The kind of factual error that ruined the launch of Google’s AI-powered chatbot will continue to worry companies using the technology, experts say, as the market value of its parent company continues to fall.
Investors in Alphabet slashed their shares another 4.4% to $95 on Thursday, marking a market value loss of about $163 billion since Wednesday, when shareholders removed about $106 billion of shares. Actions.
Shareholders grew uneasy after it emerged that a demo video from Google’s rival to the Microsoft-backed chatbot ChatGPT contained a faulty answer to a question about NASA’s James Webb Space Telescope. The animation showed a response from the program, named Bard, stating that JWST “took the first images of a planet outside of our own solar system”, leading astronomers to point out that this was not true.
Google said the bug underscored the need for Bard’s “rigorous testing” before a broader public release, which was scheduled to take place in the coming weeks. A presentation of Google’s AI-powered search plans on Wednesday also failed to reassure shareholders.
This week, Microsoft, a key backer of ChatGPT developer OpenAI, announced that it was integrating the chatbot’s technology into its Bing search engine. Google also plans to integrate the technology behind Bard into its search engine.
Dan Ives, an analyst at US financial services firm Wedbush Securities, described Wednesday’s mistake as “a dark day for Google that was exacerbated by Microsoft’s strong ChatGPT day.” He added: “We think it’s a black eye to rush a demo and have it show bugs in such a key AI event.”
Charalampos Pissouros, a senior investment analyst at brokerage XM, said Bard’s incorrect response during Google’s promotional video “adds to concerns that the company is losing ground to rival Microsoft.” Nonetheless, Alphabet remains a big business with a market capitalization of more than $1.2 trillion despite declines on Wednesday and Thursday.
Google is dominant in global search, with around 90% of the market compared to Bing’s 3%, according to data firm SimilarWeb, but Microsoft has told investors that every percentage point increase in market share equates to about $2 billion in additional ad revenue.
Bard and ChatGPT rely on large language models, a type of artificial neural network, that receive large amounts of text from the Internet in a process that teaches them how to generate responses to text-based prompts. ChatGPT became a sensation after its launch in November last year, composing recipes, poems, work presentations, and essays from simple prompts.
However, it also contained factual errors, which experts said reflected flaws in the vast data set, culled from the Internet, that ChatGPT had absorbed. Large language models feed on data sets made up of billions of words and build models that predict the words and sentences that would normally follow the previous piece of text. This can lead to answers that sound plausible but are incorrect.
Michael Wooldridge, professor of computer science at Oxford University, said he expected systems based on large language models to continue to make similar errors “for the foreseeable future.” “We must never accept without question what the great linguistic models tell us, however plausible it may be. The technology is powerful and very exciting, but it makes storytellers unreliable,” he said.
Dr. Thomas Lancaster, Senior Lecturer in Computer Science at Imperial College London, said he expected problems with Bard and ChatGPT responses to continue. “We are very, very far from getting perfect answers from these models,” he said.
Referring to his own experience with ChatGPT in recent weeks, Lancaster said he couldn’t handle math equations because he was trained on a text-based dataset and had cited bogus references in essays he had generated.
The FAQ page for the new version of Bing also acknowledges the potential pitfalls, stating, “Bing will sometimes misrepresent the information it finds, and you may see answers that sound convincing but are incomplete, inaccurate, or inappropriate.”
Microsoft and Google are moving forward with AI plans, which include making the technology behind Bard available to developers, creators and businesses to build apps based on it. Microsoft has released an AI-enhanced version of its Teams communications product, while OpenAI is also producing a subscription version of ChatGPT.
OpenAI has been contacted for comment.
