英偉達執行長預測五年內實現人類水準的人工智慧，並推出人工智慧幻覺解決方案

英偉達執行長黃仁勳預測，人類層級的人工智慧（AI）可能在未來五年內實現。他建議，透過特定的測試來定義 AGI，其中軟體的表現明顯優於人類，這可能會導致其實現。黃也解決了人工智慧中的「幻覺」問題，提出了一種解決方案，要求人工智慧在提供回應之前透過進行研究來驗證答案。解決這個問題可能會對金融和加密貨幣等產業產生深遠的影響，這些產業的準確性至關重要，而生成式人工智慧系統目前仍很有限。

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

英偉達執行長預測五年內將實現人類水準的人工智慧，並推出人工智慧幻覺解決方案

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).

加州聖荷西 - 2023 年 3 月 20 日 - 在 Nvidia GTC 開發者大會上發表開創性演講時，Nvidia 首席執行官黃仁勳 (Jensen Huang) 預測，人類水平的人工智慧 (AI) 可能在未來五年內實現。黃仁勳宣布這項消息之際，人工智慧界的樂觀情緒日益高漲，大型語言模型 (LLM) 的發展取得了重大進展。

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.

實現人類層級的人工智慧面臨的關鍵挑戰之一是「幻覺」現象，即人工智慧系統產生不準確或虛構的資訊。這個問題的出現​​是由於目前法學碩士培訓技術的局限性，法學碩士很難區分現實世界的事實和幻覺。

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，黃在他的主題演講中提出了一個簡單的解決方案：要求人工智慧系統在提供回應之前通過研究來驗證其答案。這種方法旨在解決人工智慧系統根據有限或不準確的數據產生響應的趨勢，從而導致產生虛假或誤導性資訊。

"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."

黃說：“我們可以透過要求人工智慧進行研究來解決幻覺問題。” “通過迫使人工智能在提供答案之前檢查其答案，我們可以顯著降低產生幻覺的風險。”

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.

雖然目前有幾種人工智慧模型提供了為其輸出提供來源的功能，包括微軟的CoPilot AI、Google的Gemini、OpenAI 的ChatGPT 和Anthropic 的Claude 3，但幻覺問題的徹底解決可能會對金融和加密貨幣等行業產生深遠的影響。

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，在準確性至關重要的環境中使用生成式人工智慧系統需要謹慎。例如，ChatGPT 的使用者介面會警告使用者潛在的錯誤，並建議交叉檢查關鍵資訊。

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.

在金融和加密貨幣中，準確性至關重要，因為它直接影響利潤和損失。因此，目前對生成式人工智慧系統的依賴僅限於這些領域。

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，黃提出的解決方案可能會改變這種動態。如果人工智慧幻覺問題得到徹底解決，這些人工智慧模型就有可能獨立運行，在沒有人工幹預的情況下執行交易並做出財務決策。

"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."

「解決人工智慧幻覺問題將為全自動交易系統打開大門，」黃說。 “這對金融和加密貨幣行業的影響將是巨大的。”

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.

黃也強調了對標在人工智慧發展中的重要性。他建議，透過特定的測試來定義通用人工智慧（AGI），其中軟體的表現大大優於人類，可以加速其實現。

"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."

「我們需要為 AGI 設定明確的基準，」黃指出。 “通過制定可衡量的目標，我們可以跟踪我們的進展並確保我們走在正確的道路上。”

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.

未來五年內人類水平人工智慧的潛力代表著科技發展的一個重要里程碑。黃提出的人工智慧幻覺解決方案提供了一條充滿希望​​的前進道路，解決了阻礙真正智慧系統發展的主要障礙。