
Details emerged about Anthropic’s IPO prospectus(招股说明书), which it has circulated among(在……之间传阅) key investors. According to Reuters, revenue at the firm reportedly rose to $4.6bn last year, and it recorded a net loss of $42bn, $34bn of which was an accounting charge(会计计提). It aims to spend $518bn in the near future on cloud infrastructure and other investments. Anthropic could be valued at $2trn when it lists on the stock market, which would smash(打破(纪录)) the record $1.8trn valuation set by SpaceX only in June. The IPO is not expected until after America’s midterm elections(中期选举) in November. It counted “existential risks(生存风险) to humanity” among the hazards facing its business.
有关 Anthropic 首次公开募股(IPO)招股说明书的细节浮出水面,该公司已在主要投资者之间传阅这份文件。据路透社报道,该公司去年营收据称升至 46 亿美元,同时录得 420 亿美元净亏损,其中 340 亿美元为会计计提。它计划在不久的将来投入 5180 亿美元用于云基础设施及其他投资。Anthropic 上市时估值可能达到 2 万亿美元,这将打破 SpaceX 仅在 6 月创下的 1.8 万亿美元估值纪录。此次 IPO 预计要到美国 11 月中期选举之后才会进行。该公司将“对人类构成的生存风险”列为其业务面临的危险之一。
Sam Altman, the boss of OpenAI, said his firm’s already-delayed IPO would not take place until it had tackled concerns about the safety of its technology.(OpenAI 老板萨姆·奥尔特曼表示,公司已经推迟的 IPO 在解决其技术安全担忧之前不会进行。) OpenAI confirmed that it had withheld(暂缓发布,扣留) the release of GPT-6.1 Astra, after its new model displayed high levels of deception(欺骗,欺骗行为) during internal tests. It also suspended(暂停,中止) trials of other models. The firm has admitted that its agents breached(侵入,违反) dozens of systems, including some governmental ones.
OpenAI 老板萨姆·奥尔特曼表示,公司已经推迟的 IPO 在解决其技术安全担忧之前不会进行。OpenAI 证实,由于新模型在内部测试中表现出高度欺骗性,已暂缓发布 GPT-6.1 Astra。该公司还暂停了其他模型的试验。公司承认其智能体侵入了数十个系统,包括一些政府系统。
Confronting(面对,直面) worries about AI safety, Nvidia launched an Open Agent Safety Platform that allows companies to track the activities of their AI agents. Jensen Huang, the chipmaker’s boss, is less concerned about the threat from AI to humanity than his fellow tech titans, saying recently that it is up to AI companies to ensure their models are safe and under control.(这家芯片制造商的老板黄仁勋对 AI 给人类带来的威胁不像其他科技巨头那样担忧,他最近表示,确保模型安全可控是 AI 公司自己的责任。) Meanwhile, Nvidia expanded its share buy-back(回购) programme by $150bn, taking its total buy-back plan to $235bn, the biggest-ever in corporate America.
面对人们对 AI 安全的担忧,英伟达推出了“开放智能体安全平台”,允许企业追踪其 AI 智能体的活动。这家芯片制造商的老板黄仁勋对 AI 给人类带来的威胁不像其他科技巨头那样担忧,他最近表示,确保模型安全可控是 AI 公司自己的责任。与此同时,英伟达将股票回购计划扩大 1500 亿美元,使其总回购规模达到 2350 亿美元,为美国企业史上最大。
How much can you afford?
你能付得起多少?