The Role of Data in AI

Data serves as a medium that communicates the state of the world to machines. At its core, any complex system boils down to a set of actors that communicate through data. Without data, machines can’t react to the physical world we live in.

Artificial Intelligence (AI), with its various subfields like Machine Learning, Deep Learning, and Natural Language Processing (NLP), exists to make sense of this data. Beyond catchy phrases like “Data is the new oil,” many startups are striving to build new business models around data and, more specifically, how to derive the most value from it.

However, discerning startups with a robust AI business model from those merely riding the AI hype is crucial. How can investors accurately value an AI startup?

The Three Buckets of AI Startups

AI startups generally fall into three categories:

1. Startups Solving Already Solved Problems with AI
   These startups leverage AI to address problems that already have solutions, often without adding significant value. They tend to use the AI buzzword for hype. These are generally not good investment candidates unless there is another compelling reason.
2. Startups Offering Significant Improvements
   These startups use AI to improve existing processes dramatically, offering a 10x improvement. An example is optimizing traffic dispatch systems with AI. These startups present a good investment opportunity with a potential for short-term ROI.
3. Startups Enabling Previously Impossible Solutions
   These startups leverage cutting-edge AI technologies, like Generative Adversarial Networks (GANs) or inductive graph reasoning, to do things that were previously impossible. They are typically strong investment candidates with long-term ROI potential.

Key Aspects to Investigate in AI Startups

Investors should focus on two main aspects when valuing an AI startup:

1. Algorithms
   In the fast-evolving field of AI, algorithms are continually being improved upon by major companies and academic institutions. Therefore, algorithms alone are not a long-term defensible competitive advantage. Investors should ensure that the AI team genuinely understands the algorithms they are using and can explain why their specific approach is suitable for their use case.
2. Data
   Access to unique data sources or innovative methods of obtaining data can provide a significant competitive edge for AI startups. Consumer AI startups that rely on data they do not own will struggle to compete with giants like Google, Facebook, and Amazon, which have vast amounts of consumer data.Investors should examine the properties of the underlying data closely. Important questions include:
   • Is the data stationary, or does its distribution change over time?
   • Does the data require human intervention for processing?
   • Does the data fit a specific distribution, such as a bell curve or a power law?

Additional Considerations

Besides algorithms and data, two other critical aspects are the composition of the AI team and the balance between fully automated and manually operated components within their infrastructure. These factors will be explored in further detail in a subsequent article.

Conclusion

Investing in AI startups requires a nuanced understanding of both the technology and the business models. By focusing on the true value-add of AI solutions, the uniqueness of the data, and the expertise of the team, investors can better navigate the promising yet complex landscape of AI entrepreneurship.

Stay tuned for more insights on evaluating the composition of AI teams and the automation levels in AI startup infrastructures.