The Emerging AI Food Chain
The introduction of generative models with unexpected emergent behavior has captivated the world’s attention, as evident by a surge in Google search and the slew of tech companies pivoting to become “AI first” companies.
That said, the flood of recent news in generative AI can be overwhelming, even to those who actively follow the field. The breadth and depth of the developments signify a seismic shift in the structure of the tech industry and society at large, with new personas (e.g. prompt engineers) and sectors that are spawning organically.
It’s worth zooming out from all the news and view the forest instead of the trees. Let’s break down this emerging “AI food chain” and analyze them conceptually by comparing them to earth’s natural ecosystem.
Digging deeper, lets expand upon that analogy and describe the key players in each group of the food chain.
| Natural ecosystem | AI ecosystem | |
|---|---|---|
| Ingredients | sunlight, water, carbon dioxide | data, compute, algorithms |
| Process | Plants capture energy from sunlight to produce oxygen (O2) and chemical energy stored in glucose (a sugar) via photosynthesis. | Academic / industry research labs “incubate” data, compute and algorithms to produce models via research and development. |
| Producers | Photosynthesizing organisms (autotrophs) such as plants and algae that capture energy from sunlight to produce oxygen (O2) and chemical energy stored in glucose (a sugar). | ML experts such as researchers, software engineers, data scientists, and computer scientists that capture flops from compute clusters and data to produce model weights and “energy” stored in the form of large language models or latent diffusion models. |
| Primary Consumers | Herbivores such as deer, mice, and elephants, that feed on the output (i.e. glucose) of the producers to sustain and grow. | Domain experts such as journalists, lawyers, physicians, bankers, etc, who consume the large ML models to drive decision-making processes in their respective fields, and to create new derivative products and services. |
| Secondary Consumers | Omnivores and carnivores like snakes and Komodo dragons that consume the outputs of primary consumers as their daily diet. | Lay users such as non-tech savvy citizens, kids, or seniors, that benefit by using the goods and services produced by primary consumers. |
| Decomposers | Fungi and bacteria that play a crucial role in breaking down organic waste into nutrient-rich soil, which sustains the producers. | Bots and data scrapers that collect publicly available data byproducts generated by upstream consumers and producers, then transform it into nutrient-rich datasets that can in turn be used by producers to enhance the large models. |
Just as in a natural ecosystem, the AI ecosystem is driven by a complex web of interactions between producers, consumers, and decomposers. By contextualizing the players in generative AI into this “AI food chain,” we can begin to navigate this sea of recent developments and understand where each new technology or company fits into the overall ecosystem.
And just like the natural food web, the interrelationships within the AI food web can be so intricate that a disruption to one component can cause a domino effect through the rest. While it’s not immediately obvious how actionable this kind of systems thinking can be, I do think we can map the rules and behaviors of the natural ecosystem to the AI ecosystem, and by embracing this analogous approach, we can better understand how new entrants into the ecosystem could affect the incumbents.