Chart: Will AI Go Rogue?

Following this week's release of GPT-4, OpenAI's new multimodal model accepting image and text inputs rather than ChatGPT's text-only prompts, people on social media have been marveling about the new engine's results in performing a variety of tasks, such as creating a working website based on a simple sketch, outperforming humans in a variety of standardized tests or writing code. But as people are only beginning to understand the capabilities (and limitations) of artificial intelligence models such as ChatGPT and now GPT-4, there's also growing concern over what the rapid advancements in AI could ultimately lead to. "GPT-4 is exciting and scary," New York Times columnist Kevin Roose wrote, adding that there two kinds of risks involved in AI systems: the good ones, i.e. the ones we anticipate, plan for and try to prevent and the bad ones, i.e. the ones we cannot anticipate. "The more time I spend with AI systems like GPT-4," Roose writes, "the less I'm convinced that we know half of what's coming." According to Ipsos Global Advisor's 2023 Predictions, many people seem to share Roose's reservations with regard to artificial intelligence.