How to navigate your engineering team through the generative AI hype

How to navigate your engineering team through the generative AI hype

Generative AI has been gaining significant attention in recent years and is now approaching the peak of inflated expectations in Gartner's Hype Cycle. As an engineering team leader, it is crucial to navigate through this hype and separate realistic projects from those that may not be anchored in reality. By peeling back the "how" and focusing on the "what" of the idea, you can discover projects with strong stakeholder support.

One approach to harnessing the power of generative AI is through the use of GPT models and other pre-trained models from sources like HuggingFace. These models can be "fine-tuned" with domain-specific examples, resulting in dramatic improvements in results. However, it is important to note that this process requires time and effort to curate a meaningful dataset for tuning.

When exploring opportunities in the generative AI value chain, it is essential to identify the two main categories of applications. The first category involves using foundation models with some customizations to build applications. These customizations can include creating a tailored user interface or adding guidance and a search index for documents to help the models better understand customer prompts. This approach allows companies to leverage existing models while still providing a unique user experience.

The second category represents the most attractive part of the value chain, where fine-tuned foundation models are used to deliver outputs for a specific use case. Unlike training foundation models, which requires massive amounts of data, is expensive, and time-consuming, fine-tuning foundation models can be completed in days with less data and lower costs. This makes it accessible to many companies looking to harness the power of generative AI.

To create proprietary data for fine-tuning, companies can implement feedback loops driven by an end-user rating system. This could include star ratings or thumbs-up, thumbs-down ratings to gather data on the quality of generated outputs. By leveraging user feedback, companies can continuously improve their models and provide better results over time.

As with any emerging technology, dedicated generative AI services will likely emerge in the future to help companies fill capability gaps and navigate the business opportunities and technical complexities associated with generative AI. These services can provide expertise and support to companies as they strive to build out their generative AI capabilities and create innovative applications.

In conclusion, navigating your engineering team through the generative AI hype requires a careful evaluation of realistic projects and a focus on the "what" rather than the unrealistic "how" of ideas. Incorporating fine-tuned foundation models and creating proprietary data through feedback loops can lead to successful generative AI applications. Here are three actionable pieces of advice to consider:

1. Start small and focus on specific use cases: Instead of trying to tackle generative AI on a broad scale, start with a specific use case that can benefit from fine-tuned models. This allows for more focused efforts and quicker results.

2. Invest in data curation: To achieve meaningful results with fine-tuned models, invest time and effort into curating a high-quality dataset. This will ensure that the model's outputs align with the desired outcomes and provide value to users.

3. Embrace feedback loops: Implementing an end-user rating system or feedback loop is crucial for continuously improving generative AI applications. By gathering feedback, companies can refine their models and deliver higher-quality outputs over time.

By following these actionable pieces of advice and staying grounded in reality, engineering teams can successfully navigate the generative AI hype and unlock the full potential of this emerging technology.