The Lunar Perspective

Doom, Transformative, or Bubble?

Whether you are an investor or not, the question of where AI is heading is significant.

Probability of Doom – P(Doom)

How high is the probability that AI will doom humankind? We have already seen rogue actions from AI models penetrating systems without an explicit prompt to do so. These actions occur because the models are built on a concept called Recursive Self‑Improvement (RSI). RSI is designed for the model to improve itself by writing better code, improving its architecture, and designing better training models. It does this by setting goals.

If these recursive loops spin very fast and humans are unable to correct a model’s goals, this could result in rogue actions and potential havoc. A model could also escape from a test environment because of a potential flaw in the test environment, as was experienced recently when an OpenAI model hacked Hugging Face’s systems. Havoc could also be caused by bad actors who use AI enabled tools (weapons) to destroy their enemies or competitors.

Recent warning calls from executives and employees of major AI model companies have urged regulation to prevent a “doom” scenario or major havoc causing events from happening. Insiders have a very good understanding of the potential consequences of rogue AI enabled tools, and we should take heed.

Transformative

When AI recently made a major breakthrough with Large Language Models (LLMs), the mood was much more optimistic: AI is a revolutionary technology that can deliver major benefits—from finding new treatments for diseases more quickly, to substantially improving productivity across industries and professions, to discovering new businesses and exploring new frontiers.

Knowingly or unknowingly, most of us already benefit from using AI tools in our daily work. In our business at Lunar Capital, we see major benefits in utilising AI for research and analysis, while ensuring that ultimate decisions remain human‑led. With proper guardrails, AI can indeed be transformative and revolutionary. In a few years, we will likely be in awe of the impact AI has on both business and personal life.

The fear of job losses is also real. How policy makers, businesses, institutions and we as individuals respond will determine the full impact on society as a whole.

AI can have devastating consequences but can also have profoundly positive outcomes for society, businesses, and individuals.

Is AI in a Bubble?

As investors, we also ask: is AI in a bubble?

In our analysis, we break down AI investment into the following sub‑categories:

There certainly are businesses that are over-valued or who may not be able to cope with an unforeseen event, but the framework above allows us to identify groupings of companies that can benefit in more than one way from AI and those that have strong moats. An example of a company that benefits in more than one way from AI: Alphabet provides compute services through its datacentres, benefits from more optimised advertising for its clients, achieves efficiencies through AI generated code, earns revenues from its Gemini AI model used by it clients, etc. We are also invested in Amazon, ASML, Microsoft, Nvidia, CrowdStrike and specific companies that we view are building good use cases using AI.

Closing remarks

In this our 10th year since we started Lunar Capital, we remind ourselves of the significant events we have endured: Brexit, US-China Trade War, Covid crash, meme stock mania, regional banking crisis in the US and Switzerland, war in Ukraine and the Middle East, tariffs, China property crisis.

There are real risks and real opportunities in AI. Selected investments and right sizing for potential risks offer better chances to ride through a downturn or an unforeseen event in the market. This may even strengthen the positions of stronger players if weaker companies fail to endure.