Image generated with AI
Even the most compelling technological innovation is subject to the highs and lows of elevated expectations and disappointment. The ‘hype cycle’ is a tool designed to show how a technology or application will evolve over time, and to systemise these highs and lows. It can be useful in evaluating whether a great idea will make a great investment, and to take the temperature of investor expectations.
Gartner published its first ‘hype cycle’ in 1995. It created an arc to demonstrate how new technologies or applications progress over time, suggesting that technologies have a relatively predictable evolution of innovation, expectations, disillusionment, enlightenment and productivity. It is an imperfect tool, but can be a useful prism to view new technologies, such as artificial intelligence ('AI').
The hype cycle is published every year, mapping evolving technologies onto the standard pathway. While some technologies will not get off the ground, and there is some hindsight bias in companies that follow the pattern, it can provide both investors and business leaders with some clues as to where to invest. Should investors commit to a certain company, or is its valuation inflated? Should a CEO pour company capital into AI on the basis that it will bring about competitive advantage? Or will this prove to be a waste of money?
The cycle identifies five key stages for every technology. The first is the ‘technology trigger’, where a new and innovative technology breakthrough appears. Its early ‘proof-of-concept’ stories gather publicity, even if no usable products exist and its commercial viability is unproven. This builds to a ‘peak of inflated expectations’, where a number of success stories make headlines (failures may be overlooked). Some forward-thinking companies may adopt the new technology, but most do not.
The next step is the ‘trough of disillusionment’, where interest wanes and early implementations do not deliver the hoped-for productivity gains. The market starts to shake out, with producers going bust or merging. Surviving providers need to improve. Next is the ‘slope of enlightenment’, where real-world, effective applications of the technology emerge as it is refined. More enterprises adopt the technology. Finally, there is the ‘plateau of productivity’, where mainstream adoption starts to take off and the technology achieves some of its early promise.
This is only a loose framework. Nevertheless, investors will recognise this trajectory from, for example, internet adoption. In the late 1990s, there was significant hype around the potential for the internet, a raft of speculative companies IPOed, only to disappoint as revenues failed to materialise. The resulting stock market slump was painful, but ultimately there were companies that delivered on the promise of the internet, particularly after the advent of the iPhone in 2007.
There are now other incarnations of the hype cycle. For example, some data analytics companies will look at media attention to predict technology trends, applying machine learning to plot the arc of rising expectations and excitement, disillusionment and reality. A recent study by CB Insights looked at the adoption of wearables, finding that while general adoption of variables had plateaued, disease-specific and clinical wearables had increased.
This year’s hype cycle puts generative AI at the ‘Peak of Inflated Expectations’, which means its next phase may be the ‘Trough of Disillusionment’ if it follows normal patterns. There are certainly signs of exuberance in AI. Big tech has spent heavily on AI, with Microsoft, Google and Amazon doing a number of blockbuster deals with AI start-ups in 2023. This accounted for two-thirds of the $27bn raised by fledgling AI companies in 2023, according to private market researchers PitchBook. Overall spending on AI groups is nearly three times as much as the previous record of $11bn set two years ago.
Of these, Microsoft’s multi billion dollar investment in ChatGPT maker Open AI has garnered the most attention. The tie-up is designed to accelerate AI breakthroughs. Microsoft has been among the first companies to market with mainstream AI products as it launched Copilot in its Windows applications.
Nevertheless, it was only generative AI and ‘augmented AI’ (where AI technologies help humans make better decisions rather than replacing them) that are at this stage in the cycle. This fits with the pattern seen in stock markets, where those on the front line of AI – chip providers, cloud computing groups, AI developers – have seen substantial share price rises in 2023, amid significant excitement about what AI could achieve.
Other types of AI technology are at a far earlier point on the adoption curve: causal AI, for example, is only just starting to be explored. This new class of machine intelligence can use reason in a similar way to humans, looking at cause and effect. Neuro-symbolic AI, which combines machine learning methods and symbolic systems to create more robust AI models, is also at an early stage. The cycle also highlights AI simulation and generative security AI as early stage technologies.
The easy conclusion from the hype cycle would be that generative AI is likely to be hit by a dose of reality in year ahead. Certainly, this has been seen in other ‘hype’ technologies, such as crypto. The value of deals struck by US venture funds halved between the first and fourth quarters of 2022, from $81bn to $41bn, according to PitchBook. Crypto deals plunged further, down more than 80% over the same period.
AI disillusionment may emerge as companies understand the level of investment needed to monetise AI, and that its advantages may only be open to a select few players with deep pockets. It is also possible that investors reassess the elevated valuations for AI companies if practical applications do not emerge as quickly as hoped.
However, while 2024 is certainly likely to be a pivotal year for AI and market leadership may shift, we don’t see an end to AI growth. It is important not to put AI in the same bucket as early internet companies with no revenues. This is a technology that could be transformative for company productivity and economic growth. We are only in the foothills of that development.
There will always be setbacks and periods where the technology appears to be moving backwards. As people begin using AI-related applications, the prevalence of errors will grow. There will undoubtedly be some difficult headlines where AI is employed in sensitive sectors such as health and defence. Equally, a fundamentally disruptive application could still be some way off.
Generative AI will be constrained by data and analytics governance, regulatory, and data security considerations, and will require human supervision. Policymakers have learned their lessons on social media and will want to ensure that AI does not come with similarly unintended consequences.
However, as the hype cycle shows, this is a necessary process for AI to become productive. A lot of the most valuable AI companies will be fundamentally new. Gartner says: “Generative AI techniques applied to data management are still in their infancy. Vendors are still in the very early stages of using this technology to power their data management tools. But with the vast investments in the area, we expect a rapid progression of product availability.”
The hype cycle would suggest there is a lot more to go for in AI, but the gains will come from different places. The next wave is likely to focus on the applications of AI and in particular, the search for a ‘killer app’, rather than the creators of the AI itself. It is a necessary evolution. It took time to see where the most productive uses of the Internet would be, and AI is no different.
Technologies do not evolve in a straight line. It takes time for the use cases to emerge and for the technology itself to be refined. AI is an exciting innovation, but many of the exciting investment opportunities from it are yet to emerge.
“These complex, opaque systems may do more societal harm than economic good. With virtually no US government oversight, private companies use AI software to make determinations about health and medicine, employment, creditworthiness, and even criminal justice without having to answer for how they’re ensuring that programs aren’t encoded, consciously or unconsciously, with structural biases.” This was the bleak verdict on AI from the Harvard Gazette unless governments take action to regulate its use.
AI providers are increasingly facing a legal and regulatory backlash. 2023 saw EU policymakers pass a major new law regulating AI. It focused on areas of significant risk, such as the use of AI by governments and companies, and for critical infrastructure such as energy. It forces new transparency requirements on cyber companies, while limiting the use of facial recognition and deepfakes. President Biden passed an executive order on AI, forcing tech companies to share test results and setting new standards for biological synthesis screening.
ChatGPT owner Open AI is currently embroiled in a major lawsuit with the New York Times, which accuses it of using its copywritten content to power its AI engine. This may force AI companies to build partnerships with content creators and ensure more careful use of data. Overall, there is a sense that policymakers were too slow to regulate social media, allowing harms to flourish. They are unlikely to make the same mistake with AI.
