Artificial Intelligence (AI) had another eventful year in 2021. From language models with trillions of parameters to record investments in AI firms to dilemmas around ethical AI, the year for the disruptive technology could not have been scripted any better. That brings us to these questions: What will be the biggest trends in AI in the coming year and beyond.
For starters, AI is already making it easier to set New Year resolutions. AI researcher-writer Janelle Shane has built an AI for the Smithsonian called “Resolutions Generator” that turns up some really wild and weird ideas. Trained on the powerful GPT-3 model, the AI Resolutions Generator is also infused with some of the resolutions humans have put online with parameters set wide. If you want a New Year’s resolution beyond losing those 15 kilos or reading one book every week, this is a great place to start.
However, AI is destined to scale new heights in 2022 and will inch closer to justifying Alphabet CEO Sundar Pichai’s claim that the impact of AI will be even more significant than that of fire or electricity. We are set to see machines that can compute and come to an inference faster than before while NLP brings more value. With that in mind, these are the major AI and ML trends to follow in 2022.
Language AI will become transformative
Language AI can be described as the ability of computers to effectively understand all human language. Language AI was one of the heated topics in 2021 and this year, we could see it take the centre stage. Language AI, if done right, could transform the way we engage with brands, businesses and organisations around the world.
The field of natural language processing (NLP) has become powerful enough to be commercialised at scale and that will lead to a revolution in language AI. Forbes predicts that we will see record amounts of money go into NLP startups while leading players like Hugging Face and Cohere could become unicorns this year. The key thing to watch is whether entrepreneurs are able to identify and unlock value by optimising and automating language-based activities.
Improved cybersecurity
Cyberattacks have been on the rise and each year, the bad actors come armed with more complex tools. In 2022 and beyond, we are likely to see bad actors use AI to improve their skills and even personalise their attacks on enterprise infrastructures. The World Economic Forum has stated that cybercrime is a more significant threat to society than terrorism.
In order to tackle this, we will see companies invest more in AI and ML-based solutions to either thwart cyberattacks or take pre-emptive action. One of the ways it could play out is the development of AI tools to analyse higher network traffic and recognise nefarious actors even before they deploy their attack. Cybersecurity could see major disruption from AI tools in 2022.
Midsize businesses will get AI-fuelled faster
In 2022, we will see midsize businesses adopt AI technologies like automation faster than some larger enterprises. The COVID-19 pandemic has made it clear that digitalisation is the way for sustainable business outcomes and midsize businesses are expected to go all-in by offering cloud-based services and adopting advanced technologies like artificial intelligence (AI) and automation.
These companies, according to IDC and SAP, will also link all of their business applications together. We should also see midsize businesses adopt predictive analysis, conversational AI, intelligent chatbot, modern ERP and other cloud-based business technology platforms. In October 2021, Deloitte found that only a few organisations are completely AI-fuelled today. The analysis of AI-fuelled organisations could change in a big way in 2022.
Multimodal AI will be on the rise
While the demand for complex computer systems increases, single type AI models – used for audio processing, computer vision or natural language processing – doesn’t always cut it anymore. Enter the space of Multimodal AI-systems which are able to combine multiple input variables and combine them into a single intelligent solution. The microphones in your self driving car might hear you say you fancy a cup of coffee, the camera notice a road sign with a Starbucks nearby and automatically steers you to it.
Increased productivity
In 2021, AI was deployed by enterprises to automate some of the repetitive tasks and monotonous work across business verticals. While many feared this could lead to robots taking over human jobs, the end result was that robots and AI did not lead to a job apocalypse. In fact, automation led to human workers becoming more efficient.
We will see more companies adopt AI technologies, automation and robotics in 2022, to not replace their existing workforce but to supplement them. We will see humans master these programs and it could also fill for the talent exodus that has been a major issue for companies.
Adoption of low-code and no-code technologies
App development is no longer exclusive to those who know how to code. Thanks to the development of low-code platforms, anyone can become a developer but no-code technologies will take this to a new level by offering rich functionality and the ability to turn projects around in a fraction of the time it takes to build using traditional coding options.
A combination of low-code and no-code technologies will allow people with no IT skills to not only upskill themselves but also bridge the gap between idea and working application. From faster deployment, reduced cost and time to agility and allowing anyone to develop, low-code and no-code technologies supported by AI will lead to enhanced innovation.
AI will continue to move to the edge
While the cloud has all the power you need to go through large datasets and train complex models, there is the downside of a steep price tag. Moreover, sending all your data in real time to the cloud isn’t always the most practical solution. Therefore, we’ve seen the move from cloud based AI solutions to edge devices with build in chips to preform AI-tasks on the fly.
Chipmaker Nvidia is very bullish on the move to edge devices, as well as research done by IBM, which shows that 94 percent of executives will implement edge computing in their organizations in the next five years. Nvidia suspects that Edge Management will become a focus point for IT-staff and sees big opportunities for industrial AI and IoT-solutions.
AI models will get larger
This is a no-brainer as far as one of the trends of AI and ML in 2022 to follow. We have seen the branch of AI called natural language processing (NLP) significantly transform in 2021. Towards the end of 2020, OpenAI announced GPT-3 as the successor to GPT-2 with 175 billion parameters. But the reign of GPT-3 as the largest AI model was short as companies raced to develop ever-larger transformer-based models.
In 2021, we saw models from Google (1.6 trillion parameters) and the Beijing Academy of Artificial Intelligence (1.75 trillion parameters) break the trillion parameter barrier. This race to build outsized transformer-based AI models will continue and it could come from none other than OpenAI in the form of GPT-4 before it gets dethroned again.
‘Small and wide’ datasets will also flourish
While big tech will be able to crunch the data needed for these large scale datasets and additional cost of compute power – many organizations will also benefit from ‘small and wide’ datasets. As we have seen in the recent pandemic, many models could not cope with an unexpected change in human behaviour. A shift to small and wide datasets would mean leaner AI which can be more flexible in rapidly changing environments while keeping cost down.
According to Gartner, by 2025, 70% of organizations will be compelled to shift their focus from big to small and wide data, providing more context for analytics — and making AI less data-hungry.
AI tools built for video
When Squid Game came out in September on Netflix, it led to increased network traffic forcing South Korea Broadband Corporation to sue the streaming giant for increased maintenance work. The streaming platforms have been a boon for those locked inside their homes during the pandemic and they are only expected to grow in the coming years as more people cut chords in favour of streaming services.
This will require building deep learning-based products and capabilities tailored for the delivery and consumption of video. From how your video reaches the final device to video search, video editing and video generation, AI tools could set a new war for data dominance between Amazon Prime Video, Netflix, HBO Max and Disney+.
The next AI hub
The US and China have become magnets for attracting AI talent and building some of the most data-centric businesses. However, Forbes predicts that Toronto will become the next most important AI hub in the world. Technically, the modern AI was invented in Toronto thanks to the work of pioneers like Geoffrey Hinton but it hasn’t generated much buzz like Silicon Valley.
However, the situation is changing fast with Google, Microsoft and IBM establishing an outpost in the capital city of Ontario in recent years. Toronto-based AI startups are leading the work in AI for drug discovery (Deep Genomics), NLP (Cohere), chatbot platform (Ada). However, Toronto will face competition from Europe, which is looking to cement its place as the next major player in AI.
Responsible AI will become an operational product
Google found itself in a soup when the search giant fired AI researchers Timnit Gebru and Margaret Mitchell, the co-leads of its Ethical AI team. There is now a growing movement to advocate for the responsible use of AI including addressing issues such as AI bias, data provenance, model explainability and auditability.
In 2022, we could see “responsible AI” becoming a product in the form of AI practises and toolkits. We should see responsible AI practises become a standard affair across tech giants like Microsoft and IBM, and startups like Fiddler Labs and Parity. The responsible AI practice will also be supported by regulators with the EU’s proposed AI Act leading the way.
Rise of AI quantum initiatives
Quantum computing has had some major breakthroughs in 2021. With technology giants like IBM, Microsoft and Google (which even managed to create the first ever ‘time crystal’ last year) investing heavily in the technology, and the rise of more startups and scaleups who want to leverage the endless compute power of quantum – we are optimistic about the applications for AI in the upcoming years. Thanks to Quantum AI companies and organizations will be able to crunch more data and generate more complex AI models.
Sustainability
One of the big trends in 2022 will be sustainability within the industry as well as how tech can help increase sustainability. AI and machine learning will also become sustainable in 2022 with the research name-checking AI. We are already seeing the use of AI by major fashion brands and small startups to create a more sustainable world.
However, the need to report on the carbon footprint of their AI tools could deter some companies from adopting a sustainable practice. In 2022, every company will focus on their environmental, social and governance goals as they aim to reduce its carbon footprint.
Reinforcement learning will become important
The real-world potential of reinforcement learning will become clear in 2022. Even though it has been around for decades, the dominant approach to AI today is supervised learning. This works by collecting a lot of data, labelling them and feeding it into an AI model and allowing AI to learn useful patterns about the world.
With reinforcement learning, the AI model is allowed to open-endedly explore its environment instead of asking it to pay attention to a certain element like supervised learning. The approach allows the AI model to learn about the world as it goes and it optimises only based on a particular objective. The triumph of DeepMind’s AlphaGo should encourage more companies to adopt this AI paradigm.