The overall theme of Splunk’s four-part 2021 Predictions report is the rapid acceleration of digital transformation, driven by the specific event of the COVID-19 pandemic, and the momentum of data technologies that have brought us into a true Data Age. Nowhere is that acceleration going to be more transformative than around the application of artificial intelligence and machine learning.
AI/ML was a hot topic before 2020 disrupted everything, and over the course of the pandemic, adoption has increased. We’ve seen it particularly in terms of security use cases, but security is far from the only arena. Already, it seems like artificial intelligence is everywhere. John Sabino, our chief customer officer, notes in the report that every software vendor is claiming AI/ML as a secret sauce in its solutions, and there’s a danger of fatigue as AI/ML becomes something everyone talks about, but no one ever quite sees.
Despite that, meaningful applications of machine learning in particular are already common. We see machine learning having an impact in everything from how recruiters parse stacks of resumes to how businesses analyze subtle trends in customer behavior; from improving user experience with everything from how web pages are served and products are recommended to intelligent chat features. And developments go far beyond business. Deep learning techniques produced a recent breakthrough in protein folding, which has applications in developing effective medical treatments, using enzymes to break down industrial waste, and more. It represents a considerable advance in AI development.
As we see machine learning adopted by more organizations, for more purposes, there are three innovations that I am keeping an eye out for in the near future:
- ML models that continually learn — with minimal supervision. The current pattern of humans building and deploying models simply does not scale. Models that can learn with fewer human-provided labels and more unstructured data may not appear in 2021, but we’ll see them in the next few years.
- Advances in adversarial learning and explainability. AI/ML is only as good as the data it learns from, and it’s possible to poison a data set. You can foil the image recognition systems that guide driverless cars. You can certainly turn a chatbot racist. Part of better governance is understanding why the model is behaving in the way it does. That explainability, coupled with solutions to withstand bad data or sabotage, will be a significant development.
- Ethical AI strategies. In the next few years, there will be further development of frameworks and practices for preventing bias in the algorithms that increasingly affect our daily lives. Part of the solution is ethics training, as is bringing in outside experts to consider the effects of an algorithm. For instance, if you’re going to use machine learning to help your bank determine who qualifies for a mortgage, consult with economists, urban planners and experts in the biases that have been seen in the real estate market and society in general. A parallel strategy: slowing down to consider the effects of our algorithms. Silicon Valley’s mantra of “move fast and break things” should apply to how software functions, not society.
The Emerging Technology Predictions report goes deeper into these topics, and other AI/ML predictions, including a stellar use case in medical research. It also covers 5G, AR/VR, blockchain and more. These are technologies that are going to reshape our world, and it’s fascinating to look ahead even as the future is unfolding.