What is ChatGPT's Atlas? Charlie Marchant explains OpenAI's new AI-powered browser, its "Agent Mode" capabilities, and the pros and cons of using it.

Click on the whiteboard image above to open a high-resolution version!

Welcome to Whiteboard Friday. I'm Charlie from Exposure Ninja, and today I'm going to be talking about ChatGPT's Atlas. I'm going to take you through what it is, why it's different from other browsers, how it works, and the pros and cons of actually using it. 

So ChatGPT's Atlas is an AI-powered browser from OpenAI. 

And there are two special things about it. Our first, it integrates a ChatGPT sidebar so that you can have conversations on your usual ChatGPT account about what you're browsing on the internet. And the second is it integrates something called agent mode, which can actually carry out tasks on your behalf. 

So how does ChatGPT's Atlas actually work? It uses a core architecture called O.W.L., and what O.W.L. does is enable the browser and the LLM integration.

So let's say that you're searching for a new pair of running shoes so that you can regularly do parkruns and 5ks. You might have already had a conversation with ChatGPT before about your running, about your foot shape, about your goals or ambitions, for example.

Then you actually make a search on the ChatGPT browser, on Atlas, for the best running shoes. But perhaps that's not going to give you the exact result that you want to have to find the right pair of running shoes for you. So you're going to interrogate a bit further.

What you can do is open the sidebar and actually have a conversation with ChatGPT about what you're browsing. 

Let's talk a bit about how that actually works. 

User input

So that's where the user inputs. It's where you input your query because you're trying to find the best pair of running shoes for you personally, not just what ChatGPT's browser thinks is the best pair of running shoes. 

Context understanding

So then what happens is there's a context understanding that goes on. It analyzes what your current browsing session is, and that could be multiple tabs. 

Now just so you know, the model can see what you're browsing only with your permission. So you can switch this on or off in ChatGPT's Atlas at the moment. 

Information retrieval

Then what happens is there's an information retrieval. If ChatGPT feels like it already has enough information in its local context, it can use that to answer your query directly. But more often than not, what it might do is use its retrieval APIs in order to search for newer, up-to-date information to give you a more complete and better answer, specifically to what you've asked. 

Reasoning and answer generation

Then it moves into reasoning and answer generation. 

The LLM fuses multiple things -- the page context of what you've been looking at on the browser tabs in ChatGPT Atlas, the user intent, what it thinks you're actually asking for to try and give you the answer that is most relevant to you, the retrieval of the supplied sources if it had to use the API to search live on the internet for answers, and relevant memory.

So it can look at your past chats and conversations that you've had in ChatGPT and decide if they're relevant to also add into its answer. 

Output 

Then what it does is formulate it into an output, and that output is either going to be a direct answer in text format, just like you get on ChatGPT when you use it in its own tab, or it might be a multistep agent mode set of tasks that then get done on your behalf.

That's particularly useful, for example, if you're doing something like flight booking or holiday research, or trying to find the cheapest tickets to a gig, and you don't want to do all of those tasks yourself the manual way. 

So there are some things about ChatGPT Atlas that are great, and there are some things that are not so great. And we're going to have an explore of what those are.

Some of the great things about ChatGPT Atlas are that it has this embedded chat user interface. You can ask questions live about what you're actually browsing on the internet, and you can ask it to do things like create summaries or use that information in different ways.

Secondly, it's aware of the context. Compared to using ChatGPT in a totally separate browser or tab, in ChatGPT's Atlas, ChatGPT can see what you have open on all of your browser tabs and interrogate that information with you. It means it's much more contextually aware of what you're probably trying to achieve with your searches.

The third thing is it feels faster than traditional search, and that's because it can help you move down a purchase journey much quicker by having the conversations back and forth.

And lastly, agent mode can actually do some tasks for you, which ultimately we'd all like to make our search a bit easier. 

But what's not so great? ChatGPT's Atlas is quite a new browser, and at the moment, there are some data security and privacy concerns, which I would really urge you to look at before you make the switch to using Atlas.

Secondly, it's still in development. And when things are in development, there can sometimes be glitches, and we see that with agent mode as well. There are times that agent mode doesn't fully complete the tasks and actually requires user intervention as well.

And fourth, it's currently only working on Mac operating systems for now. But we think that it will roll out further.

So this has been my conversation about what ChatGPT Atlas is and how you can actually use it. I'm Charlie Marchant from Exposure Ninja. You can find me on LinkedIn anytime. Have a fantastic rest of your Friday.