The explosion of both performance and hype for automated (or artificially intelligent) systems over the last several years has been mostly due to successes in a sub-discipline of AI – machine learning. While these systems are nowhere near some form of general artificial intelligence, they have proven to be incredibly effective at specific tasks – as long as they are fed enough examples.
At their core, these are systems that are good at predicting the outcome of novel events based on past events, with the strength of being able to process huge amounts of data to find connections. As we generate and collect more and more data by interacting with technology and the internet, machine learning has fast become integral to many of the services we use every day.
Although it’s often a surprise to many, AI is widely used by big publishers – and has been since 2012. Platforms like Reuters, the New York Times, and the BBC all use automation at various stages of the writing process. AI systems streamline workflows by scouring the internet to aggregate information, find breaking news and fact-check. AI is great at visualising data and writing effective summaries.
But it’s not limited to supporting writers. AI tools are commonly used to write entire articles, particularly in data-centric areas like sports and finance, and have been used for years by platforms like The Washington Post and Yahoo. These systems can be trained to write in a variety of styles, and have found use from news coverage to reports by banks and investment firms.
The internet is a big place, and sifting through it all to show you the content you want has become a primary mission for many digital services. This is where recommender systems come in. These AI tools power almost every facet of our online experience – from which posts Instagram shows you, to your Netflix home page, to your Google search results. If you’re seeing something online, it’s likely been delivered specifically to you by AI.
Importantly, it is these exact systems which enable the real business model of online platforms – advertising. AI models are fed incredible amounts of data of people’s habits across the web – from search history to shopping habits – which allow platforms like YouTube and Facebook to sell ads targeted specifically to you.
AI powers large parts of digital maps services from both Google and Apple. These systems fuse street view, aerial, and satellite images into the feature-rich and constantly updated maps we end up using. Google even uses machine vision to update business hours by looking through street view images to try and find signs in shop windows!
Machine learning-based vision has been widely adopted for industrial processes and increasingly in the agricultural sector. AI gives assembly lines the ability to automatically assess quality and detect errors, and allows robots to pick ripe fruit from trees. Image recognition is used to identify crop pests and automate the laborious task of labelling wildlife camera data for tracking and conservation efforts.
AI has been behind almost all translation systems in past years, which have come a long way despite being nowhere near perfect. While these models typically need to be trained on large amounts of text that has already been translated, recent advances have even enabled translation for languages that don’t have a wealth of bilingual data. Earlier in 2022, this allowed Google Translate to add support for both Tsonga and Sepedi.
Leaps and bounds in using AI to process natural language have lead to the widespread use of chatbots for customer service. Machine learning is also what allows for modern voice recognition on your phone, and generating artificial speech for personal assistants like Alexa and Siri.
AI is also heavily integrated into various levels of security systems. Your email service uses automated systems to block spam, viruses and phishing attempts. Your bank very likely uses AI to verify your transactions as you make them – to detect which ones are unusual for you and likely to be fraud. Similarly, modern biometric systems, like the ones you may have on your smartphone, use AI for face recognition and fingerprint verification.
Enabled by the rise of Big Data, machine learning systems have become powerful tools both for efficiently automating tasks previously done by humans, and using large amounts of information to perform completely new tasks and analyses. These systems continue to improve, and with powerful financial incentives for automation, AI will undoubtedly compute its way into ever more parts of our lives.
IOL Tech