It has been a wild year in every quarter, and AI development is no exception. On the whole, the year has been mixed for AI, as there have been both notable advances and new revelations about abusive applications of the technology. And the market for AI technologies appears to have plateaued, with a recent global survey finding no increase in AI adoption in the enterprise. This helps to explain why Element AI, a once high-flying startup that built AI applications for enterprises who otherwise lacked the requisite skills, was ultimately not able to survive on its own.
A new report on AI adoption by IndustryLab found that implementing AI within an enterprise often runs into people challenges, such as fear of change and job loss as well as a lack of relevant skills. According to the report, 87% percent of survey respondents faced people challenges in their AI implementations. These issues remain a substantial barrier to enterprise AI adoption. It is no wonder progress has been slow within businesses, giving the appearance of a plateau.
But despite such resistance, AI technology continues to move forward. Recent AI technology advances range from improved synthetic speech to safeguarding bee health, creating a next-generation food system and developing new recipes, improving treatment for breast cancer, uncovering government corruption, and building smarter traffic lights. These and other advances are part of why a PwC study estimates that by 2030 AI will boost global economic output by more than $15 trillion. Alphabet’s Sundar Pichai famously claimed AI is more profound that electricity or fire. At least one major data analytics platform believes 2021 will be the year of AI as several large sectors including oil & gas, fintech, and drug research firms will increasingly embrace the technology.
So has AI really plateaued or are we just witnessing a pause before a new period of steep adoption? We would expect such a pause to result from cognitive dissonance — the advance of AI meeting fear, resistance to change, and uncertainty about whether the tech will live up to the hype. At one extreme are predictions such as one from Vladimir Putin that whoever becomes the leader in AI will become the ruler of the world. At the other extreme is an analysis of 40 of the largest AI startups that suggests these companies are not having a great impact, either on change or on the economy. If the latter is true, we may be at the beginning of the next AI winter, with expectations once again exceeding reality.
Consequently, the crystal ball for AI is decidedly cloudy. We are either on a plateau with the risk of falling into a chasm, or we’re readying for the next round of innovation. Most likely, there are two paths playing out in parallel: continued advancement of technical capabilities and the very human challenges of implementation.
2020: A year like no other
While AI adoption in the enterprise has slowed, major breakthroughs in AI research this year are a reminder that this is an area of technology capable of unleashing exponential change.
Natural Language Processing in the form of GPT-3 developed by OpenAI could be the precursor for the first artificial general intelligence (AGI), a massive advancement. GPT-3 “learns” based on patterns it discovers in data gleaned from the internet, from Reddit posts to Wikipedia to fan fiction and other sources. Based on that learning, GPT-3 is capable of many different tasks with no additional training, able to produce compelling narratives, generate computer code, autocomplete images, translate between languages, and perform math calculations, among other feats, including some its creators did not plan. This apparent multifunctional capability is a departure from all existing AI capabilities. Indeed, it is much more general in function.
With 175 billion parameters, the model goes well beyond the 10 billion in the most advanced neural networks, and far beyond the 1.5 billion in its predecessor, GPT-2. This is more than a 10x increase in model complexity in just over a year, making it arguably the largest neural network yet created.
Another significant advance comes from DeepMind with AlphaFold, an attention-based deep learning neural network that may have solved a nearly 50-year-old challenge in biology: determining the 3D shape of proteins from their amino acid sequence. Proteins are the building blocks of life, responsible for most of what happens inside cells. How a protein works and what it does is determined by its 3D shape. Until now, determining the structure of proteins has been difficult, laborious, expensive, and prone to failure.
The AlphaFold system outperformed around 100 other teams in a biennial protein-structure prediction challenge called CASP, short for Critical Assessment of Structure Prediction. On protein targets considered to be moderately difficult, the neural net achieved prediction accuracy of 90%, far better than other teams; some consider it to be biology’s holy grail achievement. The advance is expected to vastly accelerate understanding of the building blocks of cells, enable quicker and more advanced drug discovery, and basically herald a revolution in biology comparable to the DNA double-helix model and the CRISPR-Cas9 genome editing technique.
As significant as these developments are, it is impossible to overlook AI’s contributions to coping with the COVID-19 pandemic. AI has helped tracking the spread of the disease to limit the number of cases, has digested and distilled the thousands of papers on the topic, and is now managing complex supply chains for vaccines, as well as combing data to track any adverse effects individuals might have in response. Imagine how much worse the impact and duration of the pandemic would be if not for AI. It is possible this “moonshot” endeavor will spur AI R&D and deployment across many sectors for years.
With enterprise adoption lagging, 2021 may not turn out to be the year of AI. But it will certainly see more breakthroughs like the ones we’ve seen this year and will carry us into the next phase of an inexorable march forward towards greater intelligence.