As of the 14th of November, seems like OpenAI is underwhelmed by the progress of their LLM. It could represent a slow down in the evolution and their quest to AGI. Nevertheless, what we got so far I believe is amazingly impressive and there are a few tools that have not yet been deployed to their full potential yet. One of the reasons for this is as Matt Clifford said: "There are no AI-shaped holes lying around". Companies' processes can be improved, but the leap in improvement is limited by the the way AI is being deployed. AI won't fill the hole of a process inefficiency. That's not how it should work. AI is here to create new processes. Instead of a step improvement, AI should review the whole goal of the process from the ground up, and a new process that leverages AI should come out. I would like to give an example: email clients.
At the moment email clients, working on a protocol older than HTTP, are still the main means of async communication in business. People communicate via email and the email client allows people to exchange data.
Email client with AI is an evolution of the usual email client, by the use of AI to improve the current process of reading and sending messages. Outlook for example allows you to read and summarize emails with AI copilot, write an answer or check the grammar.
A redesigned approach to message exchange that leverages AI should start from first principles: why do we use emails? Humans need to exchange information, learn and share data. We can isolate these 2 phases, learning and sharing. What are the best and most efficient ways of learning? It does depend on each individual. As a means of comparison, current email clients place text on a screen for you to read. An obvious evolution that can leverage AI is the creation of multimodal information delivery: audio, video, text. I personally really love the audio format as it takes my eyes away from the screen and I can listen to it while doing a light physical activity. So, can AI gather information and reshape it in a custom format for their users?
Sharing data: sharing starts with the creation of new data and then with the process of delivering the data to other people. Again the creation can be approached in a custom manner and while I would love to just chat back to the AI to create an answer to an email, sometimes I need to check and provide more accurate data via a spreadsheet or an image or a deck. However, the whole process can be sped up with the use of AI. In this world, does the AI become a CRM as well as an email client? Can it autopopulate data from internal information, email, external links like LinkedIn and search?
Apple Siri, Google AI and Microsoft CoPilot are fantastically palced to make this happen. Will they?
I am working to make this come true for a specific use case, Private Equity, hoping that whatever Microsoft or Apple come up with, won't be as well suited for them and their unique needs of financial reporting. Key for us will be speed of execution and data gathering, so that we can lock some clients in before being disrupted by AGI.
2024-11-15