A few months after chatbot ChatGPT wowed the world with its uncanny ability to write essays and answer questions like a human, artificial intelligence (AI) is coming to internet search.
Google, Bing and Baidu, three of the world’s largest search engines, last week integrated ChatGPT or similar technology into their search products, allowing people to get direct answers instead of just receiving a list of links. and participate in conversations. After entering a word or question. How will this change the way people interact with search engines? Are there risks in this form of human-machine interaction?
Microsoft’s Bing uses the same technology as ChatGPT, developed by OpenAI in San Francisco, California. But all three use Large Language Models (LLM). LLM creates persuasive writing by echoing the statistical patterns of text encountered in large databases. Google’s AI-powered search engine Bard, announced on February 6, is currently being used by a small group of testers. Microsoft’s version is now widely available, but there is a freely accessible waiting list. Baidu’s ERNIE Bot will be available in his March.
Prior to these announcements, several small businesses had already released AI-powered search engines. “Search engines are evolving into this new state where you can actually start talking to them, and you can talk to them like you would talk to a friend,” said the San Francisco-based company who co-founded Perplexity last August. says computer scientist Aravind Srinivas. His LLM-based search engine that provides answers in conversational English.
change trust
Compared to traditional Internet searches, the highly personal nature of conversations can help shake up perceptions of search results. Alexandra Ullmann, a computational social scientist at the University of Zurich in Switzerland, believes that people are inherently more likely to trust answers from chatbots that engage in conversations than answers from independent search engines. says.
2022 survey1 According to a team based at the University of Florida in Gainesville, participants who interact with chatbots used by companies like Amazon and Best Buy are more likely to trust the organization when the conversation feels more human. I understand.
This is beneficial and makes searching faster and smoother. However, increasing trust can be problematic given that AI chatbots make mistakes.google bard asked a question You confidently get it wrong about the James Webb Space Telescope in its own tech demo. ChatGPT also tends to create fictitious answers to questions it doesn’t know the answer to. known in this field as hallucinations.
A Google spokesperson said Bard’s error “underscores the importance of a rigorous testing process, which we’re launching this week on our trusted tester program.” However, some speculate that if such errors were discovered, they could erode users’ trust in chat-based searches, rather than increase their trust. Sridhar Ramaswamy, CEO of Neeva, an LLM-powered search engine, said: The mistake erased $100 billion from Google’s value as investors sold the stock out of fear of the future.
lack of transparency
Compounding the inaccuracy problem is the relative lack of transparency. Search engines typically present the user with a source (a list of links) and leave it up to the user to decide what to trust. In contrast, little is known on which data LLM trained. Encyclopedia Britannica or gossip blog?
“The method is completely opaque. [AI-powered search] This can have huge repercussions if the language model fails, hallucinates, or spreads misinformation,” says Urman.
If search bots get it wrong enough, Urman says, it can subvert users’ perceptions of search engines as unbiased arbiters of truth, rather than enhancing trust with their conversational abilities.
She conducted unpublished research that suggests current confidence is high. She looked at how people perceive existing features that Google uses to improve their search experience. This is called a “featured snippet” and an excerpt from the page deemed particularly relevant to the search appears above the link. “Knowledge Panels” – Summary Information automatically generated by Google in response to searches about individuals, organizations, etc. Nearly 80% of those Urman surveyed considered these features to be accurate, and about 70% considered them objective.
Chatbot-powered search blurs the line between machines and humans, says Giada Pistilli, chief ethicist at Hugging Face, a data science platform in Paris that promotes the responsible use of AI. She worries about how quickly companies are adopting her AI advances.
This article is reproduced with permission and was first published on February 13, 2023.
