I wanted to share a little bit about the philosophy for why we built Supercog.
We believe that Language Language Models (LLMs) are power tools that offer amazing capabilities to people that know how to use them. But so far these tools are mostly helping the software engineers. We set out to help everyone else get leverage out of this technology.
We don't think AI is ready to replace anyone's job, and frankly we aren't interested in replacing jobs. But we do think that AI systems can be offloading a lot of dreary tasks that absorb some portion of everyone's work day. Searching for information, collating and summarizing it, moving info from point A to point B, visualizing info, or delivering the right info to the right person at the right time - these are all tasks that people spend some amount of time on, and some people spend LOTS of time on some versions of these tasks.
Your Digital Office Assistant
The language abilities of the LLM makes it amazing for dealing with unstructured data like emails, and for understanding instructions in plain English. If you can put an LLM into your organization with access to lots of information, it can be very useful at quickly digging up the answer to lots of questions. And this has been the initial "killer app" for LLMs in the enterprise - "like ChatGPT but with my private company info".
But beyond answering questions, the frontier (the "leading edge") models can also dissect and analyze data from documents. You can ask your LLM to summarize info from long docs, or to extract useful facts like interpreting a chart or scanning through the rows of a spreadsheet.
And thus talk to my docs and deeper analysis are becoming additional popular use cases for using LLMs at work.
The move to agents
The reasoning capabilities of the LLM is what really triggered our early excitement about building this platform. It turns out that an LLM is quite good at devising a plan to solve a task, and then executing that plan using a set of tools that you give it to take action. With these capabilities we realized that Supercog could actually take action and perform work on your behalf, simply by following the English instructions that you give it.
All we had to do was to give the LLM a large set of tools, and it would be able to accomplish real work at the office. Because where do most "knowledge workers" spend their time during the day? They spend a lot of it using the various SaaS tools of the modern workplace. Google Docs/Office 360/Email/Slack/Salesforce/JIRA - this is where a lot of modern work gets done. So if we can connect our "LLM powered" agent to these systems, then it should be able to accomplish real work for us.
So this was the vision for Supercog, to unleash the power of the LLM - safely! - within the enterprise to give every person the chance to use these new "bicycles for the mind" as Jobs would put it.