By Cornelius Zeeman & Christoff Marais,
Fairtree Portfolio Manager & Equity Analyst
“The central issue in economic prosperity is innovation.” – Michael Porter
Labour Productivity, one of the key drivers of long-term economic growth, is defined as the output of a country per labour hour worked. Rising productivity is normally a good driver of rising real wages, which improves a country’s standard of living.
For the last few decades, however, Labour Productivity has been decelerating. This has been the result of several reasons, including a decline in the working to non-working population ratio, a plateau in educational attainment, and a slowdown in the expansion into more diverse and complex forms of production (World Bank, 2021) – see Graphs 1 and 2 below.
Graph 1: US Non Farm Labour Productivity Growth (%)
Source: U.S. Bureau of Labour Statistics
Graph 2: US Non Farm Labour Productivity Growth (%)
Source: U.S. Bureau of Labour Statistics
Can AI help alleviate this deceleration?
We think so. Numerous market pundits have been describing the recent proliferation as AI’s “Netscape moment” – the moment when products become more mainstream and accessible to the public. Artificial intelligence (AI) is the simulation of human intelligence by machines and has the potential to be one of the most transformative technologies of our time.
In this article, we will explore the potential of AI to make a positive impact on the economy. We will look at some of the ways that AI is already being used to solve real-world problems, and we will discuss the potential for AI to have an even greater impact in the future.
What are the use cases?
When researching the topic, it can be easy to be overwhelmed by the estimations for the total addressable markets, even though we are currently still far from mass adoption. The rapid improvement of AI models, as can be seen in Graph 3 below, is, however, worth emphasising. It took only around 2 years for a language model to outperform a human, whilst a handwriting recognition model from pre-2000 took close to 20 years.
Graph 3: AI system capabilities have improved rapidly | Test scores of AI applications relative to human performance
One of the first monetisable use cases that comes to mind is AI-Assisted Coding. Two of the main products currently available are Amazon Code Whisperer and Github Copilot. They are both AI-powered code completion tools that can help developers write code faster and more efficiently. They work by analysing the code that you are writing and suggesting code completions, functions, and other snippets of code that are relevant to the context.
As can be expected, this has the potential to improve the quantity and quality of programmers across industries significantly.
Some of the internal estimates by management have shown promise:
- “We’re now seeing that the developers using GitHub Copilot are 55% more productive on tasks,” commented Scott Guthrie, Microsoft EVP of Cloud & AI.
- “Internal test showed 57% faster task completion and 27% higher likelihood of success,” said Adam Selipski, AWS CEO.
PayPal has also stated that they believe that they can improve the productivity of their software developers by 30% with AI assisted tools.
Should these productivity enhancements hold true, the adoption curve might be a lot steeper than we might think. Consider the following: A productivity uplift of 20% – below Management and PayPal’s assumptions – still looks very impressive considering the Annual CoPilot Business subscription fee of $228 versus the base case productivity uplift of $24000.
AI for Advertising.
Meta is a great example of another company that has been at the forefront of creating AI-based tools to help improve the quality of advertisements on its platforms. These tools may not sound mind-blowing at first, but they provide easy ways to the millions of advertisers on their platforms to potentially improve their return on investment.
These tools allow users to expand images without sizing constraints, create new backgrounds as well as generate multiple versions of ad text, based on their original copy.
Creativity democratisation has been a major theme of AI as the barriers to image, music and art generation have been lowered. However, we believe that there is still significant room for differentiated creative material, even though some of the more easily replicable content can be at risk.
There currently exist numerous Generative AI tools, such as DALL-E 2, Midjourney and Firefly, which can generate images almost instantaneously based on a text prompt that is entered. Firefly, created by Adobe, has even gone a step further by guaranteeing users that its datasets are only trained by open-licensed images and copyright-expired public domain content – this means that Firefly is significantly less likely to generate images that contain copyrighted material – see Figure 1 below as an example of image generation by Leonardo.ai.
Figure 1: Image generated of President Joe Biden and President Abraham Lincoln shaking hands in the oval office
AI for Game Development.
Nexon’s Embark Studios, a game development company, has already been using Generative AI to increase the productivity of its developers. By using machine learning, the company has increased the speed by which it generates high-fidelity gaming environments. This aims to reduce very manual and tedious jobs and maximise the developer’s ability to focus more on creative thinking. For blockbuster games, which boast an ever-growing level of complexity, this implies thousands of hours which can be saved.
As an example, Embark has developed an automated tool to create game-ready photogrammetry in minutes – see Image 2 below:
Image 2: AI Image Generation
Source: Embark Studios, Morgan Stanley Research
Autonomous Driving might not be available today but has the potential to have a profound impact on our everyday lives. Take as an example Tesla’s latest FSD v12 (Full Self Driving) software, which is a significant departure from previous versions in that it relies less on rules-based programming and more on neural networks. This means the car can learn to drive by observing the road and other vehicles around it rather than being explicitly told what to do in every situation.
One of the key benefits of this approach is that it allows FSD to be more adaptive and handle unexpected situations more effectively. For example, if the car encounters a new type of road marking or traffic sign, it can learn to recognise and respond accordingly without having to be explicitly programmed.
Artificial Intelligence has the promise to be a possible solution to the slowdown in labour productivity that the US has experienced in the last decade. AI models and use cases are improving rapidly, with most models already performing better than humans in standardised tests. Although some use cases might be long dated, tangible examples already exist of how these models are improving our daily lives, as briefly touched on above.
Although the future is uncertain and unpredictable, AI will have an ever-increasing impact on our lives in the years to come.
Fairtree Asset Management (Pty) Ltd is an authorised financial services provider (FSP 25917). Collective Investment Schemes in Securities (CIS) should be considered as medium to long-term investments. The value may go up as well as down and past performance is not necessarily a guide to future performance. CISs are traded at the ruling price and can engage in scrip lending and borrowing. A schedule of fees, charges and maximum commissions is available on request from the Manager. A CIS may be closed to new investors in order for it to be managed more efficiently in accordance with its mandate. Performance has been calculated using net NAV to NAV numbers with income reinvested. There is no guarantee in respect of capital or returns in a portfolio. Prescient Management Company (RF) (Pty) Ltd is registered and approved under the Collective Investment Schemes Control Act (No.45 of 2002). For any additional information such as fund prices.