Today, we would like to introduce an article from the Uber Engineering blog.
Recently, after introducing AI, many people have said, "I'm not sure about productivity, but the bill is scary." This article is helpful for those people.
The title is "Running a Software Factory Efficiently at Uber Scale".
The numbers are provocative, aren't they? ^^
Uber was once introduced as a company that wasted money due to the indiscriminate introduction of AI, exhausting its AI budget early. However, this year, from February to August, the number of agent users increased 7 times, and the number of requests increased 9.4 times, but the total AI expenditure has remained almost the same since April. Moreover, when they fixed the model and re-examined it, the cost per 1,000 requests decreased by 34%, and the cost per session decreased by