AutoGPT's – Everything you need to know
What exactly are AutoGPT's? Autonomous agents are here, and most likely to stay.
** Disclaimer, this email is long and will be cut off, so make sure you click “expand” when you get to an abrupt end somewhere in the middle.
👋 Hey, I’m Linus and welcome to a 🔒 subscriber-only edition 🔒 of my weekly newsletter. Each week I write about all things AI. I cover your questions, write op-eds or create AI tutorials.
Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 13k subscribers?
Today’s newsletter is sponsored by LoopGenius
🔥 Transform Your Side Hustle with LoopGenius 🤖
Struggling to market your side project or business?
LoopGenius has your back!
🚀 Fast Landing Page: Turn ideas into reality in 30 secs
🎯 Tailored Loops: Master Twitter, LinkedIn & more
💡 Micro-Influencer Magic: Expand your reach seamlessly
✍️ AI-Assisted Content: We write, you press "send"
🎁 Unlock your marketing superpowers - Try LoopGenius FREE!
What exactly are AutoGPTs?
They are a new wave of improvements to how we interact with LLM, like GPT4. They are designed to run autonomously. Enabling the creation of new self-powered agents that complete tasks for you without human intervention.
Sounds like science fiction? Sure thing, but it’s already here. You can use AutoGPT’s today, or build your own.
The most popular AutoGPT’s at the time of writing are:
→ BabyAGI by @yoheinakajima
→ AutoGPT by @SigGravitas
→ Jarvis by Microsoft
So whats the fuzz all about?
The adoption rate of these new techniques is through the roof, for example AutoGPT’s GitHub repository has +75.000 stars (at time of writing). There are close to 200 contributing developers to the repository. Over 12k forks, and +300 issues reported. For the non-technical person, Github is a large website, where anyone or any company can host their code. And people can work together on large projects without destroying each others contributions. (Thats the easiest way I can describe GitHub, don’t burn me).
So what makes these new techniques or paradigms so powerful? Well simply put, instead of you constantly driving the AI forward. Providing it with new instructions in between every command. It kind of figure things out on its on and reflect on its ongoing work as it executes. They are like self-managed AI with simple instructions or goals set by a human driver.
So what are some of the things AutoGPT’s can do?
First it has memory, both short term and longterm
It can access the internet, the agents can browse the internet for information to better be able to complete their tasks/goals
You can have many different Agents working together to solve a task
You simply assign a goal/task and the agent automatically will work on it until it thinks the task is completed.
So, with the above information, we can allow our minds to wander around the possibilities. For example, an AutoGPT could be tasked with going through all your emails, come up with responses for each and every one of them. Rank each email based on criteria you’ve set.
Then autonomously decide which of the emails it can go a head and respond to in your name, or which to put in a list for you to just read and confirm, and off they go. The agent can also be tasked with solving specific problems, maybe organise a bunch of inbound meeting requests, and have them synched with multiple participants, and make sure nothing clashes in your calendars.
This is just one thing an Autonomous Agent could do. There is just so much here that could be done that it’s hard to really comprehend.
I think Andrey Karpathy puts it well:
Another interesting take is from Dr Jim Fan
So why is this such an incredible innovative thing?
For starters, this could be the beginning of actual AI in the workforce. Anywhere from Virtual assistants, real-estate, virtual tutors, AI paralegals, customer support agents, tele health. Just to mention a few things where these agents could have profound impact.
There are a few things that can and most likely will happen: there will be people building these agents, and there will be people paying to use them.. If there is anything I’ve seen with AI so far. It is the fact that AI has solved hard problems and people are willing to pay. Never have I seen so many startups validating their business with paying customers as I'm seeing now.
There are some tools that don’t require you to build or run things locally on your computer; one of them is godmode.space – go give it a try.
Conclusion
We're in the early days of Autonomous Agents, and people are calling this the beginning of AGI, among other things. I’m going to be a bit more cautious. But I say this, the agents are here to stay. There will be more breakthroughs, given the sheer amount of talented people now trying to improve, adapt and increase the capabilities of the systems, we will see rapid firing improvements in the next couple of weeks and months.
We've just scratched the surface, and I'm increasingly leaning into the possibility that we might soon have a ratio of 1,000,000:1 agents vs humans. It’s like these agents are indeed the next version of software in strange way.
Today’s AI tools.
🤖 Pixiebrix.com - Highlight and summarize text with AI.
✨ Character.ai - Chat with any type of persona or create your own.
🎥 Runwayml.com - Everything you need to make anything you want.
My latest YouTube video
Would mean the world to me if you subscribed to my channel – will make exclusive videos that are only for newsletter subscribers as well as public videos.
Still here?
If you are still here, thank you. I hope you enjoyed this note on Autonomous Agents. If you have questions, feel free to reply to this email or reach out to me on Twitter or reply/comment here on Substack.
Share this post with anyone you think would be interested. Use the link below.
🙏🏻 Thanks LoopGenius for sponsoring this weeks newsletter.
FYI - the Loop Genius links are broken, just need to add a "?" before the first "utm" to fix.
Let us know when you fix the link, and I'll use that one instead of direct.