gates, Elon Musk and Stephen Hawking talking about the dangers of artificial intelligence. They’re looking into the future and planning now for the threats they see on the horizon. It’s good to look ahead but what about this year? What disruptions will we see in 2015? I’ve been working with machine learning and narrow AI for the last several years but it feels like we’re at a tipping point for adoption and innovation. That’s the recipe for disruption and here’s what I see coming this year.
Open Source Is Letting the AI Genie Out of the Bottle
Open source projects like Torch and Weka have put machine learning algorithms within reach of every business. Rather than having to reinvent the wheel, these open source projects have created an advanced set of machine learning and “narrow AI” tools. That’s a big leap forward for many small data science teams. They are able to focus on new areas of work rather than spending a year just implementing the basics.
That means the pace of advancement will grow from what we saw in 2014 to what we’ll see this year. As you know, 2014 was already a massive year for machine learning with almost all large companies building mature data science capabilities. This year will be more about what startups and small businesses can do with machine learning and narrow AI.
The paradigm shift will be from highly structured business initiatives to highly innovative products. By the end of this year, several machine learning startups will have established themselves as major players in the technology space. Before open source opened the doors to machine learning, there was too high of a barrier to entry for innovative companies. Now that open source has leveled the playing field, that’s going to change.
Automation and AI
Automation is limited by the intelligence of the software running it. With machine learning and narrow AI, traditional limits of automation are being pushed aside. Cost is now the limiting factor in automating everything from driving to peoples’ jobs. As I mentioned, open source software is decreasing the cost of creating a machine learning application. That means this year we will start seeing the rise of automation.
Voice recognition is an area we’ll see rapid expansion in the machine learning and automation front. As the “intelligence” of systems and accuracy of voice recognition increases, the overall quality of automated voice systems, like the ones already in use for customer service, will improve drastically. The cost of these systems is being weighed against the cost of staffing large call centers with people. That equation is quickly tipping in favor of automation.
Call center employees aren’t the only jobs under pressure. Software developers will be looking at automation taking over some of their jobs too. There are a few startups working on automating programming at a much more sophisticated level than ever before. They’re training machine learning algorithms to program by sifting through massive amounts of open source code. There is also work being done on software test and documentation creation algorithms. Most are aiming to make programmers’ jobs easier, but replacing low level programmers is not far off and we might even see it this year.
Call centers and software developers are just two examples of the types of jobs automation is looking to replace. McDonalds is trying out kiosks in a few locations. Amazon is using robots extensively in their fulfilment centers. This year those trials will go mainstream and a lot of other companies will be jumping on the bandwagon once they see the cost benefits.
Privacy is about to get a major disruption. In the US and around the world, governments are waking up to the scope of data collection going on in business. They’re publishing reports on data, privacy and security. That’s usually a harbinger for regulation to come.
The problem is that most governments aren’t informed enough to create effective guidelines for data gathering, use and protection. That means the majority of regulations are going to be cumbersome while providing little actual protection to individual privacy. Businesses will be looking at yet another layer of compliance and regulatory oversight. However, after checking the compliance box they will continue to have virtually limitless access to and use of personal data in 2015.
If consumers thought ads and recommendations were getting creepy in 2014, just wait until the internet of things data starts coming in. The IoT provides a depth of data on individuals that’s going to allow for very precise targeting. It will allow insurance companies, lenders and many other businesses to make use of personal data to make their risk calculations. Drive in high accident zones? Your smart car’s data will likely be sold to insurance providers. Make a lot of purchases at what are considered “high risk” establishments? That data will be sold to banks that will decrease your credit score based on buying behavior.
We understand now that a lot of data is being collected about us. The disruption will come when that data collection start to impact our lives. Everything from customer loyalty systems to how much we pay for health insurance will change based on data profiling our behaviors. In the US, police departments are using machine learning algorithms to profile crime and criminals. Right now the scope of this effort is to understand how best to police high crime areas to deter criminal activity. If these programs prove effective, the use of personal data and narrow AI will increase. An algorithm to predict a person’s proclivity towards crime isn’t far off.
The next massive breach is also not far off. As more data is collected, the personal impact of a data breach will grow. Looking at the celebrity photo leaks of a few months ago it’s easy to see how much damage stolen data can do. There will be a massive data breach in 2015. The scope of the damage from that breach will determine how much better information security gets this year.
How people decide to use machine learning and narrow AI will determine a lot about privacy this year. If ethical, informed voices prevail it’ll be a good thing for almost everyone. If they don’t, we’ll be looking at abuse and an eventual backlash. People will be quick to act once an algorithm gets them wrongfully arrested or denied auto insurance.
The Good and Bad
As with any new technology change is inevitable. There will be upsides and negative consequences. While some very smart people are starting to plan for our long term artificial intelligence future, the rest of us need to look at the nearer term. Machine learning has great potential to make positive improvements but we’re going to see some downsides start to emerge this year too.
The key to success during our machine learning infancy is to remember there are people behind all those data points. We’re not collecting data on anonymized customers. There are neighbors, friends and family in our databases. The narrow AI algorithms looking for signals in all that data don’t understand what that means. Uber raising fares during a crisis reminds us that machine learning algorithms aren’t ready to sit at the top of a decision pyramid. We still need people for that. If we take the human element out of the equation, it’ll result in a lot more bad disruptions than good.
As practitioners of data science, we have to look at this year as a proof of concept for narrow AI and machine learning for businesses and people. We all want to jump from disruption to disruption as quickly as possible which is great. We can’t do that in an irresponsible way. Failures will shake confidence which could lead to a slowdown in advancement. That not a good scenario for anyone.
Vineet Vashishta is the founder of V-Squared Consulting, a leading edge data science services company. He has spent the last 20 years in retail/eComm, gaming, hospitality, and finance building the teams,infrastructure and capabilities behind some of the most advanced analytics companies in the US.
You can follow him on Twitter: @V_Vashishta.