英伟达首席执行官预测五年内实现人类水平的人工智能，并推出人工智能幻觉解决方案

英伟达首席执行官黄仁勋预测，人类水平的人工智能（AI）可能在未来五年内实现。他建议，通过特定的测试来定义 AGI，其中软件的表现明显优于人类，这可能会导致其实现。黄还解决了人工智能中的“幻觉”问题，提出了一种解决方案，要求人工智能在提供响应之前通过进行研究来验证答案。解决这个问题可能会对金融和加密货币等行业产生深远的影响，这些行业的准确性至关重要，而生成式人工智能系统目前还很有限。

Nvidia CEO Predicts Human-Level AI Within Five Years, Unveils Solution to AI Hallucinations

英伟达首席执行官预测五年内将实现人类水平的人工智能，并推出人工智能幻觉解决方案

San Jose, California - March 20, 2023 - In a groundbreaking speech at the Nvidia GTC developers conference, Nvidia CEO Jensen Huang has forecast that human-level artificial intelligence (AI) could be attainable within the next five years. Huang's announcement comes amidst growing optimism in the AI community, with significant advancements being made in the development of large language models (LLMs).

加利福尼亚州圣何塞 - 2023 年 3 月 20 日 - 在 Nvidia GTC 开发者大会上发表开创性演讲时，Nvidia 首席执行官黄仁勋 (Jensen Huang) 预测，人类水平的人工智能 (AI) 可能在未来五年内实现。黄仁勋宣布这一消息之际，人工智能界的乐观情绪日益高涨，大型语言模型 (LLM) 的开发取得了重大进展。

One of the key challenges facing the realization of human-level AI is the phenomenon of "hallucination," where AI systems generate inaccurate or fictitious information. This issue arises due to the limitations of current training techniques for LLMs, which struggle to distinguish between real-world facts and hallucinations.

实现人类水平的人工智能面临的关键挑战之一是“幻觉”现象，即人工智能系统生成不准确或虚构的信息。这个问题的出现是由于目前法学硕士培训技术的局限性，法学硕士很难区分现实世界的事实和幻觉。

However, Huang presented a straightforward solution to this problem during his keynote address: requiring AI systems to verify their answers by conducting research before providing responses. This approach aims to address the tendency of AI systems to generate responses based on limited or inaccurate data, leading to the production of false or misleading information.

然而，黄在他的主题演讲中提出了一个简单的解决方案：要求人工智能系统在提供响应之前通过研究来验证其答案。这种方法旨在解决人工智能系统根据有限或不准确的数据生成响应的趋势，从而导致产生虚假或误导性信息。

"We can solve the hallucination problem by requiring AI to do research," Huang stated. "By forcing AI to check its answers before providing them, we can significantly reduce the risk of hallucinations."

黄说：“我们可以通过要求人工智能进行研究来解决幻觉问题。” “通过迫使人工智能在提供答案之前检查其答案，我们可以显着降低产生幻觉的风险。”

While several AI models currently offer features to provide sources for their outputs, including Microsoft's CoPilot AI, Google's Gemini, OpenAI's ChatGPT, and Anthropic's Claude 3, the complete resolution of the hallucination problem could have profound implications for industries such as finance and cryptocurrency.

虽然目前有几种人工智能模型提供了为其输出提供来源的功能，包括微软的 CoPilot AI、谷歌的 Gemini、OpenAI 的 ChatGPT 和 Anthropic 的 Claude 3，但幻觉问题的彻底解决可能会对金融和加密货币等行业产生深远的影响。

Currently, the use of generative AI systems in contexts where accuracy is essential requires caution. For instance, ChatGPT's user interface warns users about potential errors and advises cross-checking crucial information.

目前，在准确性至关重要的环境中使用生成式人工智能系统需要谨慎。例如，ChatGPT 的用户界面会警告用户潜在的错误，并建议交叉检查关键信息。

In finance and cryptocurrency, accuracy is paramount, as it directly affects profits and losses. Consequently, the current reliance on generative AI systems is limited in these areas.

在金融和加密货币中，准确性至关重要，因为它直接影响利润和损失。因此，目前对生成式人工智能系统的依赖仅限于这些领域。

However, Huang's proposed solution could change this dynamic. If the issue of AI hallucinations were fully resolved, these AI models could potentially operate independently, executing trades and making financial decisions without human intervention.

然而，黄提出的解决方案可能会改变这种动态。如果人工智能幻觉问题得到彻底解决，这些人工智能模型就有可能独立运行，在没有人工干预的情况下执行交易并做出财务决策。

"Solving the problem of AI hallucinations would open the door to fully automated trading systems," Huang said. "The implications for the finance and cryptocurrency industries would be immense."

“解决人工智能幻觉问题将为全自动交易系统打开大门，”黄说。 “这对金融和加密货币行业的影响将是巨大的。”

Huang also emphasized the importance of benchmarking in the development of AI. He suggested that defining AGI through specific tests, where software outperforms humans by a substantial margin, could accelerate its realization.

黄还强调了对标在人工智能发展中的重要性。他建议，通过特定的测试来定义通用人工智能（AGI），其中软件的表现大大优于人类，可以加速其实现。

"We need to set clear benchmarks for AGI," Huang noted. "By establishing measurable goals, we can track our progress and ensure that we are on the right path."

“我们需要为 AGI 设定明确的基准，”黄指出。 “通过制定可衡量的目标，我们可以跟踪我们的进展并确保我们走在正确的道路上。”

The potential for human-level AI within the next five years represents a significant milestone in the evolution of technology. Huang's proposed solution to AI hallucinations provides a promising path forward, addressing a major obstacle that has hindered the development of truly intelligent systems.

未来五年内人类水平人工智能的潜力代表着技术发展的一个重要里程碑。黄提出的人工智能幻觉解决方案提供了一条充满希望的前进道路，解决了阻碍真正智能系统发展的主要障碍。