内容来源：https://www.wired.com/story/anthropic-says-it-will-revolutionize-biology-some-day/

内容总结：

人工智能公司Anthropic及其首席执行官达里奥·阿莫代伊曾多次宣称，该公司的模型将彻底改变生物学和药物研发。但在周一的IPO文件中，Anthropic却降低了调门，承认自身缺乏物理实验经验，并指出药物开发存在的潜在风险。无论其目标何在，看来都需要相当长的时间才能实现。

两年前，阿莫代伊曾写道，他“愿意接受”人工智能系统能在短短五到十年内推动生物学取得千年进展这一想法。今年8月，他在X平台上表示“Anthropic正在生物学和医学领域迅速加大投入”。在同一篇帖文中，他预测人工智能将在未来五到十年内使治愈大多数人类疾病成为可能，并承诺将带来“大量AI加速药物”。该公司还在9月宣传了Claude发现的一种“类Crispr”酶。

尽管新的IPO文件仍然宣扬人工智能能够“深刻加速生物学和医学的进步速度”，并将“数十年的科学进展压缩到数年之内”，但对于如何将潜力转化为现实，文件措辞则较为保守。

该公司表示，其团队“需要从事科学研究和药物发现的第一手经验”，其生命科学业务将要求公司“开展超出我们作为软件和研究机构历史经验范围的活动”，包括处理潜在危险材料、操作实验室设备以及遵守大量法规。

Anthropic还淡化了关于其酶发现的措辞，称该酶“可能具有被开发为新型生物技术工具的潜力，或许可用于基因编辑”。此前有研究人员向《连线》杂志表示，他们对该酶发现的价值和实用性持怀疑态度，而一位一直在研究这些特定酶的科学家则怀疑Claude使用了他的研究成果。（该公司表示，其模型和分子生物学团队均无法获取该材料。）

IPO文件表明，Anthropic不再仅将自己视为生物技术行业的外部AI工具供应商，而是将生物技术发现视为其商业模式的核心组成部分。与所有生物技术公司一样，它面临着漫长的药物开发周期。任何AI辅助的发现仍须经过试管研究、动物实验，最终进行人体试验。即使人工智能能够帮助筛选出哪些新药或靶点对人体有效、哪些可能是无效的，仍需要数年测试来证明其安全性和有效性。

Anthropic在文件中承认了这些挑战。文件称：“由这些工作产生的任何候选药物或其他发现，都将面临生命科学行业特有的漫长、昂贵且充满不确定性的开发和监管路径，且可能永远无法产生收入。”（阿莫代伊本人则在8月的X帖文中主张，为即将到来的AI生成药物浪潮简化监管流程。）

鉴于向潜在投资者夸大陈述存在法律风险，美国证券交易委员会的备案文件通常比公司或其高管的公开声明更为谨慎。而此次IPO明确表明，该公司相信其技术将推动革命性进步。

但随着Anthropic进一步深入生物发现领域，它将面临与其他生物技术公司相同的难题：在无法仅靠增加计算机算力来加速的试验中证明自身价值。

中文翻译：

AI巨头Anthropic及其CEO达里奥·阿莫代伊曾多次宣称，该公司的模型将彻底改变生物学和药物研发。但在周一的IPO文件中，Anthropic的口气谦逊了不少，承认自身在物理实验方面缺乏经验，以及药物开发可能存在的陷阱。看来，无论它要走向何方，都需要一些时间才能抵达。

两年前，阿莫代伊曾写道，他“愿意接受”这样的想法：AI系统可以在短短五到十年内，让生物学取得一千年的进展。今年八月，他在X上表示，“Anthropic正在生物学和医学领域迅速加大投入。”在同一篇帖子中，他预测AI将在未来五到十年内使治愈大多数人类疾病成为可能，并承诺将带来“AI加速药物的洪流”。该公司还在九月宣传了Claude发现的一种“类Crispr”酶。

尽管新的IPO文件仍然鼓吹AI能够“深刻加速生物学和医学的进步速度”，将“数十年的科学进步压缩到数年之内”，但它在如何将潜力转化为现实方面措辞谨慎。

该公司表示，其团队“需要亲身从事科学研究和药物发现的经验”，其生命科学活动将要求它“参与超出我们作为软件和研究机构历史经验范围的活动”。这包括处理潜在危险材料、操作实验室设备，以及遵守大量法规。

Anthropic还淡化了关于其酶发现的措辞，称该酶“可能有潜力被开发为一种新型生物技术工具，或许可用于基因编辑”。（斜体为强调所加。）研究人员此前曾对《连线》表示，他们对这一酶发现的价值和实用性持怀疑态度，而一位一直在研究这些特定酶的科学家则怀疑Claude使用了他的研究成果。（该公司表示，其模型和分子生物学团队都不可能接触到这些材料。）

IPO文件表明，Anthropic不再仅仅将自己视为生物技术行业的AI工具外部供应商；它将生物技术发现视为其商业模式的核心部分。和所有生物技术公司一样，它面临着药物开发的漫长周期。任何AI辅助的发现仍然必须在试管中进行研究，然后在动物身上进行实验，最终在人体中进行试验。即使AI能够帮助筛选出哪些新药或靶点将在人体中起作用、哪些可能是无效的，仍然需要多年的测试来证明其安全性和有效性。

Anthropic在文件中承认了这些挑战。文件称：“由这些工作产生的任何候选药物或其他发现，都将面临生命科学行业特有的漫长、昂贵且不确定的开发和监管路径，且可能永远无法产生收入。”（就阿莫代伊而言，他在八月的X帖子中主张为即将到来的AI生成药物浪潮简化监管。）

美国证券交易委员会的备案文件通常比公司或其高管发表的声明更为谨慎，因为向潜在投资者夸大主张存在法律风险。而此次IPO明确表明，该公司相信其技术将推动革命性的进步。

但随着Anthropic进一步深入生物发现领域，它将面临与其他生物技术公司相同的难题：在无法仅靠增加计算机来加速的试验中证明自身的价值。

评论
返回顶部

英文来源：

AI giant Anthropic and CEO Dario Amodei have repeatedly claimed that the company’s models will revolutionize biology and drug discovery. But in its IPO filing Monday, Anthropic struck a humbler note, acknowledging its lack of experience in physical experimentation and the potential pitfalls of drug development. Wherever it’s going, it seems, it will take some time to get there.
Two years ago, Amodei wrote that he was “open” to the idea that AI systems could produce a thousand years of progress in biology in just five to 10 years. In August, he said on X that “Anthropic is ramping up its efforts very quickly in biology and medicine.” In the same post, he predicted that AI will make it possible to cure most human diseases in the next five to 10 years and promised a “deluge of AI-accelerated drugs.” The company also promoted the discovery of a “Crispr-like” enzyme by Claude in September.
While the new IPO filing still touts AI’s ability to “profoundly accelerate the rate of progress in biology and medicine” and compress “decades of scientific progress into years,” it’s reserved about how it will turn the potential into the actual.
The company says that its teams “need first-hand experience doing scientific research and drug discovery” and that its life sciences activities will require it “to engage in activities outside our historical experience as a software and research organization.” That includes handling potentially hazardous materials, operating laboratory equipment, and complying with a bevy of regulations.
Anthropic also toned down its language about its enzyme discovery, saying it “might have the potential to be developed into a novel biotechnology tool, perhaps for gene editing.” (Italics added for emphasis.) Researchers previously told WIRED they were skeptical of the enzyme discovery’s value and utility, and one scientist who had been studying these exact enzymes became suspicious that Claude used his research. (The company said neither its model nor its molecular biology team would have had access to this material.)
The IPO filing indicates that Anthropic no longer sees itself solely as an outside provider of AI tools to the biotech industry; it sees biotech discovery as a core part of its business model. And like all biotech companies, it faces the long timeline of drug development. Any AI-assisted discoveries still have to be researched in a test tube, then in animals, and eventually in people. Even if AI could help sort out which new drugs or targets will work in humans versus those likely to be duds, it will still take years of testing to prove their safety and efficacy.
Anthropic recognizes these challenges in its filing. “Any drug candidates or other discoveries resulting from these efforts would face the lengthy, costly, and uncertain development and regulatory pathways characteristic of the life sciences industry, and may never generate revenue,” the filing says. (For his part, Amodei made the case in his August X post for streamlining regulations for the coming wave of AI-generated drugs.)
US Securities and Exchange Commission filings are generally more cautious than statements made by companies or their executives, given the legal risks of overstating claims to would-be investors. And the IPO makes clear that the company believes its technology will drive revolutionary advancements.
But as Anthropic pushes further into biological discovery, it will face the same conundrum as other biotech companies: proving their worth in trials that can’t be hastened simply by adding computers.
Comments
Back to top