Is the AI hype machine broken? In the following deep dive, leading economist and Syz Bank CIO, Charles-Henry Monchau plots the potential future course of AI technology and finds a course between caution and optimism.

Sceptics of this current wave of AI investments are pointing to several economic, technical, and market-based concerns that suggest the technology might indeed be overhyped.

The first of these being economic concerns revolving around the limited short-term benefits of investments into AI.  While the picture seems clearer in the long-term, immediate returns have been underwhelming for many companies and it is still unclear exactly how a lot of companies plan on monetising their tools.

This is in part because the integration of AI into business processes is often more complicated, time-consuming, and capital intensive than initially anticipated.

Capital costs

While most experts are open to the possibility of AI tools revolutionising scientific processes, many place those revolutions 20-30 years down the road, with humans still in the driver’s seat. One factor that is cause for worry is the fact that any company seriously investing into AI infrastructure, such as data centres and chips, will face significant upfront costs, with a delayed realisation of benefits.

Another cause for concern is the rapid proliferation of AI startups. Current forecasts place the CAGR for the AI market at close to 40% for the next 6 years. The AI landscape is becoming increasingly more crowded with a large number of startups, many of which are unproven and operate with business models that have yet to demonstrate profitability.

The media and marketing hype further exacerbate the issue of market saturation. There is often a significant gap between the perceived capabilities of AI, as portrayed by enthusiastic media reports and marketing campaigns, and the actual progress and limitations of the technology.

Expectation meets reality

This discrepancy often leads to unrealistic expectations among investors and the public, potentially resulting in disillusionment when AI systems fail to meet these lofty promises. Such a situation can erode trust in AI technologies and hinder their adoption, even if they hold genuine potential.

It is also hard for people not to draw comparisons to the tech bubble of the late 1990’s and early 2000’s. This recent and not too dissimilar example of a tech bubble is certainly a driving force behind the scepticism of many.

During the tech bubble, massive investments were funnelled into internet startups with the expectation of rapid and transformative returns. As many of these companies failed to deliver, the bubble eventually burst, leading to approx. $5 trillion in loss and economic disruption.

Figures place the number of dot-com companies that survived to see 2004 at just 48%. Sceptics have good reason to worry that AI might follow a similar trajectory if the exceedingly high expectations are not met, potentially leading to an economic downturn if a significant portion of AI ventures fail to achieve profitability.

Generative AI also faces its fair share of criticisms and limitations and understanding these challenges is crucial to developing a balanced perspective on the technology’s potential and the prudence required to make these significant investments.

Quantum hurdles

Despite its advancements, Gen AI faces several technological hurdles that impede its reliability and effectiveness, such as data quality and availability. These models require vast amounts of high-quality data to function effectively but sourcing this data can be challenging, particularly in fields where data privacy and security are crucial.

Additionally, poor quality or biased data can lead to inaccurate outputs, such as Google’s Gemini model creating images of racially diverse Nazis or the Founding Father’s.

 Furthermore, training these ever more advanced AI models demands significant computational resources, which are costly and also require massive energy consumption, which raises questions about the sustainability of continued AI development. This energy hungry technology is predicted to consume twice as much energy as the whole of France by just 2030.

There are also several ethical concerns regarding the deployment of Gen AI, with privacy being at the forefront of these. Generative AI systems often require access to vast amounts of personal data and with that, comes the risk of misuse or unauthorised access to sensitive information, which can lead to breaches of privacy and trust.

Reasons for optimism

It is always important to avoid confirmation biases and analyse situations through a contrarian lens. However, it is also important not to get caught up in the scepticism and doomsday mentality, as there are just as many voices with an optimistic perspective on the current wave of AI, arguing that it is indeed not overhyped.

First and foremost, one of the most compelling arguments for the optimistic view on AI comes from its potential for long-term productivity and economic growth. Gen AI has the capacity to revolutionise industries by automating complex tasks, optimising operations, and significantly increasing efficiency.

This transformative potential is expected to yield substantial productivity gains, leading to economic growth. For instance, Goldman Sachs analyst expect a full automation of 25% of all work tasks following the technology’s adoption, with their baseline estimate implying as much as 15% cumulative gross upside to US labour productivity and GDP growth.

One such application, that is already very much available, is to use AI to generate novel ideas and provide suggestions, making the model a “collaborator” of some sort. The productivity gains also stretch into innovation in product development, in which companies are using Gen AI to analyse consumer preferences, market trends, and competitive products to generate ideas for new products and features.

As impressive as some of the language models, like ChatGPT and the image generating models, like Midjourney, are, these technologies are still very much in their infancy and expected by some experts to improve at an exponential rate. The billions being invested into AI research and development are also sure to accelerate the rate of improvement, allowing AI to tackle increasingly complex tasks, as it also augments human creativity.