- A powerful new class of large language models is making it possible for machines to write, code, draw and create with credible and sometimes superhuman results.
- Humans are good at analyzing things. Machines are even better at creating things.
- Up until recently, machines had no chance of competing with humans at creative work-they were relegated to analysis and rote cognitive labor.
- Now, machines are just starting to get good at creating sensical and beautiful things. This new category is called “Generative AI.”
The Chronology of generative AI
- Earlier, relatively small models excel at analytical tasks and become deployed for jobs from delivery time prediction to fraud classification, but are not expressive enough for general-purpose generative tasks
- Between 2015 and 2020, the compute used to train these models increased by 6 orders of magnitude and their results surpass human performance benchmarks in handwriting, speech and image recognition, reading comprehension and language understanding
Anatomy of a Generative AI Application
As applications get more user data, they can fine-tune their models to:
- improve model quality/performance for their specific problem space
- Decrease model size/costs
- Applications are great at spitting out multiple different ideas to get the creative process going
- Most AI demos are one-and-done: you offer an input, the machine spits out an output, and you can keep it or throw it away and try again. Increasingly, the models are becoming more iterative.
Eyes Wide Open
- Generative AI is still very early
- The platform layer is just getting good
- We don’t need large language models to write a Tolstoy novel
- Models are good enough today to write first drafts of blog posts and generate prototypes of logos and product interfaces
- There is a wealth of value creation that will happen in the near-to-medium-term
Better, faster, cheaper: Generative AI
- Compute gets cheaper
- New techniques, like diffusion models, shrink down the costs required to train and run inference
- Developer access expands
- Applications begin to bloom
- Killer apps emerge (Now)
Closing Thoughts
- Generative AI companies can generate a sustainable competitive advantage by executing relentlessly on the flywheel between user engagement/data and model performance.
- The platform layer of AI is just getting good, and the application space has barely gotten going.
The Market landscape
- Text is the most advanced domain.
- Code generation is likely to have a big impact on developer productivity in the near term as shown by GitHub CoPilot
- Image models have gone viral
- Speech synthesis is just getting good for consumer and enterprise applications
- Other domains: There is fundamental model R&D happening in many fields, from audio and music to biology and chemistry
- Media/Advertising: Imagine the potential to automate agency work and optimize ad copy and creative on the fly for consumers
- Applications: The short form and stylized nature of the verbiage combined with the time and cost pressures on these teams should drive demand for automated and augmented solutions
