Why You Need To Know About AutoGPTs
The Next Wave Of Generative AI
Do you ever have that feeling after reading or watching something that everything is about to change?
I had that feeling last year when I first read OpenAI’s announcement post on DALL-E 2. That’s actually the moment for me when I suddenly shifted my technology gaze toward this thing called Generative AI.
Since then, we’ve seen an explosion of innovation in generative content creation, with transformers that output text, images, code, and video. And no doubt, this content creation space will continue to progress rapidly.
But I had another one of those feelings several days ago.
And it wasn’t from a company blog, an event, or a release announcement. It was in my Twitter feed.
What I was witnessing was post after post of developers experimenting with something that is very new and different.
It was about the creation of autonomous AI agents, currently referred to as AutoGPTs.
The Rise Of AutoGPTs
AutoGPTs leverage OpenAI’s large language model, GPT-4, to automate multi-step projects that would have otherwise required back-and-forth interactions with GPT-4.
In simpler terms, AutoGPTs enable the chaining of thoughts to accomplish a specified objective and do so autonomously.
They have internet access with the ability to read/write files, and they possess short-term and long-term memory to understand what has been done.
To be clear… the goal of AutoGPTs is to “make GPT-4 fully autonomous.”
The three most popular AutoGPTs can all be found on Github, they are as follows:
- BabyAGI by Yohei Nakajima
- Auto-GPT by Toran Bruce Richards, who goes by the alias Significant Gravitas.
- Jarvis by Microsoft.
In order to run these popular AutoGPTs, you need to install them locally, and that requires some kind of programming experience since it runs on Python and requires OpenAI and Pinecone API keys.
However, there are already AutoGPT apps that run on the browser, like AgentGPT and Cognosys. BabyAGI can also run on a browser via this Hugging Face space.
Since the experimentation is currently develop-centric, we should expect AutoGPTs to become more user-friendly soon, with clean and intuitive UI interfaces.
Getting back to my liminal Twitter experience, the following example is just one thread of literally dozens that have been lighting up the twitterverse. This thread shows some of the experimentation happening around AutoGPTs…
A Move Towards Action-based AI
The AutoGPT movement is part of a bigger trend in generative AI that is moving toward action-based AI agents.
For example, Adept AI is a machine learning research and product lab that is developing AI models to automate tasks and turn text commands into actions.
A move toward action is also being sparked by OpenAI’s release of ChatGPT plugins which transforms chat from a basic communication tool into a service that curates action-based experiences.
For example, Expedia is releasing a travel advisor plugin that not only provides recommendations for destinations, flights or hotels, but will also book directly from the recommendations as well.
cool right?
What about an AI doctor app that can not only respond with medical suggestions but can also recommend specialists near you, book appointments, and manage your insurance-related questions and tasks?
This takes chat to the next level and turns AI into assistants working for you.
Final Thoughts
What makes AutoGPTs amazing is once you give them a mission, they go off and autonomously perform that objective for you. They figure out the tasks and steps required.
This is one more step closer to AGI (Artificial General Intelligence).
And while there are talks about an AI pause and government bans on ChatGPT, the next wave of generative AI is surfacing and it’s not from private or company research labs, but from Github repos and Twitter feeds.
If you liked this article, throw out some Medium love… claps, comment, and be sure to follow.
You can also support my work on Medium by becoming a member using this referral link.
