Why ChatGPT And Bard Can’t Replace Search Engines
In recent years, the rise of Generative AI tools like ChatGPT and Bard has changed the way we interact with technology. These intelligent chatbots are capable of answering any question you ask them (irrespective of whether they understand the topic referred to in the question), leading some to believe they could replace search engines altogether. However, is this really possible? This article will explore the limitations of AI chatbots and why they cannot replace search engines.
The Internet Run by Search Engines
The world wide web revolutionized the way we access and share information. And with the advent of search engines like Google, the process became even more convenient. Among the search engines, more than 90% of the share is captured by Google.
Google Search is still the most popular product of the company, and a major chunk of the revenue of the company comes through Google Ads, which is assisted by the search engine. Search Ads were Google’s main source of income in 2022. This includes ads on its search engine results page and on services like Gmail, Google Maps, & Google Play. Last year’s Google income breakdown shows that Search Ads accounted for 58.1% of the company’s overall global revenue.
Since AI tools like ChatGPT can answer literally any question you ask, some intellects suspect it to be the end of Search Engines. They also believe that OpenAI – Microsoft duo will replace Google. The internet was at a time flooded with articles predicting the end of the Google Era.
The Biggest Problem with Chatbots
One of the biggest problems with modern chatbots like ChatGPT and Bard is the issue of hallucination. These chatbots sometimes generate responses that are miles away from reality. Hence, the information generated by a chatbot always needs to be cross-checked and confirmed.
Why Do Chatbots Hallucinate?
A chatbot hallucinates due to the following two reasons:
- It doesn’t know what the reality is. The current versions of ChatGPT and Bard tend to believe every piece of information fed to them is true.
- At a more basic level, they don’t understand the difference between facts and non-factual sentences. As humans, we recognize that the same fact may be written in different ways, keeping the fact intact. A chatbot is unable to do that.
However, a chatbot can overcome this problem by verifying the results generated with authentic sources. But determining the authenticity of a source is a whole new problem for these AI tools and us!
The Need for Human Validation Favors Search Engines
Even when a chatbot becomes intelligent enough to double-check its response, an intelligent human mind will never trust it because the workings of modern AI chatbots are still a black box. Therefore, chatbots will have to provide references and a list of sources, which brings us back to the need for human-driven sources.
Only sources curated by human beings are reliable; however, even that is not a guarantee. This is why we need experts to provide us with reliable information. Meanwhile, search engines like Google provide links to the source, allowing us to check its validity.
The Internet Run by Chatbots
One may tend to wonder if search engines would remain unaffected by the rise of chatbots. The answer to this is – No. One of the best features of chatbots is that they can summarize and paraphrase content really well. The probable impact of it is that chatbots will become another component of search engines, where they will be used to summarize the content from top search results.
This is what Bing and Google have done. Google, being more cautious while making changes to its search engine, has introduced another specialized chatbot instead of integrating Bard into the system. Google AI Search will aim to provide authentic results, while Bard will be used for general interactions. This is a smart move to save the authenticity of Google Search.
Research in the Field
In a few recent research pieces, chatbots’ hallucination issue has been recognized very well. The researchers have suggested using the LLMs on top of Information retrieval. The whole process happens in the following three steps:
- At first, a complex query is broken down into simpler subqueries.
- For each subquery, an information retrieval model (search engine) finds out the top results.
- Once the top results are available, the LLM is used to restructure the output. Here the LLM would act as a summary.
But even in this case, the links from which the result has been compiled will have to be provided to validate the results.
The Recent Development in Search Engines
Bing has recently integrated ChatGPT with its search engine. It has resulted in a 15% increase in daily traffic. The Chatbot’s main function is to summarize the top search results and provide relevant links that it has used to construct its findings. This feature adds an interactive touch to web surfing. It also lets users get an overview of the discussed content in suggested links. Google has also announced a similar approach at the Google I/O 2023 event, indicating that they are following Bing’s footsteps by integrating a chatbot into their search engine.
The future of search engines appears to be heading towards a more interactive experience for users.
While AI tools like ChatGPT and Bard have revolutionized how we interact with technology, they cannot replace search engines. The problem of hallucination is one of the biggest limitations of modern chatbots. And even when they become intelligent enough to double-check their responses, human validation will always be required. As of today, search engines like Google remain the most reliable sources of information, allowing us to check the validity of sources. However, chatbots can still play a significant role in summarizing and paraphrasing results from top search results. Therefore, instead of replacing search engines, chatbots will become another component of them, used for general interactions, while search engines continue to provide authentic results.