- Alphabet chairman John Hennessy said an AI chatbot search costs 10 times more than a regular one, per Reuters.
- Analysts expect Google to incur billions in extra AI costs over the next several years.
- That's because a chatbot query requires more computing power and electricity than a typical search.
Talking to Google's chatbot Bard may cost the company more than just a faulty reveal.
John Hennessy, the chairman of Google's parent company Alphabet, said that a search on AI, like its chatbot Bard, can cost the company 10 times more than a normal keyword search, according to an interview with Reuters. That's because a search through an AI-language model requires more computing power — specifically chips — and electricity than a typical search.
Analysts say the extra costs can amount to billions of dollars over the span of years, Reuters reported.
The projected costs of AI search varies. Some analysts say Google would incur an extra $6 billion dollars by 2024 if it responded to half of its search queries with 50-word answers from an AI chatbot, per Reuters. Other analysts expect Google to rack up an additional $3 billion dollars by integrating the chatbot into its search engine, Reuters reported.
Google did not respond to Insider's immediate request for comment before publication.
These potential expenses would come head-to-head with Google's plans to reduce costs amid challenging economic conditions.
Just weeks after Google laid off 12,000 employees earlier this year, Alphabet's CFO Ruth Porat said in the firm's latest earnings call that the company will spend $500 million to reduce office space and up to $2.3 billion in severance packages. Lavish office perks will also be eliminated.
To reduce its AI-related costs, Google said it will run its chatbot with a "smaller version" of its LaMDA AI model, per Reuters. Inference costs — the expenses associated with how a chatbot generates responses — must also drop, Hennessy said, calling it "a couple year problem at worst"
Still, it's not clear how companies intend to lower AI's costs without losing accuracy, Naveen Rao, a former exec at Intel who oversaw its AI products division, told Reuters.
"How you cull (parameters away) most effectively, that's still an open question," Rao said.
ncG1vNJzZmivp6x7o8HSoqWeq6Oeu7S1w56pZ5ufonyitYyer6mdnqi2t7GMoKaon5yaeqS0wK2ZqKxdl66zsIymmLJlk6TAtXnCqKSpmZ6ueqO1y6WgqKajYrGwuMuaqaxlYmV%2FdHmR