Trends in Artificial Intelligence in 2021

Artificial Intelligence is an emerging technology that is driving use cases in practically every industry from Manufacturing to Drug Discovery. The year 2021 saw big advancements in the field of Artificial Intelligence. Like every other year, there were bigger and better models beating the previous State of the Art models on various datasets. In this blog, we will review the key trends in the field of Artificial Intelligence in the year 2021 as highlighted by the Artificial Intelligence Index Report 2022.

The Artificial Intelligence Index Report is an annual report by the Stanford Institute for Human-Centered Artificial Intelligence (HAI) an independent initiative led by the AI Index Steering Committee, an interdisciplinary group of experts from across academia and industry.

According to the Artificial Intelligence Index Report 2022 by Stanford Institute for Human-Centered Artificial Intelligence, the following key trends were present in the world of Artificial Intelligence in year 2021-

Private investment soared, while investment concentration intensified

The private investment in Artificial Intelligence in 2021 increased to $93.5 billion, more than double from $44 billion in 2020. This has been the largest year over year increase in private investments since 2014.

Meanwhile, the concentration of investments in Artificial Intelligence has also increased. The number of newly funded startups has also dropped for the third year in a row.

United States and China dominated cross-country collaborations on AI

There have been some geopolitical conflicts in recent years between the United States and China, but despite this the United States and China had the greatest number of cross-country collaborations in AI publications from 2010 to 2021, increasing five times since 2010. This is more than twice the number of cross-country collaborations between the United Kingdom and China, which comes next to only United States and China in the number of cross-country collaborations.

Language models are getting better, but also more biased

While transformer models like GPT-2, BERT, and GPT-3 keep getting bigger, their capacity to reflect biases in the training data also increases. A 280 billion parameter model developed in 2021 shows a 29% increase in elicited toxicity over a 117 million parameter model considered the state of the art as of 2018.

The rise of AI ethics

Deep Learning models are Black Box models and their ability to inherit biases from the training data has pushed for research in the field of Fair AI and Explainable AI(XAI). Research focussed on fairness and transparency has increased by a ton since 2014 with a five-fold increase in related publications. Researchers with industry affiliations contributed to 71% more publications year over year at ethics-focussed conferences in recent years.

AI becomes more affordable and higher performing

In recent years, there has been a steady decrease in the cost for training a Deep Learning model but an improvement in training time and performance. This can be attributed to both, optimizations in software, as well as improvements in hardware. Since 2018, the cost to train an image classification system has decreased by 63.6%, while training times have improved by 94.4%.

Data- The Digital Gold

The increase in the data available to train Deep Learning models was one of the factors that enabled AI to grow as fast as it did. This trend has continued in 2021. Out of the 10 benchmarks in Artificial Intelligence Index Report, 9 were trained with extra data. This implicitly favours private entities with a vast amount of datasets.

Increase in Legislation on AI

According to an AI Index analysis of legislative records on AI in 25 countries, the number of bills containing “Artificial Intelligence” that were passed into law grew from just 1 in 2016 to 18 in 2021. The highest number of AI related bills passed in 2021 were in Spain, the United Kingdom, and the United States.

While the increase in legislation will increase regularization and push for fairness in Artificial Intelligence while also bringing in funding. An increase in legislation will also amount to an increase in compliance costs.

Robotic arms are becoming cheaper

The median price of robotic arms has decreased by almost half in the past five years from $42,000 per arm in 2017 to $22,600 in 2021. This can be attributed largely to robotics research becoming more accessible and affordable.