Vinay Prasad

Prasad 认为把 AI 临床研究称为“1 期试验”毫无意义，因为患者早已广泛使用 AI。

但如果用常识来看，我会说这不是一项 1 期研究。在这种情境下，1 期研究毫无意义。患者早已在用 AI 解答他们关于医学以及其他各种话题的疑问。

原文 (en)

But if I put on my common sense hat, I would argue that it's not a phase 1 study. Phase one studies are nonsensical in this context. Patients are already using AI to navigate their questions and understanding on medicine and every other topic under the sun.

Prasad 批评让医生给多个 AI 模型回答打分的研究，指出论文发表时模型已更新，结论缺乏价值。

为什么我们要向三个不同的 AI 提问，然后让随便几个医生给结果打分？等论文发表时，这些 AI 都已经变了。这到底回答了什么问题？

原文 (en)

Why are we asking questions to three different AI machines and then having random doctors grade the results? By the time that paper is published, the AIS are all different. What does that answer exactly?

Prasad 表示若声称 AI 能改善患者结局，就应通过与常规做法对比的随机对照试验来验证。

相反，很多人相信 AI 能改善患者结局。那样的话，你应该做的是……剧透一下……请击鼓……一项随机对照研究。其实很简单：把人随机分配到维持现有做法组或整合 AI 组。

原文 (en)

Instead, many believe that AI improves patient outcomes. In that case, you should do… spoilers to follow… drum roll please.. a randomized study. It turns out it's simple. You randomize people to doing things the way you're currently doing them, or to integrate the AI.

Prasad 主张 AI 试验应采用住院时长、30 天死亡率等对患者有意义的终点，而非问卷调查。

难点在于：你要衡量对患者真正重要的东西。你要衡量他们的住院时长，或 30 天死亡率。

原文 (en)

Here's the tricky part: you measure something that actually matters to patients. You measure their length of stay in the hospital, or their 30-day mortality.

Prasad 警告目前尚无关于医院 AI 的随机结局试验，而 AI 却已被推入临床工作流程。

但我们没有这样的研究。我一项都还没见过。与此同时，AI 就像那台“推车电脑”，有人推着它跟着我们到处走。

原文 (en)

But we don't have those studies. I’ve yet to see a single one. In the meantime, AI is the cow. Someone's pushing it wherever we walk.

Prasad 部分认同 Amodei，认为 FDA 很难监管变化飞快的医疗 AI，且不应监管所有数字健康和健康 AI 应用。

我认为他的话有几分道理，但不会像他那样走到底。核心的真相是 FDA 试图监管 AI 本身，也就是医疗 AI。比如你戴的手表会提供信息，FDA 想对此做很多监管。这很难监管，因为技术变化的速度远快于我们的评估能力。

原文 (en)

I see kernels of truth in what he's saying, but maybe I don't go all the way he goes. Where I see the big core truth is the FDA tries to regulate AI itself, medical AI. You wear a watch that's supposed to give you information and the FDA wants to do a lot of that regulation. That's a very tough thing to regulate, because this technology is changing so much faster than our ability to judge it.

Prasad 认为药企应在临床试验中投资 AI，但对 AI 在分子发现和基础科学中的作用不如 Amodei 乐观。

如果我是制药公司，我一定会投资 AI。在临床试验方面，AI 能让我们做得更便宜、更快、更高效。但在分子发现和基础科学方面，我不像 Dario 那么乐观，有几个原因。

原文 (en)

If I was a pharmaceutical firm, I would definitely invest in AI. There's so many things around running clinical trials that we can do better with AI, help us run them cheaper and faster and more efficiently. But when it comes to molecule discovery and basic science, I'm not as optimistic as Dario for a couple of reasons.

Prasad 认为 AI 药物发现受限于不可靠的生物学文献和人体验证瓶颈，而非缺乏想法。

第一，输入模型的许多癌症生物学信息几乎堪称造假研究。用垃圾训练 AI，它能学到什么？第二，AI 乐于提供假设，但仍需在人身上验证。我怀疑限制因素不是想法，我们的想法多到研究不过来。

原文 (en)

One, a lot of the information that is fed into models about cancer biology is basically as close to fraudulent research as you can imagine. If you train AI on garbage, what can it learn? The second challenge is AI will be happy to give you hypotheses to study, but you'll still need to run it with a few people to validate those hypotheses. I suspect the limiting reactant is not coming up with ideas; we have more ideas than we can study with people.

Prasad 称他任内 FDA 只发生了很小的改变，体系基本依旧。

也许我们只改变了 FDA 的 4% 或 2%。FDA 是个 18000 人的机构，仍有数万人在那里。但这个体系基本和以往一样。

原文 (en)

Maybe we changed 4 percent, 2 percent of the FDA. The FDA is an 18,000-person agency. There's tens of thousands of people still there. But the system is largely the same as it's always been.

Prasad 怀疑公共卫生能真正改革，预计其将分裂。

但我认为真正的改革不可能实现。我猜它会分裂成不同的公共卫生团体。

原文 (en)

But I don't think a real reform is possible. I suspect it'll splinter into different public health groups.

Prasad 称美国早期研究监管过于繁重，对小公司和罕见病往往不合适。

在 FDA 任职期间，我常常同情企业人员讲述的他们被迫跨越的障碍。美国早期研究阶段的监管非常繁重，很多出于善意，但并非都适用。

原文 (en)

During my time at the FDA, I was often sympathetic to the people who work at companies who told me about the hurdles they were forced to jump through. The regulation in this early side of research is tremendous in the United States. A lot of it is well-intentioned, but not all of it is fit for purpose.

Prasad 称 FDA 经历让他更倾向自由主义，质疑统一的安全有效性标准。

由华盛顿郊区的单一庞大政府机构为美国每个人设定这条标准，如今对我来说更难接受。

原文 (en)

The idea that a single monolithic government agency in the suburbs of D.C. can set that bar for every single person in America, that to me is a tougher pill to swallow now.