![Should You Fear AI in Finance?](https://sp-ao.shortpixel.ai/client/to_webp,q_glossy,ret_img,w_851,h_315/https://emerj.com/wp-content/uploads/2018/04/AIFinance.jpg)
Science Magazine’s report on Friday that an artificial intelligence system was caught stealing banking customers’ money may have made you rethink vesting your funds in the burgeoning technology. But have no fear – the article was an April Fool’s joke.
While there are still valid and plenty of concerns about AI, the threat that the so called “Deep Learning Interface for Accounting (DELIA)” may embezzle your funds is not one of them.
In fact, AI systems are a remarkably valuable resource in finance. Established banks and startups alike implement AI to detect fraud. Financial management firms use AI algorithms to build and manage customers’ portfolios. AI even allegedly controls most of the activity on Wall Street.
If Science Magazine’s gag had you scrambling to get your funds into human hands, let the following positive examples reestablish that lost confidence.
Fraud detection
If you’ve ever been contacted by your bank concerning a questionable transaction, the tip may well have come from an AI algorithm. In a process called data mining, these AI are capable of combing through troves of data to identify patterns and aberrations that might signify fraudulent activity.
Fraudsters leave “footprints” that can help systems recognize their maneuvers. For example, certain countries and times of day raise the risk that a transaction might be fraudulent. Discordant shipping, billing, and IP addresses and perpetual attempts with slightly different account numbers raise red flags that cause the AI to mark the transaction as questionable. From this data, AI develop a “risk score” that depends on the needs and preferences of a particular institution.
Though the system eventually functions independently, it’s initial training is guided by humans. Many of these systems are taught to recognize potential fraud through supervised training, in which random samples are manually classified as fraudulent or non-fraudulent. The algorithm then trains from these manual classifications to determine the legitimacy of future activities on its own.
We covered an interesting case study of AI for money laundering detection with Ayasdi which might be of interest to readers with a focus on fraud.
Insurance applications
The insurance industry is – in many ways – the original “big data” industry, and it might be ripe for AI disruption.
One of the more interesting recent applications of AI in insurance involves a combination of AI and IoT. Called “Snapshot”, Progressive’s latest IoT device sits in a driver’s glove box and detects a variety of vehicle data, including:
- Speed of acceleration
- Speed of deceleration
- Aggressiveness of turns
- Distances traveled
Using this data, Progressive is able to assess a more accurate insurance risk for that particular user. So, a 25-year-old male with a Porsche might be able to reduce his relatively high rates (given his vehicle and demographics) by tangibly proving his safe and responsible driving skills on the road in real time.
This follows a greater theme in insurance: Drinking in more individual data to make better individual predictions of risk. In many ways this mirrors the value of machine learning in finance in general.
Our full article on AI in insurance covers these applications in greater depth, including some of the other AI initiatives at large firms like Progressive, State Farm, and others.
Account security
A number of financial technology startups have also turned to AI to supplement and support their services.
In Hong Kong, mobile banking startup Neat plans to apply machine learning technology to analyze users’ payment data. According to the South China Morning Post, the feature will detail behavioral trends and recommend local offers to users based on their activity. Interestingly, the company calls itself “an artificial intelligence company doing banking, not the other way around.”
Meanwhile in Berlin, mobile banking startup Number26 has considered using AI in a different way. Self-billed “Europe’s most modern bank”, the company already uses AI similarly to Neat, by monitoring user payment activity, offering behavioral insights and recommendations. But Number26 is also considering applying an AI algorithm to detect unauthorized card use as well.
Stock market management and financial planning
According to the BBC, machines currently manage most of Wall Street’s activity. Trading times have been diminished from seconds to milliseconds.
Wealth management firms such as Charles Schwab and United Bank of Scotland have begun to advise their clients via AI algorithms. Schwab Intelligent Portfolios’s customers first fill out a questionnaire to identify their investment aims and aversion to risk. With this information, Schwab hands control of accounts over to an algorithm that “builds, monitors, and rebalances” a customer’s portfolio, making real-time decisions about where and how to invest money.
But computers aren’t infallible – a computer glitch in 2012 saw Knight Capital lose over $400 million in just half an hour. To mitigate such computer error, US financial regulators require trading to stop if stock prices fluctuate more than 10 percent in fives minutes.
Meanwhile, Charles Schwab implements human checkpoints to monitor significant changes. According to the BBC, when the Dow suddenly lost 1,000 points in August of last year, Schwab brought together a group of humans to determine whether Schwab Intelligent Portfolios was making wise investment decisions.
AI aren’t perfect, but they serve a vital purpose in the financial industry. Science Magazine’s April Fool’s gag may have been in good humor, but it revealed our pragmatic sensitivity to letting AI have too much free reign over our finances.
Credits: Blutegruppe/Corbis