<\/p>\n
US software developers specialising in artificial intelligence (AI) and machine learning<\/a> (ML) engineering appear to be earning significantly more than their peers in other countries, the latest global survey from Stack Overflow has reported.<\/p>\n The survey of 49,000 responses from 177 countries found that in the US, the median salary for an AI and ML engineer is $189,500 per annum, over $50,000 more than the global median.<\/p>\n UK AI and ML developers who took part in the survey earn $149,756, almost $40,000 less than those who work in the US. In France, the salary gap is higher, with the median annual salary for French AI and ML engineers at $104,413. In India, AI and ML engineers can expect to earn just $17,434 a year.<\/p>\n While 47% of the developers who took part in the survey say they use AI tools daily, Stack Overflow reported that more developers actively distrust the accuracy of AI tools (46%) than trust them (33%).<\/p>\n The survey found that experienced developers are the most cautious, with the lowest \u201chighly trust\u201d rate (2.6%) and the highest \u201chighly distrust\u201d rate (20%), indicating a widespread need for human verification for those in roles with accountability.<\/p>\n The majority (54%) of the 11,202 developers who responded to questions about how they use AI in their workflow say the primary use is to search for answers. Just over a third (35%) use AI to generate content or synthetic data<\/a>, while 33% use it to help them learn new concepts.<\/p>\n The other main use is around documentation, with 31% saying they use AI to document their code and a quarter using AI to create or maintain documentation. Around a fifth of the developers polled use AI for debugging and to learn about the codebase.<\/p>\n However, based on the responses of 31,476 developers, the biggest single frustration among the people polled (66%) is dealing with \u201cAI solutions that are almost right, but not quite\u201d, which often leads to the second-biggest frustration: \u201cDebugging AI-generated code is more time-consuming,\u201d at 45%.<\/p>\n