Welcome to 2024: Time to Start Kicking A$$ with AI - Bootstrap your multi-agent workflows the easy way with AutoGen Studio on Tanzu Kubernetes Grid
None of us will ever forget the impact AI had in 2023 ... getting access to ChatGPT I felt like Clark Kent busting out his cape, I felt like I could take on anything. And with enough time, I could have endless conversations with the chatbot and learn more than I ever thought possible.
After we spent some time talking to the out-of-the-box chatgpt model, we learned the model didnt know everything, and we could potentially add differentiated value by finding ways to get the bots to know about things they were not trained on. So we figured out how to feed our own data to the bots all without needing a PhD or massively expensive and complex model training by using retrieval augmented generation (RAG). At first to use RAG you had to get your hands dirty, there are tons of different knobs one can turn when trying to get a chatbot to answer different types of questions about different types of data, what is the optimal chunking strategy, how do you setup a vectordb, how do you automate the retrieval and insertion of contextual data in a prompt etc etc ... but by the fall of 2023 when OpenAI introduced the assistants API and the ability to create "GPT's", now I can just log in and click a button to upload whatever documents I want the chatbot to know about and thats it.
And I dont mean to give all the credit to OpenAI here as well before the assistants API was out scores of libraries to sdk's to simple free open source and commercial applications could let you chat with your pdf or other documents in simple and easy ways.
All of that is profoundly empowering and productivity enhancing ... if we werent all watching it unfold before our eyes it would be hard to believe that what we have seen so far is just the tip of the iceberg.