立即打开
下次疫情爆发,这些技术或许能拯救人类

下次疫情爆发,这些技术或许能拯救人类

Jeremy Kahn 2021-01-16
这次可以在疫苗分配上帮上忙

为全球人口接种新冠病毒疫苗,这大概是人类在分配和物流上遇到的最大挑战之一。因此,有些人希望人工智能和区块链技术能够协助完成这项任务。

IBM公司的区块链业务主管杰森·凯利说:“这是在解决我们有史以来最大的数据难题。”

直到目前为止,这个难题的解决速度依然非常缓慢。

在美国,现在只有大约400万人已经接种了一剂以上的新冠疫苗,接种比例刚刚超过总人口的1%。而在全世界范围内,疫苗的接种速度就更慢了,有些国家还没有任何人接种疫苗。以色列算是目前接种率最高的国家,接种率也才刚刚达到12%。

新冠疫苗的分配,涉及到至少四个既独立又有所关联的问题:一是何时将多少疫苗运送到何地,这就涉及到需求预测的问题;二是要监测分配网络的瓶颈,这涉及供应链管理的问题;三是疫苗的生产商、分配方和接种者都需要确保这批疫苗是合法依规生产的,是满足医学标准的,剂量是没有问题的,这就涉及品控的问题;最后还需要对接种者的不良反应进行监测,这就涉及到后续跟踪问效的问题。

无论是各国政府还是各大制药公司,都希望可以在上述的每个步骤中使用到新技术。

低温与高价

目前,美国和欧洲已经批准了几支疫苗上市,对这些地区来说,精确的需求预测是尤为重要的。因为这些疫苗必须在超低温下保存,而且它们的价格也相对较高,对于政府来说,疫苗是浪费不起的。

有些公司正在帮助美国的医院系统和各州政府对有限的疫苗资源作合理分配,IBM就是其中之一。

IBM的政府事务全球总经理蒂姆·佩多斯表示,IBM正在使用该公司的沃森健康分析(Watson Health Analytics)软件,将各地的人口统计学数据、公共卫生状况以及人们对疫苗的态度相结合,以预测各地对疫苗的需求,同时确保疫苗得到公平分配。

在发展中国家,需求预测和供应链管理方面的挑战则更加严峻。

Macro-Eyes是位于西雅图的一家人工智能公司,该公司由本·菲尔斯创办,他以前曾经用机器学习技术搞过金融市场数据分析,以寻找有利可图的交易信号。

现在,他也在使用相同的技术来预测市场对药品和其他医疗服务的需求。在这方面,该公司已经与斯坦福大学的卫生系统在美国展开合作,同时它在非洲也完成了几个项目,包括帮助坦桑尼亚加强儿童的疫苗接种工作。

在非洲的几个项目中,这家公司采用了大量的数据,包括卫星地图、卫星成像图、某一地区的手机用户数量、社交媒体上的帖子以及政府官方数据等等,以预测某一地区会有多少人会去就医。每个数据集本身可能都是边际值,但是通过汇总分析大量的数据集,Macro-Eyes就能够做出准确的预测。

Macro-Eyes的系统可以将坦桑尼亚儿童疫苗接种需求的预测准确率提高96%,将疫苗浪费率下降至2.42%。

如今,Macro-Eyes也希望能够在新冠疫苗的分配上,为各国政府,包括美国各州,做出同样的贡献。

菲尔斯指出,在疫苗分配上,效率是至关重要的,因为现在疫苗的需求远远大于供应,每一剂疫苗都很珍贵。而且有些疫苗的售价也比较高,大家是浪费不起的。很多地方还要操心冷藏的问题,尽管有些疫苗——比如阿斯利康的产品,也可以在普通冰箱里保存。

他说:“我们不能因为把一批疫苗送错了地方,就把没用完的30剂扔掉,这样的错误是犯不起的。但如果我们严格按照人口分配疫苗,那么一些地区依然会面临严重分配不足的问题,而另一些地区则会存在严重分配过剩的问题。”

区块链和机器学习

一旦分配网络建立并运行起来,就要密切关注它的运行情况,跟踪疫苗在供应链中的流动。而这也是人工智能和区块链等技术能够发挥作用的另一个领域。

IBM公司推出了一款“基于对象”的供应链管理软件,它可以近乎实时地追踪每一针疫苗的位置,并将疫苗与接种者进行匹配。佩多斯表示,在新冠疫情爆发的早期阶段,IBM就已经使用该软件追踪个人防护器材的供应了。

Celonis是一家帮助企业追踪实时业务流程的软件公司,该公司也有一款追踪个人防护器材供应链的软件。它也希望这款软件能够用在疫苗分配中。

该软件的机器学习机制可以用来预测分配中潜在的瓶颈问题,并且就如何绕过这些瓶颈给出建议。

美国的技术和业务流程外包公司Genpact也为制药公司开发了一款软件,用来帮助他们监测各个批次的药品在供应链中的流转。

该公司的药物安全负责人埃里克·桑德尔表示,新冠疫苗的监测是很有挑战的,因为不同批次的疫苗很可能是由不同的授权生产商在不同的药厂里生产的,彼此之间可能存在差异。每批疫苗的存储也会带来额外的问题。而持续追踪每个人接种的每一剂疫苗,对追踪疫苗的安全性也是至关重要的。

此外还有另一个挑战。这些供应链软件大都是针对单个企业用户设计的。但对于新冠疫苗来说,它的供应链有必要由多方共同监测,包括制药企业、物流企业、医院、药店以及政府的各个职能部门。

但这些企业和部门并非都在使用相同的软件,其中有些企业彼此甚至是竞争关系,所以它们基本上不会愿意共享数据,或者出于监管、安全和合规方面的原因而不能分享数据。

对于这个问题,IBM公司的区块链业务主管杰森·凯利认为,区块链技术(也就是比特币等数字货币的底层技术)恰恰在这方面大有可为。区块链技术能够为每一剂疫苗在供应链中的流转留下安全可信的记录,而且参与这个过程的每个部门和企业都可以使用。

凯利介绍道,IBM目前正在与几家制药公司讨论建立基于区块链技术的解决方案,以创建“一个最低限度可行的生态系统”来支持疫苗分配。

不良反应

一旦人们接种了疫苗,疫苗制造商和政府卫生部门还需要监测接种者是否出现不良反应或者是罕见并发症。虽然在临床试验阶段,疫苗已经在几万人身上进行了测试,但有些异常反应很可能只有在接种了几百万人之后才会暴露出来。

很多地方的政府都要求,医生和制药公司需要对患者用药后的所有异常症状进行报告。即便有些药物只给很少的人使用,这些规定也会为各地带来许多有关不良反应的报告。事实证明,这些异常症状绝大多数时候都是“假警报”,要么与药物本身无关,要么实际上并无任何危险。但有的时候,它们也确实暴露了某些之前没有被注意到的重大安全隐患。

现在需要接种新冠疫苗的人如此之多,这些报告的数量可能会非常庞大,大到我们完全审查不过来,无法及时从中发现隐患迹象。

因此,有些国家已经在寻求人工智能技术的帮助。

英国卫生监管机构已经与Genpact公司签署合同,用机器学习软件对那些“黄牌”报告进行筛查——也就是医生和患者报告了异常不应反应,且有必要引起重视的报告。

Genpact研发的这个系统已经于2020年12月上线,它能够自动接收文本信息并进行编码,然后搜索有可能涉及重大安全隐患的症状模式,并将其提交给监管机构作进一步调查。

Genpact公司的首席执行官泰格·塔加利安介绍道,这款软件已经接受了多种写作训练,既可以理解医生在报告症状时所写的专业术语,也能够理解老百姓的通俗化表达。

Genpact之所以可以迅速部署这个系统(合同签署后仅三个月,该系统便投入了使用),是因为Genpact公司以前就为制药行业的客户开发过更加复杂的类似系统。特别是在美国,因为美国食品与药品管理局对药品上市后的监测报告制度有严格的要求,药品的任何安全隐患都必须上报。

因此,Genpact的人工智能软件不能像在英国那样,仅仅筛查政府提供的表格,它还要筛查医学期刊文章,甚至是社交媒体上的帖子,以寻找任何应该引起重视的不良反应线索。

有些技术专家也在感叹,人工智能技术为何没有在疫情期间帮上多大的忙。实际上,在疫情刚刚爆发时,一些人工智能软件就已经发出预警,指出一种令人担忧的新型呼吸道病毒似乎正在流行,但人工智能技术显然无力阻挡病毒的大流行。

人工智能技术对疫情的流行病学建模和政策制定也只起了微不足道的影响。另外,它对寻找治疗方法和开发疫苗的帮助也较为有限。

因此,有人打趣说,等到下次爆发疫情的时候,人工智能技术或许就准备好了,不过不是这一次。不过在确保疫苗快速安全分发上,这项技术还是能够证明其价值的。(财富中文网)

译者:朴成奎

为全球人口接种新冠病毒疫苗,这大概是人类在分配和物流上遇到的最大挑战之一。因此,有些人希望人工智能和区块链技术能够协助完成这项任务。

IBM公司的区块链业务主管杰森·凯利说:“这是在解决我们有史以来最大的数据难题。”

直到目前为止,这个难题的解决速度依然非常缓慢。

在美国,现在只有大约400万人已经接种了一剂以上的新冠疫苗,接种比例刚刚超过总人口的1%。而在全世界范围内,疫苗的接种速度就更慢了,有些国家还没有任何人接种疫苗。以色列算是目前接种率最高的国家,接种率也才刚刚达到12%。

新冠疫苗的分配,涉及到至少四个既独立又有所关联的问题:一是何时将多少疫苗运送到何地,这就涉及到需求预测的问题;二是要监测分配网络的瓶颈,这涉及供应链管理的问题;三是疫苗的生产商、分配方和接种者都需要确保这批疫苗是合法依规生产的,是满足医学标准的,剂量是没有问题的,这就涉及品控的问题;最后还需要对接种者的不良反应进行监测,这就涉及到后续跟踪问效的问题。

无论是各国政府还是各大制药公司,都希望可以在上述的每个步骤中使用到新技术。

低温与高价

目前,美国和欧洲已经批准了几支疫苗上市,对这些地区来说,精确的需求预测是尤为重要的。因为这些疫苗必须在超低温下保存,而且它们的价格也相对较高,对于政府来说,疫苗是浪费不起的。

有些公司正在帮助美国的医院系统和各州政府对有限的疫苗资源作合理分配,IBM就是其中之一。

IBM的政府事务全球总经理蒂姆·佩多斯表示,IBM正在使用该公司的沃森健康分析(Watson Health Analytics)软件,将各地的人口统计学数据、公共卫生状况以及人们对疫苗的态度相结合,以预测各地对疫苗的需求,同时确保疫苗得到公平分配。

在发展中国家,需求预测和供应链管理方面的挑战则更加严峻。

Macro-Eyes是位于西雅图的一家人工智能公司,该公司由本·菲尔斯创办,他以前曾经用机器学习技术搞过金融市场数据分析,以寻找有利可图的交易信号。

现在,他也在使用相同的技术来预测市场对药品和其他医疗服务的需求。在这方面,该公司已经与斯坦福大学的卫生系统在美国展开合作,同时它在非洲也完成了几个项目,包括帮助坦桑尼亚加强儿童的疫苗接种工作。

在非洲的几个项目中,这家公司采用了大量的数据,包括卫星地图、卫星成像图、某一地区的手机用户数量、社交媒体上的帖子以及政府官方数据等等,以预测某一地区会有多少人会去就医。每个数据集本身可能都是边际值,但是通过汇总分析大量的数据集,Macro-Eyes就能够做出准确的预测。

Macro-Eyes的系统可以将坦桑尼亚儿童疫苗接种需求的预测准确率提高96%,将疫苗浪费率下降至2.42%。

如今,Macro-Eyes也希望能够在新冠疫苗的分配上,为各国政府,包括美国各州,做出同样的贡献。

菲尔斯指出,在疫苗分配上,效率是至关重要的,因为现在疫苗的需求远远大于供应,每一剂疫苗都很珍贵。而且有些疫苗的售价也比较高,大家是浪费不起的。很多地方还要操心冷藏的问题,尽管有些疫苗——比如阿斯利康的产品,也可以在普通冰箱里保存。

他说:“我们不能因为把一批疫苗送错了地方,就把没用完的30剂扔掉,这样的错误是犯不起的。但如果我们严格按照人口分配疫苗,那么一些地区依然会面临严重分配不足的问题,而另一些地区则会存在严重分配过剩的问题。”

区块链和机器学习

一旦分配网络建立并运行起来,就要密切关注它的运行情况,跟踪疫苗在供应链中的流动。而这也是人工智能和区块链等技术能够发挥作用的另一个领域。

IBM公司推出了一款“基于对象”的供应链管理软件,它可以近乎实时地追踪每一针疫苗的位置,并将疫苗与接种者进行匹配。佩多斯表示,在新冠疫情爆发的早期阶段,IBM就已经使用该软件追踪个人防护器材的供应了。

Celonis是一家帮助企业追踪实时业务流程的软件公司,该公司也有一款追踪个人防护器材供应链的软件。它也希望这款软件能够用在疫苗分配中。

该软件的机器学习机制可以用来预测分配中潜在的瓶颈问题,并且就如何绕过这些瓶颈给出建议。

美国的技术和业务流程外包公司Genpact也为制药公司开发了一款软件,用来帮助他们监测各个批次的药品在供应链中的流转。

该公司的药物安全负责人埃里克·桑德尔表示,新冠疫苗的监测是很有挑战的,因为不同批次的疫苗很可能是由不同的授权生产商在不同的药厂里生产的,彼此之间可能存在差异。每批疫苗的存储也会带来额外的问题。而持续追踪每个人接种的每一剂疫苗,对追踪疫苗的安全性也是至关重要的。

此外还有另一个挑战。这些供应链软件大都是针对单个企业用户设计的。但对于新冠疫苗来说,它的供应链有必要由多方共同监测,包括制药企业、物流企业、医院、药店以及政府的各个职能部门。

但这些企业和部门并非都在使用相同的软件,其中有些企业彼此甚至是竞争关系,所以它们基本上不会愿意共享数据,或者出于监管、安全和合规方面的原因而不能分享数据。

对于这个问题,IBM公司的区块链业务主管杰森·凯利认为,区块链技术(也就是比特币等数字货币的底层技术)恰恰在这方面大有可为。区块链技术能够为每一剂疫苗在供应链中的流转留下安全可信的记录,而且参与这个过程的每个部门和企业都可以使用。

凯利介绍道,IBM目前正在与几家制药公司讨论建立基于区块链技术的解决方案,以创建“一个最低限度可行的生态系统”来支持疫苗分配。

不良反应

一旦人们接种了疫苗,疫苗制造商和政府卫生部门还需要监测接种者是否出现不良反应或者是罕见并发症。虽然在临床试验阶段,疫苗已经在几万人身上进行了测试,但有些异常反应很可能只有在接种了几百万人之后才会暴露出来。

很多地方的政府都要求,医生和制药公司需要对患者用药后的所有异常症状进行报告。即便有些药物只给很少的人使用,这些规定也会为各地带来许多有关不良反应的报告。事实证明,这些异常症状绝大多数时候都是“假警报”,要么与药物本身无关,要么实际上并无任何危险。但有的时候,它们也确实暴露了某些之前没有被注意到的重大安全隐患。

现在需要接种新冠疫苗的人如此之多,这些报告的数量可能会非常庞大,大到我们完全审查不过来,无法及时从中发现隐患迹象。

因此,有些国家已经在寻求人工智能技术的帮助。

英国卫生监管机构已经与Genpact公司签署合同,用机器学习软件对那些“黄牌”报告进行筛查——也就是医生和患者报告了异常不应反应,且有必要引起重视的报告。

Genpact研发的这个系统已经于2020年12月上线,它能够自动接收文本信息并进行编码,然后搜索有可能涉及重大安全隐患的症状模式,并将其提交给监管机构作进一步调查。

Genpact公司的首席执行官泰格·塔加利安介绍道,这款软件已经接受了多种写作训练,既可以理解医生在报告症状时所写的专业术语,也能够理解老百姓的通俗化表达。

Genpact之所以可以迅速部署这个系统(合同签署后仅三个月,该系统便投入了使用),是因为Genpact公司以前就为制药行业的客户开发过更加复杂的类似系统。特别是在美国,因为美国食品与药品管理局对药品上市后的监测报告制度有严格的要求,药品的任何安全隐患都必须上报。

因此,Genpact的人工智能软件不能像在英国那样,仅仅筛查政府提供的表格,它还要筛查医学期刊文章,甚至是社交媒体上的帖子,以寻找任何应该引起重视的不良反应线索。

有些技术专家也在感叹,人工智能技术为何没有在疫情期间帮上多大的忙。实际上,在疫情刚刚爆发时,一些人工智能软件就已经发出预警,指出一种令人担忧的新型呼吸道病毒似乎正在流行,但人工智能技术显然无力阻挡病毒的大流行。

人工智能技术对疫情的流行病学建模和政策制定也只起了微不足道的影响。另外,它对寻找治疗方法和开发疫苗的帮助也较为有限。

因此,有人打趣说,等到下次爆发疫情的时候,人工智能技术或许就准备好了,不过不是这一次。不过在确保疫苗快速安全分发上,这项技术还是能够证明其价值的。(财富中文网)

译者:朴成奎

Vaccinating the global population against COVID-19 is one of the most immense distribution and logistical challenges humanity has ever faced. Some are hoping that artificial intelligence and blockchain technology can help with the task.

“This is about trying to solve the biggest data puzzle of our lifetime,” says Jason Kelley, who heads blockchain services for IBM.

So far, solving that puzzle has proved painstakingly slow. Only about 4 million people in the U.S.—just over 1% of the population—have received at least one dose of a COVID-19 vaccine. Worldwide, the progress is even more sluggish, with some countries yet to vaccinate any of their citizens. Even Israel, which has vaccinated the largest portion of its population so far, has given first jabs to just 12%.

The distribution of the COVID vaccine involves at least four separate but related problems: how much vaccine to ship where and when. That’s demand forecasting. Then that distribution network needs to be monitored for bottlenecks. That’s supply chain management. Furthermore, the pharmaceutical companies making the vaccine, those administering it, and people receiving it all need assurance that the batch of vaccine is legitimate and made to the correct standard, and that the right dose is administered. That’s quality assurance. Finally, those receiving the vaccine need to be monitored for any unusual side effects. That’s adverse event surveillance.

Governments and companies are hoping to use new technologies in each of these steps.

Low temps and high prices

Accurately forecasting demand for the vaccine is particularly important for some of the first vaccines that have been approved for use in the U.S. and Europe because of the ultralow temperatures at which they must be kept and their relatively high prices. Governments can’t afford to let doses go to waste.

IBM is among the companies trying to help U.S. hospitals and state governments manage the limited supplies of vaccines available so far, according to Tim Paydos, the company’s global general manager for government industry. This involves using IBM’s Watson Health Analytics software to marry zip-code–level data on demographics and health status with information on people’s attitudes toward vaccinations to try to forecast demand and also ensure vaccines are distributed equitably, he says.

In the developing world, the challenge of demand forecasting and supply chain management is even more acute. Macro-Eyes is an A.I. company based in Seattle. It was founded by Ben Fels, who had once used machine learning to scour financial market data for minute trading signals. Today, he uses similar technology to look for indicators that will enable Macro-Eyes to forecast demand for medicines and other health care offerings. On this front, the company has worked with Stanford University’s health system in the U.S., but it has completed several projects in Africa, including one to bolster childhood immunizations in Tanzania.

In its African projects, the company uses a wide range of data—including satellite imagery and maps, the number of mobile phone users in a certain area, social media posts, and official government data—to try to predict how many people will show up for health care at any one place. Each data set on its own may be of marginal value. But by combining lots of data sets, Macro-Eyes is able to make accurate predictions.

Macro-Eyes’ system was able to improve forecasts for childhood vaccination demand in Tanzania by 96% and reduce wasted dosages to just 2.42 vials per 100 shipped. Now Macro-Eyes is hoping to help governments around the world—including possibly some U.S. states—do something similar with COVID-19 vaccines.

Ensuring efficiency is even more important with these vaccines, Fels says, since demand far exceeds supply, making each dose precious. Some of the vaccines are also relatively expensive per dose, making waste costly. Concerns about cold storage will be an issue in many places, even with vaccines such as AstraZeneca’s that can be kept at normal refrigerator temperatures. “We can’t afford to throw away 30 doses because we sent them to the wrong place,” Fels says. “But if we allocate the vaccine strictly according to population we are going to severely under-allocate to some sites and severely over-allocate to others.”

Blockchain and machine learning

Once a distribution network is up and running, keeping tabs on how it is functioning and tracking doses as they move through the supply chain is another area where A.I. and technologies such as blockchain may play a role.

IBM markets “object-based” supply chain management software that can track the location of every vaccine vial in as near real time as possible and match the vial to the people vaccinated with the doses contained in that vial. It had already used this software earlier in the pandemic to help track supplies of personal protective equipment, Paydos says.

Celonis, another software company that helps businesses build dashboards to track business processes in real time, has also seen its software used to track PPE for health system customers and is now hoping that it can be adopted to handle vaccines too.

Building on top of this software, machine learning can be used to predict potential distribution bottlenecks and to potentially suggest ways to work around them. Genpact, the U.S.-based technology and business process outsourcing firm, has developed software for its pharmaceutical industry customers that helps them track batches of drugs as they move through a supply chain. The COVID-19 vaccines will be particularly challenging, says Eric Sandor, the head of Genpact’s pharmacovigilance A.I. business, because different batches may be produced by different contract manufacturers at different facilities, resulting in variations among them, and there may be further issues around the storage of individual lots of vials from within each batch. Keeping track of exactly which lot and batch was used to vaccinate each individual may be critical to tracking any safety issues with the vaccine, Sandor says.

There’s another challenge too: Most of these supply chain software packages were designed to be used within a single organization. But with the COVID-19 vaccines, there is a need to track supplies through a chain that is controlled by many different parties—drug manufacturers, courier companies, hospitals and pharmacies, and even various branches of government—which don’t all use the same software. What’s more, some of these companies may be competitors that are generally reluctant to share data, or they may be unable to share data easily owing to regulatory, security, and compliance concerns.

That’s where IBM’s Kelley thinks blockchain technology, the digital ledger system that underpins cryptocurrencies such as Bitcoin, can play a vital role. This kind of digital ledger could provide a trusted, secure, and verifiable record of the chain of custody for every vial of vaccine that could be used by every organization involved in the process, he says. IBM is currently in discussions with several drug manufacturers about signing up for this blockchain-based solution in order to create “a minimally viable ecosystem” to launch it, he says.

Side effects

Once people have received inoculations, the vaccine makers and government health agencies will need to monitor these people for signs of unusual side effects or rare complications. While the vaccines have been tested on tens of thousands of people during clinical trials, there may be side effects or safety issues that only become apparent when millions receive injections. Many governments require doctors and pharmaceutical companies to file reports for any unusual symptoms patients experience after being given a drug. Even for medicines given to far smaller numbers of people, these rules can result in a large number of “adverse event” reports being submitted. The vast majority of these usually end up being false alarms, with the symptoms either unrelated to the drug in question or not an indication of any danger. But sometimes they do point to a critical safety issue that wasn’t picked up previously. Because the COVID-19 vaccine is being given to so many people, the volume of these reports is likely to be massive—far too many for humans to review fast enough to pick up any signs of a serious problem before it’s too late.

That’s why some governments are turning to A.I. to help. The British health regulator has contracted with Genpact to deploy machine learning software that can screen its official “yellow card” reports—which doctors and patients use to report unusual side effects that could be a cause for concern. The system Genpact built, which went live in December, takes in plain text, automatically codifies it, and searches for patterns that could be indicative of an emerging safety issue, flagging this to the regulator for further investigation, Sandor says.

Tiger Tyagarajan, Genpact’s chief executive officer, says the software has been trained on many different types of writing, so that it can understand both the medical terminology a doctor might use in reporting symptoms as well as the more colloquial expressions a member of the public might use. He says Genpact was only able to deploy this system quickly—it was up and running three months after Genpact received the contract—because the company had already built even more sophisticated versions of it for its pharmaceutical industry customers, particularly in the U.S. where the Food and Drug Administration requires stringent post-approval surveillance and reporting of any safety concerns with a medicine. In those cases, Genpact’s A.I. software doesn’t just look at a single government form, as it is doing in the U.K., it also scans medical journal articles and even social media posts for evidence of unusual symptoms that could be a cause for concern.

Some technologists have lamented that A.I. hasn’t been a big help during the pandemic. While some A.I. software helped sound early warnings that a worrisome new respiratory virus seemed to be circulating, the technology certainly didn’t help prevent the pandemic. And its impact on epidemiological modeling and policymaking has been minimal. It’s had limited impact in the quest to find COVID-19 treatments and develop vaccines.

Some have quipped that A.I. would be ready to combat the next pandemic, but not this one. Still, in helping to ensure that vaccines are distributed quickly and safely, the technology may yet prove its worth.

热读文章
热门视频
扫描二维码下载财富APP