Building AI-powered meeting-summarizing apps seems to be one of the ways that Amazon CTO Werner Vogels spends his time. In a recent blog post, Vogels introduced Distill, an open-source app developed by his team at the Office of the CTO. Distill takes audio recordings of meetings and transcribes them, generating a summary and a list of action items. It can even integrate with platforms like Slack for easy sharing.
What sets Distill apart is its reliance on various Amazon products and services. AWS Transcribe handles the transcription, while Amazon S3 provides storage for the audio files. The summarization process is powered by Bedrock, Amazon’s generative AI development suite. By utilizing these resources, Vogels demonstrates the capabilities of Amazon’s technology.
So why create another meeting summarizer when there are already countless tools available? Vogels likely saw it as an opportunity to explore and experiment with his resources and spare time. As the CTO of Amazon, he has the luxury of pursuing hobbyist programming projects like Distill. In fact, he’s even considering porting the code from Python to Rust, showcasing his dedication to continuous improvement.
One interesting aspect of Distill is its flexibility in choosing the AI model for summarization. By default, it uses Sonnet, a mid-range model from Anthropic’s Claude 3 family. Given Amazon’s significant investment in Anthropic, this choice makes sense. However, any model hosted in Bedrock can be used, including those from Meta, Mistral, AI21 Labs, and Cohere. This flexibility allows users to experiment with different models and find the one that suits their needs best.
Vogels acknowledges that AI is not infallible and that mistakes can occur in the summaries generated by Distill. He emphasizes the importance of critical thinking and human involvement in the process, stating that while the app speeds up their processes, it still requires discernment. However, one could argue that the need for human intervention in summarization diminishes the purpose of an automatic summarizer. Nevertheless, Vogels remains supportive of the technology his employer provides.
In conclusion, Werner Vogels’ creation of Distill showcases his passion for exploring new technologies and utilizing the resources available to him. By building an AI-powered meeting summarizer, he highlights the capabilities of Amazon’s products and services. While the app may have its limitations, it serves as a reminder that AI is not a perfect solution and that human involvement is still crucial. Vogels’ commitment to innovation and his employer’s technology solidifies his position as the CTO of Amazon.