業界領袖探討邁向通用人工智慧（AGI）的軌跡，預測其將在幾年內到來。Elon Musk 特別預測，AI 可能在大約 5 年內超越人類智慧的總和。儘管發展迅速加速，Musk 堅持認為超級智慧構成人類生存風險的機率為 10% 到 20%。同時，像 Demis Hassabis 這樣的人物將此時刻視為關鍵的歷史門檻，並傾注整個職涯以達成 AGI。

與高階主管所擁抱的宿命般加速發展相反，AI Now Institute 發表了一篇評論，突顯了重大的懷疑態度。該機構指出，大多數 AI 研究人員懷疑擴展目前的大型語言模型將能實現 AGI。此外，他們主張 AGI 的論述主要作為一種行銷手法，使科技公司能夠籌集數十億美元的資金。這造成了脫節，少數領袖在缺乏更廣泛社會授權的情況下，基於推測性的敘事做出關鍵決策。

為反制這種技術解決主義，Mark Zuckerberg 建議將 AI 能力分配給個人，以確保遵循人類價值觀。然而，有效的監督需要的的不僅是工具的分配；它要求健全的治理。監管機構、公民社會組織與民選官員必須對 AI 部署行使合法權力，而非維持被動的乘客身分。企業必須與這些機構交涉以維持使用者信任，確保人類在抵達一個仍未被充分討論的目的地之前，能獲得掌控自身未來的能動性。

Industry leaders debate the trajectory toward artificial general intelligence (AGI), forecasting its arrival within a few years. Elon Musk specifically predicts that AI could exceed the sum of human intelligence in about 5 years. Despite this rapid acceleration, Musk maintains a 10% to 20% probability that a superintelligence could pose an existential risk to humanity. Meanwhile, figures like Demis Hassabis view this moment as a pivotal historical threshold, dedicating entire careers to reaching AGI.

In contrast to the fatalistic acceleration embraced by executives, the AI Now Institute published a critique highlighting significant skepticism. The institute notes that the majority of AI researchers doubt that scaling current large language models will achieve AGI. Furthermore, they argue the AGI narrative functions largely as a marketing gimmick enabling tech companies to raise billions of dollars in funding. This creates a disconnect where a small group of leaders makes critical decisions based on speculative narratives without a broader social mandate.

To counter this technosolutionism, Mark Zuckerberg suggests distributing AI capabilities to individuals to ensure adherence to human values. However, effective oversight requires more than tool distribution; it demands robust governance. Regulators, civil society organizations, and elected officials must exercise legitimate power over AI deployment, rather than remaining passive passengers. Companies must engage these institutions to maintain user trust, ensuring humanity acquires agency over its future before reaching a destination that remains largely undiscussed.