The introduction of ChatGPT Search was almost inevitable since ChatGPT emerged nearly two years ago. OpenAI’s chatbot, along with its competitors, is increasingly seen as an alternative to Google’s search engine. The recent launch of a dedicated search alternative by OpenAI signals a significant intent in this direction, featuring many advantages and few (yet important) drawbacks.
Perplexity showed us the way. Founded in August 2022, even before the launch of ChatGPT, AI startup Perplexity began to gain recognition in early 2023. Its conversational search engine demonstrated that a viable alternative to traditional search engines like those offered by Google and Microsoft could exist.
What’s more, it offered clear and coherent answers to user queries, including citations and links to the sources used. Perplexity’s answer engine can make mistakes. After all, the large language models it relies on aren’t perfect. However, it still poses a real threat to conventional search engines.
ChatGPT Search’s big limitation. After testing ChatGPT’s new features, TechCrunch reports that ChatGPT Search has a clear disadvantage. It struggles to provide answers to brief queries, which are the most common type. With chatbots like ChatGPT, it’s essential to articulate queries clearly. Otherwise, the responses may be inaccurate or misleading.
Image: TechCrunch
In fact, this is what happened to the author of TechCrunch’s report. He searched for “Nuggets score” in reference to an NBA game and received incorrect information. ChatGPT indicated that the Denver Nuggets were winning, even though they were losing to the Minnesota Timberwolves at that time. In contrast, Google consistently displayed the correct score in real time.
Criticism. Those who criticize the future of chatbots as search engines indicate that they may not have a place in this area. Computational linguist Emily Bender from the University of Washington recently said, “LLMs are nothing more than statistical models of the distribution of words in text.” She added, “If someone uses an LLM as a replacement for search, and the output they get is correct, this is just by chance.” While she’s essentially correct, it’s important to note that LLMs are continuously improving, leading to increased accuracy in their responses. As such, the answers seem less like a mere “chance.”
Not always a disadvantage. In reality, the significant disadvantage of ChatGPT when it comes to real-time searches diminishes for many other types of queries where that factor is less critical. For example, a search for “Nuggets vs. Timberwolves” on Perplexity and Google demonstrates how both platforms function.
Perplexity provided the correct score along with interesting facts about the game, while Google focused on displaying the score and breaking it down by quarter. If you only wanted to see the score, Google made it clear and accurate. However, if you sought more detailed information, Perplexity aggregated data from various sources (all of which were cited) and presented it in a richer context about the game.
AI searches can hardly make stuff up. The reason is current generative AI systems excel at summarizing text. That’s precisely what Perplexity and, presumably, ChatGPT Search do. They analyze content from multiple sources and summarize it to formulate their answers. They don’t fabricate information, or at least not that often. The real challenge for AI search engines is to trust the correct sources for extracting that information.
Free from ads and SEO (for now). AI search engines like Perplexity and ChatGPT Search currently offer a respite from the issues associated with traditional search engines like Google, notably advertising and results influenced by SEO strategies. One main criticism of Google’s search engine is that its results have declined in quality over time. In contrast, AI alternatives provide direct answers and include sources, allowing users to verify the information and trust what the chatbot presents.
Natural language. There’s room for improvement in the responses provided by AI search engines. In the above NBA example, including a link to a match highlight video would’ve been useful. However, the advantage lies in the straightforwardness of the answers. This approach aligns with Google’s ongoing efforts to present information more succinctly without requiring users to scroll through multiple search results to find what they need.
Additionally, AI chatbots deliver answers in clear and simple language, and they’re increasingly capable of generating human-like responses, even offering spoken answers that can be terrifyingly realistic.
The magic of conversational search. While it doesn’t seem that Perplexity and ChatGPT Search will replace Google in the short term, they’re gradually becoming more attractive alternatives for many scenarios.
For example, if you ask Google to write code to solve a function in programming, it’ll typically provide links to pages with that code. In contrast, ChatGPT can write the code for you. If the code doesn’t work, you can “chat” with it, explaining the error message so it can help refine the solution. Engaging in conversations and discussing the answers provided to improve them is one key advantage of these conversational search engines. Traditional search engines like Google aren’t designed for this purpose.
Rivals that will coexist. The rapid evolution of conversational search engines poses a real threat to Google, but the company is actively working to adapt to changing times. Its AI initiatives, like Project Astra, indicate that Google is determined to remain competitive, even if some of its current proposals haven’t succeeded.
It’s likely that, for the foreseeable future, traditional search engines like Google will coexist with AI search engines, allowing users to benefit from both options as needed. Whether AI search engines will eventually replace traditional ones remains to be seen. Complete reliance on them seems challenging at this point.
Image | ChatGPT | Google
Related | ChatGPT Spanish Teacher Mode: How to Use It to Learn Spanish From Scratch and for Free
View 0 comments