It’s hard to imagine a technology world before Generative Artificial Intelligence (GenAI), yet it was only a couple of months ago when we were all first introduced to ChatGPT and Bard. As with many new technologies, these powerful tools immediately set off a wave of blue sky thinking about the opportunities they may present.
GenAI is still in its early stages, but it could transform how we interface with our preferred devices or favorite applications. As one vice president of product recently quipped, “Gen AI is my new best friend.”
→ Given the nature of the event, we brought onboard our cloud partner, Google Cloud, who generously enabled access to their enterprise GenAI, large language models, and AI toolkit for participating teams.
→ Open Source have a long, rich history helping the people achieve more in the flow of work.
→ Quality and accuracy of AI applications are significantly influenced by the data and information used to train and benchmark them. In the case of GenAI and LLMs, which are trained on a broad corpus of information, they can also be fined tuned to specific use cases using more specialized data sets.
→ As with many hackathons, teams were encouraged to rapidly build prototypes and share the result of their work with each other. Collaboration was encouraged both in-person and over Microsoft Teams.
We used this opportunity to engage the broader organization on security best practices and ethical AI. There are many well publicized examples of companies’ code being shared in the consumer versions of GenAI such as ChatGPT that have led to security leaks. We have also seen examples of biases being created by the choices of training data. Some of our event educational webinars have explored these issues and gave participants the opportunity to reflect on practices that will be essential to leverage this new capability moving forward and reinforce our commitment to using GenAI in an ethical manner.
Finally, we explored the economic impact of using LLMs. These models are compute-intensive and have the potential of changing the economics of SaaS businesses if not designed properly. We explored enterprise needs to operate multi-corpus and multi-model environments that are fit for purpose and can leverage feedback to avoid model drift. These are difficult engineering problems, but I was excited by the creativity and resourcefulness of our teams.