How ChatGPT and Bard AI cost
“Each AI query, like the Bard chatbot, can cost a company 10 times more than searching for information about a keyword in a Google search,” said John Hennessy, Alphabet chairman. Reuters. “The reason is that searching large language models requires a lot of computing power and power consumption.”
Hennessy also noted that AI costs could decrease over time as they are refined and optimized for performance. However, it took years to perfect this process, from chip performance, amount of data for AI training to usability and performance.
AI-enabled chatbots like OpenAI’s ChatGPT or Google’s Bard will be worth billions of dollars in a few years before companies can actually turn a profit, analysts say. Morgan Stanley estimates that Google Search processed 3.3 billion searches last year. Assuming Bard answers half of those questions, it will cost the company another $6 billion or more if the answers are longer than 50 words.
Meanwhile, according to SemiAnalysis, a chip technology research and consulting firm, to add ChatGPT-style AI to its search engine, Google will also need to spend $3 billion to invest in Tensor Processing Units (TPU) and associated infrastructure.
What actually makes the AI search more expensive than current models is its computing power. Conventional search engines don’t need a lot of computing power, they simply crawl the entire page and return relevant results in the form of links. However, AI query systems need chips with fast computing capabilities to process information and exchange data in complex text form. In addition, these vehicles consume more electricity than usual.
Earlier, Tom Goldstein, a professor of computer science at the University of Maryland, also suggested that OpenAI could spend at least $100,000 a day running ChatGPT. The estimate was made at a time when ChatGPT wasn’t that popular, so the actual figure could be many times higher in a context where this super AI already has over 100 million users.
According to SemiAnalysis, the company should rely on the engine running for a number of years rather than just racing for a short time. So the big boys aren’t tough enough to look for AI in the form of ChatGPT-style chatbots.
According to Paul Dougherty, CTO of technology company Accenture, cost is one of the top two reasons the search and social networking giant with billions of users didn’t adopt chatbots sooner. “One is accuracy, the other is precise scaling,” he said.
Alphabet is now being forced to take the plunge as its search AI model begins to improve and threatens the company’s market share. Among them, Microsoft’s competitor launched ChatGPT-based Bing AI and claims that “a new war has begun”.
Bao Lam (in accordance Reuters)