How are enterprises using GenAI or even trying to use Gen AI? Here are 5 big ideas from a report by Scale AI team
Open source models usage is increasing; Closed-source is still winning though
Model preferences continue to evolve and remain a key decision for an organization’s AI strategy. The largest increase in usage came from closed-source models with
86% of organizations using these models compared to 37% the year
prior. This is likely due to a combination of factors. Many organizations have existing contracts with cloud service providers who in turn have partnerships with closed source model developers, making usage of closed-source models easier. Many closed-source models also outperform open source models out-of-the-box. Despite that, open-source model usage still increased from 41% to 66%. This is likely due to the flexibility opensource models provide for fine-tuning and hosting.
Operational efficiency is the primary driver
61% of organizations stated improved operational efficiency as the leading driver behind adopting generative AI. Improved customer experience came in second at 55%. Despite growing adoption, there are still a number of challenges that stall widespread use of generative AI. 61% of respondents cited infrastructure, tooling, or out-of-the-box solutions
not meeting their specific needs.
Processes like RAG and fine-tuning introduce the complexity of integrating external data sources in realtime, ensuring the relevance and accuracy of retrieved information, managing additional computational costs, and addressing potential biases or errors.
And the fav model is..?
Deploying and Customizing AI Use Case
Coding copilots are becoming mainstream with technical users being early adopters of solutions like GitHub Copilot, CodeLlama, and Devin. Model vendors have responded to demand for content generation with prompt templates that guide users to effective content creation questions for functions including Marketing, Product Management, and Public Relations.
You must be logged in to post a comment.