I think we’ve got two waves happening right now:

Wave 1 is about the maturing of the LLM platforms. These include the foundation models, frameworks like LangChain and others, commercialization of these tooling frameworks, and more clarity for compliance teams to help
approve the right tooling for enterprise stacks. All this tooling is currently being architected into enterprise stacks.

Wave 2 is happening in the early adopter crowd, and this includes:

• multi-modal functionality (take a picture and ask it a question),
• text to voice (just talk to ChatGPT, or capture the way I speak),
• video generation.

All of these technologies are moving from research to early commercial products, but they’re not yet recognized and product market fit is incipient (What use cases? Which killer apps?).

“Strategies, Prototypes, and Challenges in GenAI Deployment”

Going back to wave 1, what I’ve seen much of is that most large enterprises are (have) refreshed their AI strategies to incorporate GenAI. They’ve also stood up task forces to accelerate AI, so this means weekly cadences with business, technology, security, legal, compliance, HR, etc. And everyone is building prototypes in order to learn the capabilities. You don’t code, you don’t learn. This requires heavy lifting from compliance departments.

So to summarize, large enterprises have setup strategy and task forces, to build prototypes and sort out their compliance. In the background, the tech companies in the forefront are moving multimodal models from the research lab to production. These have not impacted the market yet. There’s only so much the market can absorb at one time.

I think many large enterprises are at the beginning of the cycle of deploying GenAI at scale. So the economics of that is only now starting to be seen as a key aspect. Compute costs are high, and maximizing capabilities (longer contexts, faster latency) requires more compute not less. There’s a need to setup proper infrastructure that’s right fit in every enterprise with thoughtful economics (right resolution of the model which can 10x the cost up or down, proprietary models vs open source, cloud or alt cloud/on prem, etc).

Usual disclaimer: Views are all my own, not my employer.

What are you seeing in the market on GenAI today? Does this resonate from where you stand?