Artificial intelligence saw rapid growth in 2020. The global pandemic and the resulting digital transformation forced upon companies and individuals as we moved from offices to homes accelerated AI usage and development.
And it seems that this pace will not slow down in 2021. Here are a few predictions on how AI will continue to dominate in the coming year.
AI investment will skyrocket
As noted in PWC’s annual AI predictions survey, this is probably the most straightforward prediction to make.
In that report, PWC noted that 86% of its respondents say that AI will be a “mainstream technology” at their company in 2021. Whether it is being used to offer better customer service, help stakeholders make better business decisions, innovate existing products and services (and create new ones), achieve significant cost savings, or increase productivity, a majority of organizations will use AI to give them a competitive advantage.
Indeed, you could argue that the use of AI within your company will be mandatory soon to survive, in the same way that Gartner once claimed that every business – large or small – needed a website by the year 2000 to stay alive.
This leads us to another prediction: one that allows all businesses to take advantage of what AI offers without hiring specialists.
AI as a Service will explode in 2021
While demand for AI is ever increasing, the talent pool is reducing. It isn’t easy to hire data scientists at this time, and it will become even more difficult as the need to incorporate AI in your business becomes more urgent.
One solution is to make use of the many AI as a Service platforms that are appearing. Using these platforms is a cost-effective and fast way to adopt artificial intelligence and integrate it into existing systems for smaller businesses.
We’ve already witnessed these solutions scale dramatically in 2020, and we expect that to accelerate further in 2021. Indeed, the future of digital transformation is more likely to come from easy-to-use, easy to adopt platforms that bring the power of AI to every business, rather than just those that can afford to hire experts and build their own solutions.
Initially, we expect AIaaS platforms to be adopted for improved customer service, data analysis, and financial reporting. Still, we expect a wide range of AI capabilities to join the ever-growing AIaaS industry.
We’ll see more deep fake legislation, but it won’t fix the problem
Deep fake audio and video have been on the rise over the last three years. Famous examples, such as the fake Obama video that circulated in 2018, have highlighted the dangers of deep fake videos in the political sphere. A recent incident in India showed how deep fakes could have a real effect on voting and elections.
And while the state of California in the US passed a bill that made it illegal to circulate deep fake videos of politicians within 60 days of an election, it is clear that laws won’t deter the perpetrators and distributors of such videos.
In 2021, we expect to see more legislation around deep fakes – both punishing production and distribution – but what is needed is a better way to identify them. The Rochester Institute of Technology (RIT) in New York has built its own deep fake detection software, and a browser plugin called Reality Defender is helping to identify fake videos.
That being said, the answer may be non-technical. In the case of India’s deep fakes, a group of people noticed a slight anomaly in the mouth movements and raised the alarm. As more negative deep fakes are being circulated, it may be that a simple awareness campaign will be the best way to counter the effects.
There will be considerable advances in Federated Learning
With privacy and security becoming mainstream topics after several high profile cases and documentaries such as The Great Hack, it is becoming essential to ensure data privacy at all times.
Federated Learning helps to achieve that. Google describes how FL works in this way concerning mobile phones:
It works like this: your device downloads the current model, improves it by learning from data on your phone, and then summarizes the changes as a small, focused update. Only this update to the model is sent to the cloud, using encrypted communication, where it is immediately averaged with other user updates to improve the shared model. All the training data remains on your device, and no individual updates are stored in the cloud.
FL enables devices like mobile phones to collaboratively learn a shared prediction model while keeping the device’s training data instead of requiring the data to be uploaded and stored on a central server.
FL moves model training to the edge; smartphones, tablets, IoT devices, or even “organizations” such as hospitals that are required to operate under strict privacy constraints. Having personal data remain local is a substantial security benefit.
And since models sit on the device, the prediction process works even when there is no internet connectivity.
The number of papers written about FL has exploded in the last two years, and it looks like 2021 will be the year that FL will become a mainstay for anyone working in machine learning.