This weekend I started an exciting experiment. I build my AI agent which not a rule based agent such most of the Bank's chatbot.
I started to feed personality also into this agent, and as I'm running alone in my environment, I can also feed my personal data - such my Banking transactions, financial goals, current assets.
The most difference compared to ChatGPT - is that I'm sure that the data is only available for myself.
I can't wait to see how my experiment goes over time! I think the future is more about this direction - and it is practically not expensive.
With a simple rented infrastructure I can get it for 40€ per month.
Over time, the cost will radically decrease, especially if we move into the direction of multimodal models, where we can turn on and off model parts which required for the answers.
First I build it as a text based interface, but over time, I would like to make available a speech interface, where I can talk with the agent with my voice.
To enhance my productivity, I can load information via text interface and via automation - such my bank and investment transactions.
My expectation to have better insights to my data, and make my data cleanup easier such tagging transactions and enrich data such labeling merchants where I make transactions.
Finally the rise of the publicly available opensource models make available these for everyone who have minimal programming background.
Again - who uses tools what available can outperform others who still uses the old mechanical ways.