Model agnostic AI Agents
The launch of Google Gemini is another reason why you should consider other options.
You are allowed to change yourself at any time. You can change beliefs, study new fields, and “start over” at any time. While some are happiest without change, I like to believe that most people want to improve, and continue to reach for something greater. This is a luxury we have as humans in the Golden age of information access. We can download books in an instant, meet in a virtual office, and connect with others like us through social media. We aren’t trapped with one toolset, and neither should your AI agent.
You may have seen OpenAIs new feature “create a GPT”. IBM Watson X lets you train your own models on your own data. Anthropic Claude2, Meta’s Llama2, and xAI’s Grok, provide alternatives to chatGPT. Google Bard has largely been a disappointment as many eagerly waited for Google to release something truly special. My team Magick ML has been talking Gemini with the Google team since April. We have been preparing for the model agnostic agents that you don’t know you need, and frankly, we like Google.
The next wave
As AI, and what LLMs can do, becomes more commonplace, more people are realizing that AI can help them get more done, or free them to work less and live more. We have been gearing up for the age of the AI agent, as we rely on these virtual beings to perform tasks, and coordinate other task agents on our behalf. At some point everyone will have an AI extension of themselves, their very own custom tailored version of Siri, Alexa, or Google Assistant.
Google has been AI-first since 2017. I for one am honored to have been in contact with them at all. You’re interested in us? I asked myself. I told my team this is not normal and that we must remain hungry because eventually more would catch on and we would see more competitors enter the landscape. Today, much like the Metaverse, the future we saw coming is here, and all the bigs are drawing their lines in the sand.
I hate those fucking lines in the sand. I am a capitalist, but above all, I believe in interoperability and freedom of choice. Most people have not been obsessing over AI agents for very long. The giants know this and want to capture market share. We want to capture the giants, and put them into little nodes like lego pieces that you can assemble and swap out on demand. Unless you want to buy a new AI agent every year like you do your phone, you will want these little lego pieces to be readily available and easy to use.
AI agents are meant to be useful, not spam
Today, our Magick agents are solving real user problems. One is fielding inquiries for a refund issued to their users. Answering tough and sometimes emotional questions kindly, professionally, and with care. Another fields complex questions about blockchain, current prices, and details about the company. We have more that can spin up markdown architectural diagrams, assist with grant writing in your own voice and so much more.
We see a world where artists can bottle their skills into an agent, and license it to the world. You can get AI agents with certified seals of approval from your favorite writers, painters, and singers. Imagine making music with Grimes, writing a novel with Stephen King, or crafting art in the certified style of Hayao Miyazaki. Imagine having both the Worlds knowledge AND the Worlds talent at your fingertips.
As Gemini rolls out this month, we will be working swiftly to integrate as much of these nodes as possible. Gemini is multi-modal, meaning it can work with images AND text, video AND tasks, Calendar AND email. We want to pair models like Gemini with useful tools like Zapier, AWS, and Gitlab. We want to create a world where you can get rewarded for creating agents, and integrations, that solve real users problems. Ultimately, we want it to be as easy as having a conversation, to spin up complex AI agents that you can harness, and trust.
See creating AI agents is one thing, but deploying, scaling, and monetizing are really hard. If you choose the wrong walled garden, it gets more difficult to leave. Anyone who has ever gone through a website modernization effort, migrating from one content management system to another, should know the pains of choosing the wrong walled garden. Anyone who has ever built their own CMS knows the pains of rolling your own. We believe a modular, open ecosystem, will drive down complexity and increase what is possible.
The future is bright, take action
Eventually your agents will have access to the marketplace. They will be able to suggest new integrations and skills, and even swap our language models on demand. You will say what you need, and the agent will find the best skills for the job. Like drawing on an endless book of spells, your agents will be capable of practical magic. Pretty soon you will be able to run an entire LLM on your phone, locally. This means you can have encrypted access to AI agents without feeding the models upstream. This means people in remote places without internet will even be able to benefit from these technologies. Better yet, one day we may not have phones, but simply compute devices for our AI agents. A lot more is coming in 2024.
Thanks for reading this far!
If you're interested in staying up-to-date on the latest Metaverse Interop news and insights, sign up for our newsletter today. In addition, don't forget to subscribe to my YouTube channel, the Metaverse Interop Show, for even more content. Whether you're an avid reader or prefer video content, we've got you covered.
I hope that you will consider trying Magick ML, and that you will explore new models like Google Gemini. I hope you are excited about the future we all get to share.