The recent unveiling of OpenAI’s ChatGPT, an Artificial Intelligence chatbot with 100 million monthly active users, has sparked a new race in the AI search engine industry. Microsoft quickly jumped on the bandwagon and announced that it would integrate ChatGPT-style search into its Bing search engine and Edge browser. Google, not to be outdone, announced its own AI search tool, Bard, the day before.
However, both companies have struggled with the reliability of their AI search tools. Google, being the world leader in search, has had years to incorporate AI into its search engine, yet Bard is still in its beta stage with a limited group of testers. The launch of Bard was plagued by errors, including an embarrassing mistake in a pre-made GIF that showed a user asking Bard for facts about the James Webb Space Telescope, one of which was untrue. This has led to the parent company Alphabet losing as much as 8 percent of its share price on the day of the launch.
Microsoft also faced its own issues with the ChatGPT-based search. Although no obvious errors were displayed at the launch event, the software is still in its beta stage and has a “hallucination problem” where there’s a 15 to 20 percent chance that it might make stuff up. Sarah Bird, Microsoft’s Head of Responsible AI, acknowledged this issue and added that previous versions of the software could even help users plan a school shooting, which has since been disabled.
This highlights the danger of launching AI search tools too early, as initial popular response can make or break a product’s reputation. Microsoft is no stranger to this, having experienced it with the infamous Clippy, a paperclip assistant that dispensed unwanted advice. ChatGPT, with its tendency to sometimes give garbled results or mundane variations of “I can’t answer that,” could become the Clippy of the AI search engine industry if enough casual users of Bing get unreliable results.
In conclusion, while the AI search engine industry is exciting and holds immense potential, companies must be mindful of the reliability of their AI tools before launching them to the public. The “hallucination problem” is a real issue that must be addressed, and companies must ensure that their AI tools are able to provide accurate and trustworthy information to their users.
