Privacy and the future of data mining in the new AI-powered internet
The shift towards LLM search marks a significant change in the search industry. It presents a chance to address data privacy violations in a more effective manner. However, there are questions about whether users truly desire this level of privacy protection. The evolving internet landscape offers a platform for these discussions to take place. It will be interesting to see how this shift impacts the future of online search.
"Alexa, how many goals has Cristiano Ronaldo scored for Manchester United?" asked the user. The voice replied, "Cristiano Ronaldo played 346 games for Manchester United between 2003 and 2009, scoring a total of 145 goals." Compare this instant reply to a traditional search engine, and we know where the future of the search business is headed — a user types the keywords "number of goals Cristiano Ronaldo for Man U". Within microseconds, a list of web links appears, featuring a detailed career track record and a table of statistics.
The user clicks on one of the links and painstakingly scrolls through the page, looking for the exact number. In the case of simple queries like goals scored or runs made, the search is easy. Often, the search queries are more complex than obtaining statistics of star players. It's messier when numerous websites have conflicting accounts or nuanced opinions about historical facts or controversial news.
Search Engine Evolution
Google Search bypasses this problem by effectively filtering the most credible sources and placing those results at the top of the page. The search rankings created a pecking order in the internet world. Even then, it failed to reduce the manual work or reading time of the users. No wonder, then, that the search engine giant has shifted to an AI Overview that mimics the instant responses of Alexa, Siri, or large language models (LLMs) like ChatGPT and Grok.
Ask Llama whether it's true that regular fasting is good for cancer, and it immediately responds in the affirmative. Follow up with a question about credibility, and it provides three or four research paper citations and names the University of Southern California's study, which concludes that fasting can make cancer cells more sensitive to chemotherapy and protect normal cells.
AI vs. Traditional Search Engines
In contrast to the homework and extended reading that search engines provide, LLMs are short, crisp and pointed in their responses, even though they are as likely as search engines to show inaccurate responses. The convenience of LLMs over search results is undeniable because it saves a few clicks and reading effort, particularly for those looking for instant responses and not long-winded descriptions.
The internet is free, but at what cost? Search engines operate on a business model wherein their service is free for internet users. The primary source of revenue is advertisement — generating cash flow by selling ad space on search results.
Privacy Concerns and Revenue Generation
The "Sponsored Ads" atop the regular search results is prime real estate sold by businesses that want to attract clicks and eyeballs to their websites. The pay-per-click (PPC) model earns money only when the user clicks on the promoted content. Big Tech companies in the search business are, therefore, capable of increasing their profits by displaying ads that appeal to individual users, thus enhancing the probability of a click.
Collecting user data enables search engines to determine appealing content and provide personalised custom ads rather than generic content. Unsurprisingly, search engine giants have been accused of privacy violations over the past 15 years because the free model is, in reality, funded by the value of personal information of users that is used to run personalised marketing campaigns.
Rise of AI and Data Privacy
With the rise of artificial intelligence (AI) through LLMs, the business model of search engines is facing disruption. Targeted ads cannot be displayed if users shift from time-consuming internet searches to AI models like ChatGPT, Llama, or Grok. The digital advertising model cannot be sustained if the user data cannot be utilised for income generation. AI giants like OpenAI, Grok and Meta are fast chipping into the market share of the search engine industry.
Search engine users are slowly shifting to AI LLMs for their information needs. Running LLMs is an expensive affair, and the new internet search is more expensive than the traditional model, which means it needs better user targeting and monetisation capabilities.
Future Implications and Challenges
Companies have deployed new data mining techniques, best illustrated by this example: check your location history on Google Maps today by asking ChatGPT a simple question — "Tell me about myself." In most cases, it provides very precise information about the user based on their ChatGPT usage.
The world is set to witness a repeat of privacy breaches by AI companies earlier attributed to search engine giants. The search business is ending, and the era of LLM search is here. As the new internet evolves, we have an opportunity to fix the data privacy violation challenges, but do we want this? — Ankush Tiwari is the founder and CEO of pi-labs.ai, a cybersecurity and intelligence solutions company.