What will happen over the next year in machine learning and AI innovation? Today, we will share five AI predictions of data scientists and tech experts.
Commercial AI will go mainstream
The trope “there’s an app for that” is quickly becoming “there’s an AI for that.”
Over the next year, AI services and technology that has been around for severals years will shift from the “early adopter phase” to “general adoption phase.” We’ll see an increase in popular awareness about AI as a utility, just like the Internet and Wi-Fi.
In 2019, AI will become a standard feature of mobile phones, smartphones, and smartphone apps. Your iPhone is already programmed to use AI and machine learning for several features such as face ID, Siri, and location services. This year marked the beginning of a new era of AI consumer-based apps like TikTok, a short-form mobile video social app.
Voice assistants for cars
AI assistants like Siri, Alexa, and Cortana will become available in your car. Voice assistants in the home have already taken off in popularity among consumers in 2018. The average person on the street might not know about natural language processing, but they definitely have at least heard of Siri and Alexa.
In 2019, new car models will come with built-in voice assistants. Toyota has already began integrating Amazon Alexa in new Toyota and Lexus models. BMW, Mercedes-Benz, and Ford have also began implementing voice assistants in their newest models. Next time, you can ask Alexa for directions or parking information without whipping out your phone. Conveniently, data scientists are also designing the car voice assistants to sync together with the voice assistants that you might already own at home.
Small-scale economic disruption
Automation in 2019 will not result in mass unemployment overnight. But if say, 10~20% of jobs are lost to automation, even that is enough to cause a disruption in the job market and general economy. For example, many workers drive a vehicle for a living, and they might be one of the first groups to lose their current jobs to automation.
As AI innovation improves, businesses should start thinking about the pros and cons of automate tasks using technology. The most obvious advantage of automation is cutting costs. The cost of employing and managing humans might soon be higher than the cost of implementing and maintaining technology, for many industries.
But for some industries, automation might still be risky to the point of being unethical. For example, although AI innovation in the medical field is advancing at great lengths, we still need humans to double check the results of an algorithm that diagnoses patients with cancer. Similarly in law, some lawyers and judges are using machine learning algorithms to help predict trial outcomes or determine prison sentences for criminal defendants. These algorithms, however, are far from replacing lawyers and judges.
Machine learning will become pervasive to most industries
Across most industries, artificial intelligence is being used to automate a wide range of tasks in various industries, for example:
- Finance and banking: According to an Accenture report, more than half of Fortune 500 companies have gone out of business since 2000, and AI is set to take this disruption to a new level. Yet, the same report also suggested that if financial institutions invest enough in AI, they can expect potential savings of between 20~25% across IT operations. For example, financial fraud detection systems used to depend heavily on a complex set of rules, but machine learning systems can more quickly and accurately detect unusual activity to flag for security teams. AI can identify location, transaction anomalies, verify customer place of business, and flag sensitive cross-border movement.
- Ecommerce: Chatbots can be embedded in websites such as online stores, or a through a third-party messaging platform like Facebook messenger, and Twitter and Instagram’s direct messaging. Chatbots allow businesses to automate customer service. For businesses with a young customer base, chatbots are more likely to increase customer satisfaction. 60% of millennials have used chatbots, aond 70% of them reported positive experiences.
- Journalism: AI is proving to become one of the most useful tools in the newsroom, and has already driven data-led stories and made way for more investigative pieces. For example, The Washington Post used its robot reporter, Heliograf, on Election Day to cover congressional and gubernatorial races. In fact, Heliograf was used to publish 850 articles in one year.
Better access to large, high-quality datasets
Unfortunately, this last one is more of a hope than prediction for 2019. The shortage of AI training data is a tremendous ongoing blocker for future innovation. Data scientists need large amounts of high-quality AI training data to train machine learning algorithms. In general, increasing the quality and quantity of AI training data leads to faster and more accurate algorithm performance.
One of the reasons why data scientists don’t have enough data is because big companies invested in data collection often do not make their datasets publicly available. Instead, they hoard their data, perhaps due to privacy concerns or fear of losing to their competitors. For example, most social media platforms such as Facebook and Twitter published a developer policy that prohibits sharing archived posts.
We at Gengo hope that in 2019, these companies will make their datasets available to the public, so everyone can use it to build and improve machine learning technology. Data scientists would be quick to point out that data alone does not give any company an edge over the competition. In the meantime, we recommend checking these 12 open social media datasets and 50 open datasets for machine learning.
Building a large, high-quality dataset used to be a time-consuming and grueling task for data scientists. But now, the advent of cloud technology has made the task much easier. We at Gengo have also taken advantage of the recent tech advancements, to provide our AI training data service. AI training data must be first be prepared manually by humans. Gengo’s diverse crowd of over 21,000 contributors, ain all major time zones and almost every country, can quickly prepare large, custom AI training datasets for your machine learning projects. We also source qualified contributors for each project, and manage the entire process.