The Stochastic Mindset
AI tools will change the way we work by changing the way we think.
Most descriptions of AI’s potential to change how we work are by now familiar: AI will automate repetitive tasks, increase worker impact, improve job satisfaction and disrupt entire industries. But how will using AI tools change the actual experience of work? Will AI change the way we think?
AI is the next era of computing, and one of its hallmarks is moving from deterministic to probabilistic or “stochastic” outputs. As AI spreads, workers are engaging with a new software paradigm that calls for a new mindset. In a recent episode of Sequoia’s Training Data podcast, Dust co-founder GabrielHubert introduced the concept of the “stochastic mindset” as “the biggest shift in the use of the tools that we have since the advent of the computer.”
The stochastic mindset moves us from having minimal leverage on a task and 100% certainty of its outcome to 100% leverage on a task and far less certainty on the exact manifestation of its outcome. This transition is the difference between doing something yourself and delegating a task to someone else.
Randomness and uncertainty are a part of life, but in the context of modern knowledge work we have been trained to optimize for certainty and predictability. The stochastic mindset invites us to move from rote workflows to iterative development of tools, content, strategy and more.The stochastic mindset moves us from having minimal leverage on a task and 100% certainty of its outcome to 100% leverage on a task and far less certainty on the exact manifestation of its outcome. This transition is the difference between doing something yourself and delegating a task to someone else.
AI as exoskeleton for work
The shift to the stochastic mindset shows how the impactful productivity gains from AI will actually be achieved. By making it easier to access and synthesize information, workers will consume and produce much more information. OpenEvidence, Harvey and Dust are examples of products that take the friction out of accessing and making use of relevant information in the context of doctors, lawyers and knowledge workers more generally.
AI expands information through generalization. But it also reduces by summarizing information. In the AI age, there is uncertainty in exactly the message being delivered in exchange for speed and leverage. The quantity of information itself requires probabilistic approaches to manage and infer simplifications.
The pace of progress in artificial intelligence is incredibly fast. Unless you have direct exposure to groups like Deepmind, you have no idea how fast—it’s growing at a pace close to exponential.
— DAVID TISCH
Inverstor, Box Group
New York Times
The razor between the vines
2021
AI tools provide drafts or suggestions, not definitive answers. They will improve—but always to the point of some probability. Those reading—especially the quantum and stats enthusiasts—will be quick to assert that everything is probabilistic. Fair. But the history of computing has been one darn near determinism. AI is reaching a threshold of scale where being probabilistic is more efficient than being deterministic.
Founders need the stochastic mindset
The shift to the stochastic mindset shows how the impactful productivity gains from AI will actually be achieved. By making it easier to access and synthesize information, workers will consume and produce much more information. OpenEvidence, Harvey and Dust are examples of products that take the friction out of accessing and making use of relevant information in the context of doctors, lawyers and knowledge workers more generally.
Most importantly, by developing their own stochastic approach, founders can solve their customers’ problems too. By infusing the stochastic mindset into their products, founders can help users take advantage of these new capabilities:
- For the former, we adopt the emerging state-of-the-art approach of diffusion training combined with transformer models inspired by advanced large-language-models (LLMs) in order to train an autoregressive model that:
- Providing an initial image outside the distribution of the model can lead to unclear results.
- Following an in-depth sensitivity analysis on different configurations of the architecture alongside the data and model size.
- We hypothesize that the majority of these aspects may be addressed through scaling of the model and the datasets.
Giving humans more time to think
The shift to the stochastic mindset shows how the impactful productivity gains from AI will actually be achieved. By making it easier to access and synthesize information, workers will consume and produce much more information.
The shift to the stochastic mindset shows how the impactful productivity gains from AI will actually be achieved. By making it easier to access and synthesize information, workers will consume and produce much more information.
