Charting the Course of AI
By: Jasika Walia
This article from our research challenge appeared in Qrius.
“The development of full Artificial Intelligence could spell the end of the human race…. It would take off on its own, and re-design itself at an ever-increasing rate. Humans, who are limited by slow biological evolution, couldn’t compete, and would be superseded.”
— Stephen Hawking 
Artificial Intelligence has been all the talk in recent years. Over the last decade, we have seen massive growth in this field, with an expanding breadth of use cases in government, corporate, and consumer contexts. AI has become a part of almost everything we do, and we often do not realize it. Here in Figure 1, we see that the number of peer review papers on AI has been growing quite rapidly since 2015.
Fig. 1 – Number of Peer-Reviewed AI Publications, 2000-19 
We also see that investment in AI has been skyrocketing- doubling between 2020 and 2021- as shown in Figure 2 below.
Fig. 2 – Private Investment in AI, 2013-21 
A portion of AI’s growth is owed to some notable successes in its research and development over the last 5 to 10 years. It has performed exceptionally well in fields with widely available data and massive computational resources. These were the most high-return, low-risk analytical problems that were easiest to tackle first. The table below shows some familiar examples.
Table 1 – Use cases of AI 
While AI can perform more quickly, accurately, reliably, and impartially than humans on a wide range of problems, the technology has also suffered numerous failures. The increasing ubiquity of AI means that these failures can sometimes have a population-level impact, not just affecting a handful of individual users. Most of the failures come from data biases and a lack of understanding of the problem, as indicated in the table below.
Table 2: Failures of AI 
As the scope of AI grows, different views emerge on its path forward. However, it’s important to remember that predicting the course of any technology is a complex and error-prone undertaking. Take these contrasting predictions on nuclear energy from some forward-thinking figures of the past century.
“I have no doubt that we’ll have nuclear-powered vacuum-cleaners in say, 10 years time”
— Alex Lewyt, CEO of a vacuum-cleaner company, 1955.
“There is not the slightest indication that nuclear energy will ever be attainable. It would mean the atom would have to be shattered at will”.
— Albert Einstein, 1932. 
Despite this, there has been no shortage of efforts to try to map out the future of AI. Unfortunately, many of these attempts were forced to use broad figures or an arbitrary framework to make predictions.
For example, Figure 3 is a hype cycle chart from Gartner. This chart attempts to estimate how long to expect an AI application’s expectations to match its potential. Gartner acknowledges that applications can disappear and reappear anywhere on the chart. This is an interesting framework, and the expectations axis seems relatable; however, expectations are hard to quantify. For this reason, it is difficult to make any future projections using this chart.
Fig. 3 – Hype Cycle for Artificial Intelligence, 2021 
Despite the difficulty in making technology projections, we have conducted our own analysis by using a systematic approach. We started by listing AI’s most prominent market applications and then aimed to get its market sizing and rate the specifics of the AI application. We attempted to be as granular and quantitative in determining these figures as possible.
Background and Key Definitions
For our analysis, we have used the following definition of AI as proposed by IBM:
“Artificial Intelligence leverages computers and machines to mimic the problem-solving and decision-making capabilities of the human mind.” 
We have decided to focus on areas of Artificial Intelligence that are classifiable. Since AGI (Artificial General Intelligence – AI which can generalize concepts across different tasks) has such a range of opinions and definitions, we have lumped everything related to AGI together as one category for our research. In addition to IBM’s definition of AI, Figure 4 elaborates on what we are considering is and is not AI.
Fig. 4 – AI vs. Not AI
When we consider the history of AI, there have been several memorable moments. The start of AI showed that computers had the capacity to solve problems intelligently. Then there was a large gap of time with very little AI innovation. After the recent massive improvements in computing power, AI started to take off again. An abbreviated timeline of AI development is shown in Figure 5.
Fig. 5 – Abbreviated Timeline of Artificial Intelligence Development 
In general, there are four schools of thought – summarized in Figure 6 – related to the public discussion on AI. These are shaping the conversation about this important topic as it continues to rise in attention and funding.
Fig. 6 – Schools of Thought on the future of AI 
The utopian viewpoint is especially well represented in the current discussion. This causes a lot of the reporting on the development of AI to revolve around hype and makes it difficult to distinguish exaggerations and predictions from fact and reality. It also turns out that many alleged advances in AI are based on flimsy evidence. One review found that only 15 percent of AI studies shared their code. An example of this kind of exaggeration was in January 2019. A team from Google Health claimed in Nature that their AI program had outperformed humans in diagnosing breast cancer. In October, a group led by Benjamin Haibe-Kains, a computational genomics researcher, criticized the Google health paper, arguing that the “lack of details of the methods and algorithm code undermines its scientific value.” Researchers make dramatic claims that cannot be tested because researchers, especially those in industry, do not disclose their algorithms. 
Our research approach included the creation of a database of 60 AI applications to make a framework that easily reveals the current state of AI development. Each was rated with regard to data availability and development. To understand the rating scale, we put them on the X and Y axis of Figure 7 and divided it into 4 sections for discussion.
Fig. 7 – Rating Scale for Framework
We then marked the potential market size for each AI application as represented by the size of its bubble. The percentage of the potential market size that has been realized by the current market size is represented by the color of the bubble. The larger the currently realized share of the potential market, the greener the bubble. This key is shown in Figure 8.
Fig. 8 – Market Sizing Key
Using this key and the relationship between data availability and development, the AI applications are plotted in Figure 9. The arrow on each application represents its predicted movement over the next 3 to 5 years. Analyzing the applications, we see that applications lacking data are generally less developed and vice versa. This is expected as these tasks require more complicated algorithms to solve. We expect this trend to continue because it is part of the dynamics of how AI is developed.
Fig. 9 – AI Applications by Development and Data Availability
Let’s analyze this chart by region.
A. Upper Left – Mature – These AI applications have already shown significant progress. Most of the low-hanging fruit here is gone as industry capitalized on these areas and saw great returns early on. A tremendous amount of funding has gone into this space over the last decade. These applications will see declining returns, but there is a danger that companies overinvest in them since they are familiar with them.
Example – Advertising:
We have seen online advertising explode due to the development of AI. This field had incredible volumes of internet data waiting to be tapped into not that long ago. Now we have seen that some of the largest companies in the world filled this void and offer highly sophisticated data-driven advertising.
B. Lower Left – Untapped – These applications are near-term opportunities. This is because lots of data is available, but development potential still exists. We haven’t seen much funding in this space yet. As it stands now each of these is either waiting to be tapped into for great returns or they have some barrier to implementation.
Example – Energy Market:
AI can be used to efficiently optimize energy distribution. This is such a large project that it will take a lot of work to apply AI across all areas of this industry.
C. Upper Right – Ahead of the Curve – These applications were taken as the first attempt to have AI solve difficult algorithmic problems. As they progressed, they either got lots of funding and are shifting to become less AI-reliant technologies or are developing very slowly.
Example – Self Driving:
Self-driving cars have seen immense funding. As we saw huge potential for AI to handle the complicated task of driving, many companies moved into this space. Since there is pressure to see returns, these companies have realized where AI performs well in driving and where it does not. One example of this is programming in particular scenarios to account for variability in results.
D. Lower Right – Future Opportunities – Much more progress with AI is required to develop these applications. We need more/better data collection and management and possibly much more computing power. Investors need to be careful as the returns in this area may take a long time to be realized.
Example – Human Robots:
All of us have wondered when we will get to enjoy the services of robot butlers. However, the sheer number of situations where a robot like this will need to perform well is what makes it such a difficult task. We are much further away from a truly intelligent humanoid robot than is sometimes portrayed.
Based on the characteristics of each quadrant of the AI application chart in Figure 9, the predicted trajectory over the next 3 to 5 years is represented by the arrows. If we apply these arrows, we see a predicted version of the chart in Figure 10. Notice that the change in development over that time is dependent on both how available the data is as well as how developed it is currently.
Fig. 10 – AI Applications by Development and Data Availability in 3 to 5 Years
The changes in this predicted chart are broken down by quadrant.
A. Upper Left – These mature applications mostly will grow at a steady pace. Since they have lots of training data available, they will continue to eat away at their potential market size at a similar rate to the recent past.
B. Lower Left – These untapped applications will develop most of the four regions because they have the data available and have much more development potential. With time their development will match the trend of the other data-rich applications.
C. Upper Right – These ahead-of-the-curve AI applications will develop slowly as they have low amounts of training data for all possible scenarios. Since they were invested in early on and have more pressure to see returns, these applications will see some increase in data availability to meet their goals.
D. Lower Right – These future applications will progress faster than the upper right since there is much more room for growth. However, applications with very low data available for all possible scenarios will take a long time to develop regardless of how high the ceiling is. A large breakthrough may be required to see progress in those cases. One notable example may be a quantum computing solution.
Potential Investment Areas
There are several key takeaways from the AI application charts. Notably, we will take a deeper look at the untapped quadrant in the bottom left. Figure 11 shows the expected growth in market size for the three most promising AI applications based on our research.
There are a few different elements that make up smart cities, each with significant amounts of associated data compared to the other categories, but also with considerable room for development. However, even with substantial data availability, these are massive projects which will take many businesses/governments to implement on a noticeable scale. We expect a protracted timeline for these initiatives to mature, but their development rate may be more reliable. Here are the main applications of AI in smart cities:
Table 3 – AI applications for Smart Cities 
These are some notable companies to watch in this space:
- Zeleros – Hyperloop development
- Fleetonomy – Automobile fleet management
- Envio Systems – Building automation 
As the grid becomes increasingly more digital, we will see a shift towards AI to efficiently optimize energy distribution. This is going to be another area that will take a lot of work to apply across all areas of the industry. Some of the applications of AI in the energy market are as follows:
Table 4 – AI applications for Energy Market 
These are some notable companies to watch in this space:
- Landis+Gyr – Smart meters and communication systems
- Turntide – Smart energy management solutions
- BluWave-ai – AI-based SaaS platform 
Plant-Based Food Quality
As the trend for meat alternatives grows, a goal for manufacturers is to reproduce meat flavor so that consumers will not be deterred by a difference between the original meat and plant-based alternatives. There are two applications of AI in plant-based food quality that could see moderate growth with the right implementation:
Table 5 – AI applications for Plant Based Food 
These are some notable companies to watch in this space:
- Protera – Protein-based ingredients
- Climax Foods – Interpreting raw ingredients
- Inari – Seed design 
Based on this research, we believe that the future of AI cannot be captured in a homogenous manner. AI will progress depending on the specifics of the application. We especially see that data availability is crucial to determining this rate of development. The enormous amount of internet data has produced the means for the first wave of successful AI applications. Despite the increasing investment in AI, some promising and largely untapped areas for AI growth remain. While this analysis presents a unique way to view AI and its growth, this is not nearly the end of AI development research. We hope to continue to see many others attempt to map out this difficult market.
DISCLAIMER – This content is for informational purposes only. You should not construe any such information or other material as legal, tax, investment, financial, or other advice. Nothing contained in our article constitutes a solicitation, recommendation, endorsement, or offer by FischerJordan or any third-party service provider to buy or sell any securities or other financial instruments in this or in any other jurisdiction in which such solicitation or offer would be unlawful under the securities laws of such jurisdiction.
All Content in this article is information of a general nature and does not address the circumstances of any particular individual or entity. Nothing in the article constitutes professional and/or financial advice, nor does any information on the article constitute a comprehensive or complete statement of the matters discussed or the law relating thereto. FischerJordan is not a fiduciary by virtue of any person’s use of or access to the article or content. You alone assume the sole responsibility of evaluating the merits and risks associated with the use of any information or other Content on this article before making any decisions based on such information or other Content. In exchange for using the article, you agree not to hold FischerJordan, its affiliates, or any third-party service provider liable for any possible claim for damages arising from any decision you make based on information or other Content made available to you through the article.