AutoGen is a framework that enables development of LLM applications using multiple agents that can converse with each other to solve task. AutoGen agents are customizable, conversable, and seamlessly allow human participation. They can operate in various modes that employ combinations of LLMs, human inputs, and tools.
Git repo here: https://github.com/microsoft/autogen
You must log in or register to comment.
I’d be really interested to work on these sorts of experiments myself but it kind of spoils it when there’s software giants with infinitely more manpower doing the same experiments.
You can always put your own spin on things. That’s usually a nice thing. Or take the results of a proper study and apply it to something specific. That’ll contribute something useful. Or if you already started, contribute your own findings. Confirm the results or ask the questions they missed. These are all left to do.
In case you want to do research without them interfering: There are subjects that are of little interest to the academic world at this point or the big companies won’t do it. (Or won’t reveal the inner workings.) You can choose one of those. For example I want to read studies about AI companions and personal assistants. Like Replika AI or the combination of Alexa and LLMs. I think sociology did a bit on this (and sci-fi movies), but the perspective of computer science and AI research is still lacking. And there is potential for innovation and features left to invent. And things to study scientifically.
And I think Microsoft and other companies don’t care about issues like that. They just publish the stuff. I haven’t looked at it properly, yet, but this framework is probably very similar to like 5 other frameworks that support Agent workflows. With a slightly different focus and spin on this idea.
Ah, thanks for broadening my perspective on the possibilities
Well, there is kind of a gold rush happening with AI. Lots of people doing all kinds of stuff because there’s a giant amount of money to be made.
So everyone and their grandma are working on AI. I think you absolutely need to change your attitude, find your niche, or jump on the hype-train and be quicker than them.
But I think it’s possible (maybe less so for the last thing). And I very much hope the free and open source community and the ‘normal’ people don’t get lost in the process. Because if AI gets more and more important in our world, and research and the market are completely dominated by the few big companies that currently invest huge sums of money… I’m afraid the world will get less democratic.
So please don’t be too discouraged and leave (all of) the field to them.
There’s plenty of smaller projects around that attempt to solve similar problems, metagpt, agent os, gpt-pilot, gpt-engineer, autochain, etc
Several would I’m sure love a hand , you should check em out on GitHub!
deleted by creator
It seems reasonably realistic if you compare it to code interpreter, it was able to recognize packages it hadn’t installed and go seek them out, I don’t think it’s outside the scope for it to recognize which module wasn’t installed and install it
Even now with regular models they’ll suggest the packages you install before executing the code they provide
deleted by creator
