On using AI-enabled browsers
I use a variety of browsers, and always like to try something new. But not all that is new is good.
Although I have been using Apple products for many years, I never bonded with Safari; I found it clunky and unfriendly. Similarly, having used Firefox for a long time, I eventually moved away from it when I found better solutions.
At the moment, I use several browsers for different tasks:
- For everyday quick browsing, I use Brave. It is quick, safe, and has a nice, clean user interface. I also like the company’s approach to safety, although I am sceptical about their push for AI (Leo).
- For the University business (teaching, admin, finance), I use Edge, but with CoPilot switched off. Firstly, it is the recommended browser for our University, and is quickly accessible in lecture rooms for access to slides. Secondly, although I use Office web apps only occasionally, some only work properly on Edge. Thirdly, it is actually a solid, quick and reliable browser (with CoPilot off!).
- I use Google Docs a lot for teamwork. Many of my colleagues and I found OneDrive/ShareDrive hopeless; difficult access management, lost files, bloated interface. For any Google Drive-related work and for Meets, I use Chrome. In particular, Meets only works properly on Chrome. I do not use Gemini.
- I have recently been switching to Kagi as my default search engine; more about it in some later posts. Kagi also produces a browser, Orion, and I am testing it at the moment as a Brave alternative. Kagi/Orion does not push any AI solutions.
In the past, I also used Opera (a long time ago) and Vivaldi. I liked Vivaldi, but when I installed Brave, I happily changed to it.
This brings me to the main topic of this post, AI-empowered browsers. I have recently tried Comet, an AI-empowered Perplexity-linked browser, but have not really got a handle on it. I have been using the Perplexity app quite a lot, as I found it useful for “middle-sized” web search (i.e. to find how many and which plants a certain bacteria can infect, not to find the nearest restaurant). But somehow, I could not really see how much more the browser can do that the Perplexity app cannot.
Yet, this seems to be a tendency now, with all major players trying to push generative AI more and more. ChatGPT has just announced their browser, and I am sure others will follow. But, the push towards Agentic AI, that is AI systems that can do things for you, like order flight tickets, raises a lot of questions.
Now, I find the whole concept of an Agentic AI working in that way disturbing. You can read more about the push for AI in Brian Merchant’s excellent Substack posts, but for me, the two main issues are reliability and security. The UnHeard article focuses on security, and links to this study and this one, explaining how to convince the Perplexity Comet browser to give a hacker access to your e-mail or banking details. Mind you, they were written by Brave staff, whose company is also pushing their AI, Leo.
So, is AI safe? Firstly, the relentless push for incorporating AI into everything has already made us personally and as businesses, less secure and more stupid [paywall]. Secondly, some of the AI capabilities are definitely very useful, and I will be writing about them in my future posts. But, at least at the moment, I am very sceptical about the AI browsers and Agentic AI capabilities.
And, I promptly uninstalled Comet, although I will continue to use Perplexity (at least for now).
