AI investments soared in 2021, but big problems remain

Commentary: For all the money and talent being thrown at AI, we still haven’t solved some of its most basic shortcomings.

The artificial intelligence rich definitely got richer in 2021, according to the 2022 Stanford AI Index report. Private venture investment in AI exploded to $93.5 billion in 2021, more than doubling the 2020 tally. Even as investment levels have ballooned, the number of companies getting that money has gone down. In 2019, venture capitalists funded 1,051 AI companies. In 2020, that number dropped to 762, then plunged again to 746 in 2021, even as the size of funding rounds skyrocketed for the lucky few: In 2020 there were just four funding rounds that exceeded $500 million, but in 2021, that number climbed to 15.

SEE: Artificial Intelligence Ethics Policy (TechRepublic Premium) 

All of which may indicate that the stakes for AI keep increasing. If only we could say the same for the outcomes.

The AI gold rush

By a number of measures, interest in AI is off the charts. Take research and development, for example. According to the report, the number of AI patents filed in 2021 was 30X higher than in 2015, representing a 76.9% compound annual growth rate. Those patents, in turn, are helping to fuel a frenzy in venture funding of startups, as mentioned above. While such investment is widespread, the highest concentration of funding is in the U.S. At $52.9 billion in 2021 funding, the US funds AI at more than three times the rate of the next country (China, $17.2 billion) and over 10 times third place (United Kingdom, $4.6 billion).

Given how much money is pouring into AI, it’s not surprising that companies are feverishly looking for AI talent. According to the report, the share of job postings that mention a need for AI skills was up across the globe, with the most demand for machine learning skills (0.6% of all job postings), followed by artificial intelligence (0.33%), neural networks (0.16%) and natural language processing (0.13%). What are the hottest sectors for AI? In the U.S., the number one industry for AI jobs is Information. Last place? Waste management.

At the same time, more people than ever before are getting degrees in related fields to prepare themselves for these jobs:

In sum, there’s more talent chasing more jobs in companies getting more money. Yet AI reality can’t quite keep up.

For example, in the area of deep learning, AI expert Gary Marcus suggested that DL is “at its best when all we need are rough-ready results, where stakes are low and perfect results optional.” That’s useful, but it’s not robots reasoning with general intelligence like we sometimes imagine AI should be delivering by now.

Ask the IEEE technical crowd, and they wonder if AI is “reaching its limits.” Then there’s the heightened concern that for all its promise, we still haven’t tackled the most basic questions about AI and built-in bias.

Small wonder, then, that on Gartner’s 2021 Hype Cycle for AI, most AI-related disciples are barreling up the Peak of Inflated Expectations, preparing themselves for a slump into the Trough of Disillusionment. Just a small handful of things—like Autonomous Vehicles—are readying to leave the Trough and, in the case of autonomous vehicles, it’s unclear that so-called self-driving cars are anywhere near true self-driving. (As analyst Benedict Evans has written, “[V]ersion nine of ‘Full Self-Driving’ is shipping soon (in beta) and yet will not in fact be full self-driving, or anything close to it.”

No, this doesn’t mean there’s no substance underlying the AI euphoria. Investors are betting big on tomorrow’s potential, not today’s reality. That’s fine. But let’s not get ahead of ourselves. As David Meyers has said, “Too many businesses now are pitching AI almost as though it’s batteries included [which may] potentially lead to over-investment in things that over-promise. Then when they under-deliver, it has a deflationary effect on people’s attitudes toward the space.” We shouldn’t dim our hopes in AI, but should temper near-term expectations.

Disclosure: I work for MongoDB, but the views expressed herein are mine alone.