My circus, my monkeys: calling on Bard to make sense of one chaotic email inbox
My circus, my monkeys: calling on Bard to make sense of one chaotic email inbox
✉️ Bard launches an integration with Gmail, let's see what it can and cannot do
My Gmail inbox is a crossover of an air traffic control room and a circus.
Everything goes through it: event reservations, plane tickets, doctor appointments, e-commerce, newsletters, urgent marketing email about something somewhere being 10% off.
Is there anything in my Gmail I should be following up on? Who knows, no way to tell. It’s such a happy mess that it could definitely benefit from some AI assistance. Can we use a Large Language Model to help out?
Bard, meet Gmail
Earlier this week Bard (an experimental AI chatbot by Google) introduced a new capability to interact with Gmail and other Google Workspace services. Let’s give it a try and see what it can do!
If you want to follow along, visit Bard, click on the Extensions icon in the upper right section and enable “Google Workspace”:
Asking about a specific conversation
Let’s start with the basics, asking the chatbot about the status of a specific task that we know is tracked by some emails somewhere in our inbox:
✅ So far so good!
The 5 email limit
Below the answer itself, the UI gives us a hint as to how the response was generated, listing the emails “used for this response”:
In my experiments, no matter what query I make, the results are always based on at most 5 emails. That suggests that the integration likely works like this:
The LLM reads our initial query. Behind the scenes it generates a Gmail search query to retrieve the emails that would be related to it
It picks the 5 top resulting emails and feeds it back to the language model, prompting it to answer the original question
This means that the tool can only handle queries that can be answered by reading at most 5 emails, and only if the right emails can be identified.
Asking about many conversations
Is there anything in my Gmail I should be following up on?
Pretty good answers, but the catch above applies: as far as we can tell, Bard only uses 5 emails to answer the question. We may have 🔥 200 exploding unanswered crisis-mode conversations that the tool did not surface, so don’t be too trusting in the response.
🗺️ Gmail + Maps, shake hands
Text queries are nice, but we’re visual creatures, so let’s close with a map. This week Bard also got an integration with Google Maps, so could it bridge between Gmail and Maps to draw a map based on emails in our inbox?
Yes it could!
I love that this just worked!
We start to get a glimpse of the future in which AI systems based on language models can juggle multiple data sources and tools (the web, emails, etc.) and bridge between them to answer our questions. Not just via text, but also by making maps, charts and whatnot.
In other news
it’s our 20th weekly post 💫 ! If you have thoughts on how the newsletter is going I’d love to hear from you, add a comment below or hit reply and send me an email ❤️
70% of Gen Z use ChatGPT while older generations don’t get it, as per study from Salesforce
📝 How to Blow Up a Timeline → fascinating long-form read on the troubles at Twitter. “Usually, social networks are killed off by something exogenous. Twitter went out and bought Chekhov’s gun in the first act and used it to shoot itself in the foot in the third act.”
Postcard from Paris
Warm days seem to be over in Paris, took this one on one of the last summer evenings. Stay warm and have a great week 💫!
– Przemek