usatoday24
Programmers love AI, but they don’t trust her too much. This is confirmed by a recent survey that Stack Overflow has done and in which 49,000 professional developers in their community have participated. That conclusion is as contradictory as logic, and points to a potential transformation of this sector.
Each time they use it more. According to him Complete study84% of developers already use AI as part of their workflow, when last year that proportion was 76%. The proportion is in the line of a survey conducted in 2023 in the github community, although in that case of the 500 programmers surveyed, 92% confessed Use AI tools to program.
But every time they trust less. The other prominent data of the survey is the one that indicates that programmers trust somewhat less in the Code generated by these AI tools. If last year the confidence in precise solutions was 40%, this year that confidence is only 29%.
I spend the day correcting mistakes. The most important frustration of developers is that they are working with AI solutions that gives a rather correct, not completely correct answer. That implies that in the end developers must devote much more time to detect and correct those errorsand in fact 66% of them confess to investing longer to fix that “almost correct” code of AI.
I trust more than human experts. When correcting errors there is another unique conclusion: in complicated code fragments, 75% of respondents claim that They would ask another human programmer (and not to another model of AI) when they do not trust the answer or the code generated by the machines.
AI agents do not set so much. Although tools “Vibe Coding“As a cursor or Windsurf, they have positioned themselves as a very interesting option even for new programmers, that theoretical revolution is far from being a reality. Of course: they gain productivity, and 69% say they have seen said metric increased thanks to agents.
Will I replace an AI? Programmers continue to see these tools more as a help and assistance than as a possible substitution. The majority (64%) do not see this technology as a threat to their work, but it is also true that this percentage was somewhat higher, of 68%: there is a small increase in that threat than For Jensen Huang, CEO of Nvidia, it is inevitable.
Learning to program. This community also made it clear that it does not stagnate: 69% of them have invested time in learning new programming techniques or a new language. Here is another relevant fact: 44% have learned new things thanks to AI tools, when 37% did it last year.
Work correcting the code of a machine. The survey seems to point to a future in which programmers end up programming less and less to become something like project chiefs or software engineers. His work will no longer be the one to chop code, but probably that of Check the code generated by these systems of artificial intelligence.
Image | Sigmund
GIPHY App Key not set. Please check settings