AIs that are powered by Large Language Models (LLMs) are designed to understand and respond to human language with a degree of sophistication that allows them to "gather" user intentions effectively.
For example, if users frequently ask for technical support, an LLM might shift its tone to be more analytical, offering step-by-step solutions or troubleshooting advice. If users tend to showcase their hobbies, collections, passions and interest about a specific subject, they AI will adapt and provide future answers that are biased towards those.
As the model adapts to users' preferences, the AIs can offer more relevant, customized content.
In turn, this behavior should help provide a more tailored and intuitive user experience.
And because of this, AI companies behind them are also benefiting hugely. And probably, like they never did before.
According to Aravind Srinivas, CEO and co-founder of Perplexity.ai, in a podcast with TBPN:
" [...] So if any of the AI companies can do that, I think that could be like a thing where brands could pay a lot more money to advertise there.
But to make that happen, AI companies need to "crack memory properly."
The AI memory feature refers to a capability within LLM AI systems, where the model can "remember" past interactions with users across different chat sessions to provide a more personalized and context-aware experience.
This allows the AI to tailor its responses more effectively, offer relevant suggestions, and assist with long-term projects or recurring needs.
Srinivas added that Perplexity is even lured into gathering as much data from users as it can to fill up its AI's memory of its users, and this is "one of the other reasons" why it introduced the Comet browser, a "browser for agentic search" that should allow the company to "get data even outside the app to better understand you."
He explained that queries it gathers alone from its Perplexity main app aren’t sufficient for its AI to truly grasp its users' needs.
To fill in the gaps and offer a more complete understanding, a broader, more personal perspective is necessary, this is exactly where Perplexity’s browser can be useful.
"Because some of the prompts that people do in these AIs is purely work-related. It’s not like that’s personal," he said.
The thing is, privacy matters, and a lot of people aren't going to fall for such trap.
Since privacy-concerned users keep on increasing, especially since they start to fully grasp the extend tech companies have into their personal lives, chances are, extensively tracking users will create backlash.
This is then clarified Srinivas, who said that the the approach is "yet to be explored," and that it's "hypothetical."
Instead, to preserve privacy and retain users' trust, AI companies who wish to commercialize their AIs, need to utilize their AI's memory feature to focus on personalization—not for ads.
Srinivas views these as core capabilities essential to building a fully functional AI assistant.
Ads are optional, and so should memory, if that memory is utilized for sponsored content.
According to Srinivas, Perplexity aims to offer users three distinct solutions: First, no memory, no ads — functioning just like a traditional browser. Second, memory without ads — providing a smarter, more personalized experience. And third, memory with ads — where ads may appear only in certain sections of the feed, but not during core interactions.