Also: ChatGPT’s intelligence is zero, but it’s a revolution in usefulness, says AI expert In an interview with ZDNET, Rosedale remarked that the present capabilities of AI, such as large language models like GPT-4 and image generation technologies, resemble “alien intelligence” that can perform superhuman tasks but lacks human-like reasoning. Despite the awe-inspiring potential, Rosedale highlighted that current AI models are limited by their ability to read and write to memory at the same time – resulting in “zombie AIs” that cannot listen and reason like humans. Rosedale observed that AI’s ability to write software code is “shockingly good,” providing programmers with time-saving benefits. However, AI is also capable of creative tasks such as copywriting, generating images, and offering new insights. As AI advances, concerns about its impact on white-collar jobs and the potential for automation is growing. Also: AI could automate 25% of all jobs. Here’s which are most (and least) at risk “We finally have enough compute to do something that rivals thinking,” Rosedale said. “And that is turning over many rocks in terms of opportunities.” Among the opportunities is AI’s ability to perform tasks that rival human thinking, from text completion to image generation. According to Rosedale, AI can provide time compression and effort-saving benefits in various tasks, including programming. He shared his experience with GPT-4, which quickly wrote code to simulate Rosedale’s three-body problem, a task that would have taken him hours. “These machines are completing paragraphs for us,” said Rosedale. “… and they’re completing them in the ways we most likely would, [but] not in new ways.” Also: How to use ChatGPT to create an app And despite the many advantages, potential dangers exist. Generative AI could be used to disseminate disinformation, facilitate hacking, or write ransomware. Rosedale also voiced concern about AI’s potential to exacerbate wealth inequality and the need to make AI technologies freely available to all, similar to access to the early internet. He emphasized the importance of addressing inequality and the risks of increasing the divide between the haves and have-nots. He stressed that humanity is experiencing a paradoxical inflection point with AI: It’s powerful enough to be used as a weapon but not yet wise enough to prevent its weaponization. As for the future, Rosedale emphasized that AI developments are unpredictable, but the technology will likely continue to find new patterns and achieve superhuman performance. Nonetheless, he suggested that AI’s impact on everyday human lives may not be as significant as some expect. While AI-generated works of art may surpass human ability, they may not be as important to people as human-created art. Also: How ChatGPT works “It’s this kind of encyclopedia or super search engine that completes sentences. But it’s not thinking about how to overthrow us,” Rosedale said. “Not just yet.” You can watch our full interview with Rosedale on ZDNET or YouTube.
