Name: MultiOn
Uploaded: 2024-12-17T15:33:32.828Z
Duration: 2 min 33 s
Description: An AI platform providing autonomous agents that perform web-based tasks on behalf of users.

multion

An AI platform providing autonomous agents that perform web-based tasks on behalf of users.

MultiOn

In this video I'm going to be showing you MultiOn which is one of the most impressive personal AI agents that I've seen to date. MultiOn allows you to remove a lot of the monotonous work that you do throughout the day when you're using the internet. 

First I'm going to give you a little bit of a demonstration on how this works. I'm going to say book me a time slot tomorrow at 2: p.m. to do yoga and I'm going to click enter and pretty quickly you'll hear a voice speaking back to me. So my speakers turned up here which you should hopefully be able to hear. 

I'm navigating to the Google C under event creation page to book a time slot for a yoga session tomorrow at 2 p.m. In the event details for a yoga session tomorrow at 2 p.m. including the title start and end times and a description with a signature as instructed after the yoga session has been successfully booked for tomorrow at 2: p.m.

As you saw there, it went to Google Calendar. It filled out all the details to create that yoga session that I requested. And it did all this just from that natural language prompt that I had. You also saw there that the agent was speaking back to me in real time. 

If you plan on having it in the background while doing something else, that audio option is a really nice feature. You can see within this chat interface there's a handful of things here. So you saw me turn on the voice. You can also go and configure your agent. 

And there's also the ability where you can teach the agent. So say if we're using a platform that MultiOn might not be as familiar with as something like Google Calendar or Uber or DoorDash, something like that. Maybe you're using a proprietary CMS or a proprietary tool and you want to show MultiOn how to use it. That teach me option is intended to be a way to augment your agent's behavior within the interface. 

You can type a subsequent command here or you can hold the microphone icon and you can do something like book another yoga session at 9:00 a.m.

On Friday if I go back to the main interface here there's a number of different options. So you can go ahead and wish a happy birthday to a friend on Facebook. You can call an Uber from X to Z if you'd like. 

Now the thing with the autonomous mode that I found is right off the bat I didn't necessarily trust it to send an email out of the gate. When I was playing around with this I didn't want to just start sending emails that might have errors or send them to the wrong people because it does take these actions very quickly. MultiOn does really well on a lot of the popular websites that I found but it is general. It isn't just a gimmick you actually can go ahead and ask at things across the board. 

So if I just say research a email to draft about video sponsorships and if I just go ahead and enter that in here what I found interesting with this is observing how an agent like this performs research. When I perform research I might do it in a very particular way. I might go to Google. I might be biased to reading articles. I might not look at videos. I might look at certain resources.

Rather than others where what I found with MultiOn was really interesting is it takes a completely different approach than I would when researching a topic. What I found is that it starts to navigate the web in ways that I didn't even really think about. So if you see here on Google, it clicks a template. It's going to that first option there. 

And then what's really neat with this is you'll start to see it start to control your browser. So you'll see it scrolls down. It's giving you that information on the right-hand side here where you can either look and read through it or you can go ahead and turn on that agent speaking back to you at any time, which I'll turn on here. 

I'm scrolling further down the page to continue looking for additional sponsorship email templates and key points that should be included in a sponsorship request email as per the plan. The other nice thing with this is at any point you can go ahead and pause it. Say if this looks like a place where I can go ahead and minimize the agent here and get the information that I need, you can go ahead and do that or you can.

Go ahead and write subsequent commands or give audio commands back to it at any time. I think the latter half of the year is going to be really interesting with tools like MultiOn and the Rabbit R One device that's coming out where you're able to speak and it's able to perform actions with something like a large action model. 

Just to show you some of the other features within MultiOn, once you have it set up, it is a Chrome extension. You can go in here, set up your profile, say if you have your default details where you're typically filling out different things like your address. Maybe you're buying things, you can also include certain notes which is like your system prompt in something like Chat GPT where it weights these notes higher for the action that it takes. 

Say if you always like pizza without a particular topping or you prefer certain things like an aisle seat on an airplane, you can go ahead and put all your notes in here. Now it does give you the ability to toggle on and off the voice mode by default. And then there are a handful of options within the AI mode that you...

Can choose so there's standard Vision fast and ludicrous. Now full disclosure I'm not sure what the benefits of all of these are. I wouldn't be surprised if over time they just have a refined combination of all these things and they decide for you the best model for this type of agent. 

Then here you can also specify the trigger modes whether you want it to be autonomous or whether you want to have the approval for each of the subsequent steps that it's going to take. And then you can choose something like your language. The other thing that I thought was really neat is you can teach it skills for particular website. Say if you're on a particular URL you can give it a description and a skill name and then you can give it certain rules. Say there's certain things that you always want to do or always want to be mindful of. It gives you the ability to create these new skills per URL which I think is just really neat. 

In terms of the company itself I was reading through some of their blog post and it does really seem like the founders are motivated to alleviate us from all of that.

Monotonous work that we spend hours every week doing as they put here motivated to free up the human race from the mundane and boring TSS that we do that often take hours or days to accomplish. I'm excited with what the team is doing at MultiOn. I'm excited to see how this product evolves and eventually once it's released to a wider community. 

The other interesting thing is they do have a mobile app that's coming out that's going to be available in early 2024 on iOS. So that's going to take this to a different level where you'll potentially be able to navigate your phone autonomously with something like MultiOn. 

In terms of next steps for MultiOn, you can join their Discord if you like. You can also get connected with them on X. In terms of gaining access to this, you can sign up on their homepage here for their iOS app. 

For developers and businesses, if you're interested in signing up for API access, you can do that from their homepage as well. If you'd like to use the tool itself, I engaged with one of the founders on Twitter where I got a DM where I was able to eventually sign up and share the AI Agent Store with colleagues.

Get an email to download and install the Chrome web store extension, which is how all of this works. I'd encourage you to check it out. Follow them on Twitter. Great job to the founders and the technical team behind this tool. 

I don't say this lightly, but I do think that something like this is going to be how we interact with computers and devices in not so distant future. Getting used to something like this might take a little bit of time, but I imagine over the coming months and years that tools like this we're going to be much more familiar and within the public consciousness of how we interact with computers and devices. 

That's it for this video. If you found this video useful, please share the AI Agent Store with colleagues. Otherwise, until the next one.

Note, that there may be slight differences in text and video. Watch video for the original content.

MultiOn Video