This preprint survey, conducted by AI Impacts and other researchers, surveyed 2,778 AI researchers who had published in top-tier AI venues. The survey aimed to gather predictions on the pace of AI progress and the nature and impacts of advanced AI systems. Key findings include: 1. **AI Progress and Milestones**: Most respondents predicted that several milestones, such as autonomously constructing a payment processing site, creating indistinguishable music, and fine-tuning large language models, would be achieved by 2028. The probability of unaided machines outperforming humans in all tasks was estimated at 10% by 2027 and 50% by 2047, earlier than in a similar survey conducted in 2022. 2. **Human-Level Performance**: Respondents expected high-level machine intelligence (HLMI) to achieve human-level performance in tasks or occupations by 2047, down from 2060 in 2022. Full automation of labor (FAOL) was expected to occur by 2116, down from 2164 in 2022. 3. **Framing Effects**: There was a significant framing effect, with respondents giving later predictions when asked about the probability of milestones being met by a given year compared to the year being met with a given probability. 4. **Social Impacts**: Over half of the eleven potentially concerning AI scenarios were deemed "substantially" or "extremely" concerning by more than 30% of respondents, including the spread of false information, manipulation of public opinion, and AI enabling dangerous groups to create powerful tools. 5. **Opinions on AI Safety**: While most respondents thought AI safety research should be prioritized more, there was no consensus on whether faster or slower AI progress would be better for humanity's future. 6. **Demographic Differences**: Geographical background was correlated with expectations about the timing of HLMI, with respondents from Asia anticipating an earlier arrival compared to those from Europe, North America, or other regions. The survey highlights the wide range of views among AI researchers and the need for broader evidence from various sources to inform policies and decisions regarding AI development and deployment.This preprint survey, conducted by AI Impacts and other researchers, surveyed 2,778 AI researchers who had published in top-tier AI venues. The survey aimed to gather predictions on the pace of AI progress and the nature and impacts of advanced AI systems. Key findings include: 1. **AI Progress and Milestones**: Most respondents predicted that several milestones, such as autonomously constructing a payment processing site, creating indistinguishable music, and fine-tuning large language models, would be achieved by 2028. The probability of unaided machines outperforming humans in all tasks was estimated at 10% by 2027 and 50% by 2047, earlier than in a similar survey conducted in 2022. 2. **Human-Level Performance**: Respondents expected high-level machine intelligence (HLMI) to achieve human-level performance in tasks or occupations by 2047, down from 2060 in 2022. Full automation of labor (FAOL) was expected to occur by 2116, down from 2164 in 2022. 3. **Framing Effects**: There was a significant framing effect, with respondents giving later predictions when asked about the probability of milestones being met by a given year compared to the year being met with a given probability. 4. **Social Impacts**: Over half of the eleven potentially concerning AI scenarios were deemed "substantially" or "extremely" concerning by more than 30% of respondents, including the spread of false information, manipulation of public opinion, and AI enabling dangerous groups to create powerful tools. 5. **Opinions on AI Safety**: While most respondents thought AI safety research should be prioritized more, there was no consensus on whether faster or slower AI progress would be better for humanity's future. 6. **Demographic Differences**: Geographical background was correlated with expectations about the timing of HLMI, with respondents from Asia anticipating an earlier arrival compared to those from Europe, North America, or other regions. The survey highlights the wide range of views among AI researchers and the need for broader evidence from various sources to inform policies and decisions regarding AI development and deployment.