Computing moves to the pocket and the cloud
The 2010s were less about new kinds of machines than about where computing happened. For a growing number of people, the main computer was a phone, and much of the work behind apps ran on servers in large data centers operated by a few big companies. At the same time, a branch of artificial intelligence called machine learning produced results that surprised many researchers.
Mobile apps
By the early 2010s smartphones from Apple and from makers using Google’s Android system were mainstream in many countries. Tablets joined them, with Apple’s iPad introduced in 2010. App stores let developers distribute software directly to hundreds of millions of devices, and new services were built around phone features such as cameras, location sensors and constant connectivity. Messaging, ride-hailing, mobile payments, photo sharing and navigation became daily tools for many people. Faster cellular networks, including the fourth-generation standards, made video and music streaming on phones practical.
Touchscreens became the default input method for many users, and desktop and laptop computers were used less for casual tasks. Sales of traditional PCs flattened or declined in many markets, though they remained important for work, gaming and creative tasks. The Raspberry Pi, a tiny inexpensive computer released in 2012, also found an audience among students and hobbyists.
Cloud services
Cloud computing means renting computing power, storage and software over the internet instead of owning and managing equipment. Amazon had begun offering such services in the middle of the 2000s, and during the 2010s Microsoft, Google and others built large competing offerings. Businesses moved email, file storage, databases and entire applications to the cloud, in part because they could scale up or down as needed. Consumers used the same infrastructure through photo backup, document editing and streaming video.
The shift brought trade-offs. It reduced the need for small organizations to run their own servers, but it concentrated a great deal of the internet’s infrastructure in a small number of companies. Outages at a major provider could affect many unrelated services at once. Questions about where data is stored, and under what country’s laws, became more important.
Machine learning and neural networks
Researchers had worked on neural networks, systems loosely inspired by connections among brain cells, for decades, with periods of enthusiasm and disappointment. In the 2010s, a combination of much larger data sets, faster graphics processors repurposed for training, and improved techniques gave a method called deep learning striking results. A 2012 image-recognition contest result is often cited as a turning point. Speech recognition, translation, image tagging and recommendation systems improved noticeably, and voice assistants and smart speakers appeared in homes.
In 2016, a program called AlphaGo defeated a top professional player at the board game Go, a result many experts had expected to be years away. In 2017, researchers published the Transformer design, which later became the basis of many large language models. At the end of the decade, these models were still mostly a research topic for the general public, although the foundation for the 2020s had been laid.
Progress was real but uneven. Systems that did well on narrow tasks could fail in unexpected ways, and critics pointed out problems with bias in training data, difficulty explaining how models reached decisions, and heavy energy use for large training runs.
Social platforms, privacy and security
Social media use grew further, and platforms became central to news and politics in many countries. Disputes over misinformation, harassment and the collection of personal data produced public hearings, new rules and ongoing debate. In 2013, documents leaked by a former U.S. intelligence contractor revealed extensive government surveillance programs and prompted wider discussion of encryption. The European Union adopted the General Data Protection Regulation in 2016, which took effect in 2018 and influenced how companies handle personal data worldwide.
Security flaws also made headlines. The Heartbleed bug in a widely used encryption library was disclosed in 2014. Ransomware attacks, which lock files until payment is made, hit hospitals, businesses and city governments. In 2018, researchers disclosed the Meltdown and Spectre hardware vulnerabilities, which affected many processors and showed that long-standing design choices could carry hidden risks.
Hardware trends
Solid-state drives, built on flash memory, became common in laptops and phones, making them faster and more resistant to shocks. Screens grew sharper, and many devices adopted high-resolution displays. Chip makers found it harder and costlier to shrink transistors at the old pace, and discussion of the end of Moore’s Law became common, with more attention to specialized chips for graphics and machine learning rather than only faster general-purpose processors. Interest also grew in the Internet of Things, in which everyday objects such as thermostats and cameras connect to the network, bringing both convenience and new security concerns.
Cryptocurrencies attracted a mixture of investment, speculation and experimentation, though their long-term role remained unsettled by the end of the decade.
What the decade changed
By 2019, computing was less visible as a box on a desk and more a service woven through daily life. The same developments that made it convenient also raised questions about privacy, power and dependence that the next decade would face.