特朗普的硅谷顾问向人工智能审查制度发起战争
大卫·萨克斯（David Sacks）、埃隆·马斯克（Elon Musk）和马克·安德森（Marc Andreessen）等硅谷企业家对大型科技公司的人工智能审查威胁表示担忧，他们现在为特朗普提供技术问题建议。技术专家们对于在美国各地快速开发和采用“真实”人工智能的需求达成了相当一致的看法。
这三者可能使人工智能聊天机器人的反应成为保守派与科技公司进行持续“文化战争”的新战场。人工智能公司可以控制聊天机器人如何给出答案，以符合某些政治意识形态或推动其有偏见的内容审核。
获得正确的人工智能答案仍然是一个不断变化的挑战
有没有办法根据人工智能模型的真实程度对其进行评分？我们称之为伽利略索引。欢迎提出建议。 https://t.co/fJzwOH3JJa
—大卫·萨克斯 (@DavidSacks) 2024 年 12 月 12 日
According to Trump’s tech advisors, getting AI answers right for controversial subjects and live news events is a “constantly moving target.” Conservatives repeatedly criticized Big Tech for caving into government-sanctioned censorship, although some tech executives had begun to moderate their positions in public.
Mark Zuckerberg, Meta’s CEO, apologized to Congress ahead of the 2024 elections for bending to the Biden administration’s pressure to moderate content related to COVID-19. He said he had made a 20-year political mistake by taking too much responsibility for problems out of Meta’s control. Zuckerberg, however, promised never to repeat those mistakes.
Sacks also criticized OpenAI and Google for forcing AI chatbots to be politically correct, saying that one of ChatGPT’s earlier concerns was that it was programmed to be “woke.” He added that ChatGPT was not giving people truthful answers, pointing out that censorship was built into the answers.
“I think what’s important in training AI is to make sure that it is as truthful as possible and maximally curious because it is important that it trains to be honest even if that truth is unpopular.”
– Elon Musk, Tesla Motors CEO
Andreessen, the co-founder of a16z, believes that AI censorship has the potential to be a thousand times worse than social media censorship. He added that having everything controlled by an AI programmed to lie would be the perfect way to create the ultimate dystopian world.
Sacks made it clear that he was focused on “AI truthfulness.”
Trump’s advisor, Sacks suggests the ‘Galileo Index’ to score AI truthfulness
Trump’s tech advisors suggested that specialized agencies could undertake AI truthfulness investigations. It is, however, unclear whether the Trump administration has the investigative options and capacity. Sacks suggested the Galileo Index as a way to score AI models based on how truthful they are.
Musk said he would start something called “TruthGPT,” a truth-seeking AI that tries to understand the nature of the universe.
Sacks applauded Musk’s launch of Grok, saying that having it would, at minimum, keep both OpenAI and ChatGPT honest. Andreessen also added that with the Twitter files, Musk is doing what needed to happen broadly.
Andreessen emphasized that the American people need to find out how government pressure and censorship intertwine. He also noted that there needs to be consequences.
Users recently discovered that ChatGPT refused to answer questions about certain names, while OpenAI admitted that some names triggered internal privacy tools. Almost all AI tools, except for Grok and Perplexity, refused to answer questions about the 2024 election results.
From Zero to Web3 Pro: Your 90-Day Career Launch Plan
推荐文章
