2015: The Year Machine Learning Took Off
Introduction
The year 2015 marked a major turning point in the rise of machine learning from a niche research field to a mainstream technology poised to transform industries. As Google CEO Sundar Pichai remarked at the time, "Machine learning is a core, transformative way by which we‘re rethinking everything we‘re doing." From tech giants to startups to established companies across verticals, 2015 saw an unprecedented embrace of machine learning‘s potential to revolutionize products, services, and operations.
In this retrospective, we‘ll examine the key developments that made 2015 such a pivotal year for machine learning, and how the momentum generated then has propelled the field to even greater heights in the years since. Through an infographic timeline, discussion of major milestones, and analysis of machine learning‘s proliferation across industries, we‘ll paint a picture of a groundbreaking year that laid the foundation for an ongoing technological revolution.
Tech Giants All-In on Machine Learning
Among the most visible drivers of machine learning‘s explosive growth in 2015 were the major tech companies that dedicated massive resources to advancing the state of the art. Google, a pioneer in the field, signaled how central machine learning had become to its ambitions when it made major organizational changes to focus on AI, including promoting AI expert John Giannandrea to its core leadership team.
Google‘s DeepMind unit, acquired in 2014, made waves in 2015 with a landmark Nature paper detailing how its algorithms learned to play Atari video games at superhuman levels—a powerful illustration of machine learning‘s potential. And Google Brain, the company‘s deep learning research team, open-sourced its TensorFlow software library, kickstarting a revolution in the accessibility of machine learning tools.
Google was far from alone in its machine learning push. Microsoft, Amazon, Facebook, IBM, Baidu and others each made major investments in machine learning research and development in 2015, from creating dedicated AI labs to open-sourcing key technologies. The arms race to acquire top machine learning talent was already in full swing, with salaries for experts soaring and academics leaving universities in droves for industry positions.
A Bumper Crop of ML Startups
2015 also saw an explosion of machine learning startups aiming to harness the technology‘s power for specific applications and verticals. Over \$1 billion in venture capital was invested in AI startups in the first half of 2015 alone, and the year saw a raft of significant funding rounds, product launches, and acquisitions.
Some notable examples: Sentient Technologies, an AI company with roots in the hedge fund industry, raised a \$103.5 million Series C round in 2015 to further develop its distributed machine learning platform. Robotic process automation firm Automation Anywhere incorporatedt machine learning into its offerings and raised \$57 million. And personal assistant app Accompany, which used machine learning to scan online data to provide briefings on people and companies, launched in August 2015 and was acquired by Cisco less than three years later for \$270 million.
Other ML startups that made their mark in 2015 focused on areas like healthcare (Enlitic, Atomwise), fintech (Affirm, Numerai), marketing (Persado, Amplero), and beyond. For many, 2015 proved to be just the beginning—a launchpad for ongoing growth and success as machine learning‘s commercial applications proliferated.
Machine Learning Branches Out
The embrace of machine learning in 2015 extended far beyond the tech sector. Companies across a wide range of industries, from healthcare to finance to manufacturing, began exploring the technology‘s potential to transform their operations.
In the automotive world, Toyota announced a \$1 billion investment in AI and robotics R&D, including hiring researchers to staff new labs in Silicon Valley and MIT. Robotics firm Fanuc collaborated with Nvidia to develop machine learning techniques for training industrial robots. And a host of automakers and tech firms, from Audi to Baidu, made strides in autonomous driving powered by machine learning.
Healthcare saw a surge of interest in machine learning for applications like medical imaging analysis, drug discovery, and patient risk prediction. IBM Watson Health embarked on numerous collaborations with hospitals and pharma companies. Startups like Enlitic and Arterys developed ML-based tools for radiology, while others like Deep Genomics and Atomwise applied the technology to genomics and drug development.
In finance, 2015 brought an uptick in hedge funds launching machine learning-based trading strategies, with prominent funds like Point72 (now Cubist) and Bridgewater making significant investments in AI talent and technology. Machine learning also found applications in areas like fraud detection, credit scoring, and customer service.
And in fields like energy, manufacturing, and logistics, companies began to harness machine learning for predictive maintenance, yield optimization, demand forecasting, and more. GE‘s Predix industrial IoT platform, launched in 2015, aimed to bring machine learning to bear on data from connected machines to improve performance and efficiency.
Limitations and Challenges Emerge
For all the excitement and progress around machine learning in 2015, the year also brought a clearer understanding of the technology‘s limitations and challenges. Even as ML systems achieved human-level or superhuman performance on narrow tasks like image classification and game-playing, they remained brittle, opaque, and difficult to adapt to changing real-world conditions.
Concerns about bias and fairness in machine learning came to the fore, with examples like Google‘s photo tagging algorithm labeling black people as gorillas and a ProPublica investigation finding racial disparities in ML-based criminal risk assessment scores. The lack of diversity in the AI field also drew increasing scrutiny.
On the technical front, issues like data quality, algorithm interpretability, and the difficulty of reproducing results posed ongoing challenges. The cost and complexity of training cutting-edge ML models continued to rise, making it harder for smaller players to compete.
And even as machine learning proliferated across industries, many companies struggled to translate the technology into real business value. Gartner‘s 2015 Hype Cycle report placed machine learning at the "peak of inflated expectations," noting that it would still be 2-5 years before the technology reached mainstream adoption.
The Road Ahead
Looking back from the vantage point of 2023, it‘s clear that 2015 was indeed a pivotal year for machine learning—but in many ways, it was just the beginning. In the years since, the technology has continued to advance at a rapid clip, with breakthroughs like Google‘s AlphaFold protein folding model, OpenAI‘s GPT language models, and DeepMind‘s AlphaFold 2 and MuZero.
Machine learning has become integral to products and services we use every day, from voice assistants to recommendations to autonomous vehicles. It‘s also become a key competitive differentiator for businesses across industries, with the most successful firms often being those that have most effectively harnessed the technology.
At the same time, many of the challenges and limitations that emerged in 2015 are still with us today. Bias, transparency, and ethics remain major concerns as machine learning systems play ever-larger roles in high-stakes decisions. The environmental costs of large-scale ML model training have come under increasing scrutiny. And the concentration of cutting-edge ML capabilities among a handful of tech giants has raised alarms about the risks of centralizing such a powerful technology.
As we look ahead to the future of machine learning, the seeds planted in 2015 will continue to bear fruit—both the immense promise of the technology and the complex challenges surrounding it. Continued research and innovation in areas like transfer learning, federated learning, and AI safety give reason for optimism. But realizing the full potential of machine learning in a responsible and equitable way will require ongoing collaboration among researchers, practitioners, policymakers and society as a whole.
One thing is certain: the transformative impact of machine learning that became so apparent in 2015 is only just beginning. As computing power continues to increase, algorithms keep advancing, and more industries find innovative ways to harness the technology, machine learning‘s potential to reshape our world will only grow. Navigating that potential responsibly and inclusively will be one of the great challenges—and opportunities—of our time.
