AI Search Chatbots May Transform The Internet As We Know It. Not Necessarily For Good.
Published in the Hindustan Times
It’s hard to think of almost any other tech advancement which has caught the global imagination as fast as ChatGPT. ChatGPT, OpenAI’s artificial intelligence chatbot is able to interact with users in a conversational manner, respond to user queries with succinct synthesized responses, keep track of what was said before and answer follow-up questions. The result is a human-like interaction which has shocked and enamored the world and led to ChatGPT’s massive popularity. However there’s one pivotal distinction: unlike human intelligence, which is built on understanding and creativity, ChatGPT is based on probability, large-language models which train on massive amounts of data to learn patterns and thus predict the most probable sequence of words in a given situation. Consequently ChatGPT will occasionally confidently provide responses which suffer from problems such as inaccuracy and bias. This fundamental difference has many implications - including moral and political - on the nature and use of output produced by such models.
Nevertheless the internet has exploded with examples of people using ChatGPT in ingenious ways from writing code to completing home assignments. While many of these will remain contingent on individual adoption, one immediate possibility of mass adoption is the replacement of traditional search engines with AI LLMs to provide succinct one-point answers. In fact this is so likely a possibility that Google which controls over 93% of the world search traffic is reported to be alarmed at the highest levels and pushed to rush the release of its own version of AI-enabled search chatbot, Bard. This is not good news. This transition while seemingly simplifying life for users can undo many of the original benefits of the internet.
A traditional search engine scours the web for information related to the search query and provides a list of links to the original sources. The user must then review the information in these individual links to build a fuller understanding of her search query. Based on the quality and relevance of returned links, the user may try wording her search differently to get a new set of links. This iterative process has at least four components: exposing the user to new and different sources of information; requiring the user to review individual sources, which may contain diverse and contradictory perspectives; discern between the relative relevance and credibility of individual links; and finally synthesizing the information contained therein to build an understanding for one self. On the other hand, an AI-driven chatbot will elide all of these steps in order to “simplify” and give a unified and succinct response. This has enormous - and mostly negative - implications for the internet as we know it.
First, this will lead to massive centralisation. According to some reports, more than half of web traffic comes from organic search; this entire traffic will get concentrated in the AI-chatbot since responses will “summarise” information for the user instead of directing traffic to individual sources of information. Even for AI-driven search engines like Bing which list sources, the very nature of “synthesis” implies that many original sources will fall by the wayside. Moreover, it is likely that a disproportionate amount of searches will terminate at the AI chatbot itself instead of continuing to individual sites. It is evident that if this were allowed to happen, this will deprive individual sites of visibility and revenue and render most sites unviable. This is likely to lead to wholesale consolidation and reconfigure the internet as we know it. This also raises serious ethical concerns because the companies hosting the AI chatbots will profit from the original work done by institutions and individuals without (appropriate) compensation.
This centralisation will amplify other dangers of AI language models manifold. A handful of organizations create and maintain AI language models giving them enormous power to control the future of the internet. Since AI language models rely on the data they are trained on to identify patterns, it is possible to control the nature of their outcome by limiting the scope of their training data. It is also possible to “limit” the nature of responses by non-transparently encoding rules in the model itself. This can lead to biased and incomplete responses made all the more egregious since the user is unlikely to know of either the omissions (willful or innocent) and biases therein. Moreover, since language models do not actually “understand” information but rely on probability, it is quite possible that they are unable to identify misinformation. For language models trained on open internet data, it may be possible to seed disinformation which would then be disseminated by the AI-chatbot to unsuspecting users. Compare this to a traditional search engine which allows the user to explore a multiplicity of sources and thus provides the user with valuable input into assessment of the antecedents and completeness of the sources of information.
Supplanting traditional search with AI-chatbots will not just lead to centralisation of information online. It is also likely to lead to uniformity of perspectives offline. Given the nature of AI language models, it is likely that it will converge on similar answers to related queries over time thus resulting in a narrow range of perspectives getting solidified and presented to the users. Further, it is likely that users will accept the presented information as opposed to sifting through diverse sources to make up their own mind. This will limit users' ability to make independent connections with sources of information and perspectives and will almost certainly constrain users’ ability to think critically and make up their own mind.
The great promise of the internet has been its promise of democratization through decentralization - of allowing users to bypass gatekeeping both in the production and consumption of information and creativity. This promise has appeared to flounder in the last decade for a variety of reasons. If AI chatbots replace traditional search, it may hasten the end of this promise.
Also Read: