双语 | 为人工智能技术制定有效的知识产权战略

这是大岭为您分享IP英文的第104天：前几天我分享了IAM一篇关于人工智能知识产权保护策略的文章，永新知识产权的孟杰雄对其进行了翻译，再次感谢孟杰雄的热心翻译，也欢迎更多的伙伴投稿。

假期我将不再更新，祝大家国庆节快乐，祝愿祖国繁荣昌盛！

---

Buildingan effective IP strategy for AI

为AI制定有效的知识产权战略

As with many emerging technologies, AI raisesinteresting and often deeply unsettling fundamental ethical, moral, social,political and privacy issues. However, some questions touch on more practical issuesrelating to patentability

与许多新兴技术一样，AI产生了有趣且常常令人不安的基本伦理、道德、社会、政治和隐私问题。但是，有些问题涉及与可专利性相关的更实际问题。

Artificial intelligence (AI) is now encroaching onmany aspects of our lives and will be – according to Google’s CEO Sundar Pichai– more transformative than fire or electricity. Regardless of whether thischaracterisation will be borne out, AI is considered of such importance to bothMicrosoft and Google that both are reported to have undergone majorreorganisations to bring it to the forefront of their organisations in Marchand April of this year, respectively.

人工智能（AI）现在正在侵蚀我们生活的许多方面，并且——根据谷歌首席执行Sundar Pichai的说法——其将比火或电更具变革性。无论这种特性是否会得到证实，AI被认为对微软和谷歌都具有重要意义，据报道，两者都经历了重大的重组，分别在今年3月和4月将AI带到了各自组织的最前沿。

Figure 1.AI-related US patents and patent application filings

图1.与AI相关的美国专利和专利申请

As shown in Figure 1, the number of AI-related USpatents and patent application filings has risen substantially since 2008 fromonly a few patents in the late 1970s, which included a sewing machine with alearning mode and an automated speech recognition system. Figure 2 shows the approximatenumber of US patents granted and patent applications filed during the last fiveyears assigned to each of the listed 10 companies, which are among the mostrecognisable and dominant in the AI field.

如图1所示，自2008年以来，与AI相关的美国专利和专利申请的数量从20世纪70年代后期的几件专利（其中包括一台具有学习模式和自动语音识别系统的缝纫机）有了大幅增加。图2显示了在过去五年中所列出的10家公司中每家公司的美国专利和专利申请的大致数量，这些公司是AI领域中最知名和最主要的公司。

Although AI – particularly machine learning and itssub-field deep learning – is now being used for a broad spectrum ofapplications, a dilemma is beginning to emerge that is rooted in ourstill-limited understanding of how it works. One of the major technical issuesfacing deep learning has to do with its black-box character.

虽然AI ——特别是机器学习及其子领域的深度学习——现在被广泛应用，但是一种困境开始出现，这种困境源于我们对其工作方式仍然有限的理解。深度学习面临的主要技术问题之一与其黑箱特征有关。

Despite reassurances from several AI camps that we nowhave the tools necessary to decode the one or more mechanisms that underpin howAI does what it does, many organisations have recently voiced their concernsabout this black-box dilemma. In fact, some of the criticisms levelled at whatwe refer to here as ‘black-box AI’ have been quite pointed.

尽管来自几个AI阵营的保证，我们现在拥有必要的工具来解码一个或多个支持AI如何工作的机制，但许多组织最近表达了他们对这种黑箱困境的担忧。事实上，对于我们在这里称为“黑箱AI”的一些批评是非常尖锐的。

Figure 2. USpatents granted and patent applications filed during the last five years forthe 10 most prominent companies in AI

图2.  AI中最著名的10家公司在过去五年中提交的美国专利和专利申请

In an AI conference last December 2017, Ali Rahimi – aGoogle AI researcher and winner of the Test-of-Time Award at the 2017Conference on Neural Information Processing – asserted that machine-learningalgorithms have been transmuted into some kind of alchemy. 

在2017年12月举行的AI会议上，谷歌人工智能研究员，2017年神经信息处理会议的时间检验奖获得者Ali Rahimi断言，机器学习算法已被转化为某种炼金术。

Accordingly, AI researchers understand very little ofwhy certain algorithms work but others do not. In addition, the absence of areliable and rigorous standard for choosing one AI platform over another posesa serious concern. “There’s an anguish in the field,” Rahimi admitted. “Many ofus feel like we’re operating on an alien technology.” In view of theseconcerns, Rahimi and his cohorts presented several suggestions towardsaddressing the black-box issue in a paper presented at a subsequent AIconference in Canada, held in April 2018.

因此，AI研究人员很难理解为什么某些算法有效，但有些算法不能。此外，缺乏选择一个AI平台而非另一个AI平台的可靠且严格标准引起了严重关注。“在这个领域有一种痛苦”，Rahimi承认道。“我们中的许多人都觉得我们正在使用外星人技术。”鉴于这些问题，Rahimi和他的同伙们在随后2018年4月加拿大举行的AI会议上发表的一篇论文中提出了解决黑箱问题的几点建议。 

Yann LeCun –Facebook’s chief AI scientist in New York City – countered that spending timeto investigate how AI works sooner rather than later would detract from thedevelopment of cutting-edge AI techniques.

Yann LeCun——纽约的 Facebook的首席AI科学家——反驳说，花在探索AI如何工作的时间迟早会削弱尖端AI技术的发展。

However, releasing an extremely powerful tool foreveryone to use without understanding how it works or without implementingsafeguards would seem tantamount to courting disaster. We have seen this happenbefore in the case of the private and public sharing of hacking tools by both white(ethical) and black (malicious) hat hackers. The number of potentiallydetrimental AI capabilities could likewise increase without checks if we failto establish the knowledge and tools necessary to prevent or recover from anypotential harm that might be caused by black-box AI.

然而，发布一个非常强大的工具供所有人使用，而不了解它是如何工作或没有实施保护措施，似乎等于引发灾难。我们已经看到过这种情况发生在白帽黑客（伦理）和黑帽黑客（恶意）对黑客工具的私人和公共分享之前。潜在有害的AI能力的数量同样可能会增加，而无需检查我们是否能够建立必要的知识和工具来防止或恢复可能由黑箱AI引起的任何潜在伤害。

In addition to the argument that it is better to dosomething about a potentially disastrous situation while we still can, ratherthan pause to think about it only when irreversible damage has already beeninflicted, is the fact that many scientific advances would not have beenpossible without knowledge derived from both fundamental and applied research.Thus, allotting resources for both fundamental and applied AI research willlikely benefit everyone in the long run.

除了我们仍然可以更好地对潜在的灾难性情况采取行动，而不是仅在已经造成不可逆转的损害时停下来思考它，事实是若没有从基础和应用研究中获得的知识，许多科学进步是不可能的。因此，从长远来看，为基础和应用人工智能研究分配资源可能会使每个人受益。

To pre-empt any potential damage relating to AI’sblack-box nature, an organisation called the AI Now Institute has recommendedthat the US government and other agencies that deal with criminal justice,healthcare, welfare and education desist from using AI-based technologies.Several organisations – including the US Department of Defence, HarvardUniversity, Bank of America Corp, Uber Technologies Inc and Capital OneFinancial Corp – are also trying to address the same issue, as well as theproblem of bias involving advanced AI algorithms.

为了预防任何与AI黑箱性质相关的潜在损害，一个名为AI Now Institute的组织建议美国政府和其他处理刑事司法、医疗保健、福利和教育的机构不要使用基于AI的技术。一些组织——包括美国国防部、哈佛大学、美国银行、优步科技公司和第一资本金融公司——也试图解决同样的问题，以及涉及高级AI算法的偏差问题。

Reproducibility

再现性

As well as theblack-box conundrum, there is also the issue of reproducibility. This occurswhen researchers fail to replicate results generated by other researchers orresearch groups. While this problem can arise because of the lack of uniformityamong AI research protocols and publication practices, there are also fearsthat the two issues may be linked.

除了黑箱难题，还有重复性的问题。当研究人员无法重复其他研究人员或研究组产生的结果时，就会发生这种情况。虽然由于AI研究协议和出版实践之间缺乏统一性，可能会出现这个问题，但也有人担心这两个问题可能会有关联。

A failure tounderstand how AI arrives at its decisions during the intermediate and finalsteps of a learning process could well be the root cause of inconsistentresults generated by various AI researchers. In addition, there are othervariables that could contribute to this issue, including variations in thefollowing:

不能理解AI在学习过程的中间和最后步骤如何做出决定，成为各种AI研究人员产生的不一致结果的根本原因。此外，还有其他变量可能导致此问题，包括以下变量：

l algorithms used bydifferent researchers;

l the size, sourceand quality of training data used;

l the installedversion of the CPU/GPU/ASIC firmware; and

l the type orarchitecture of CPU/GPU/ASIC used to run an algorithm.
       *不同研究人员使用的算法；

l *所用训练数据的大小、来源和质量；

l *已安装的CPU/GPU/ASIC固件版本；以及

l *用于运行算法的CPU/GPU/ASIC的类型或体系架构。

Bug or feature?

错误还是特征？

Some contend thatAI’s black-box character is not a real issue at all. Science as a whole hasbeen faced with exactly the same interpretability issue for as long as we canremember. Yet, science – despite its shortcomings and failures and despitebeing constantly hounded by reproducibility issues in many research fields –continues to thrive and is now advancing at a faster rate than ever before.

有人认为AI的黑箱特征根本不是真正的问题。只要我们记得，科学作为一个整体就面临着完全相同的解释性问题。然而，科学——尽管存在缺点和失败，尽管在许多研究领域不断受到重复性问题的影响——仍在继续蓬勃发展，现在正以前所未有的速度发展。

Physicians, forexample, have been providing diagnoses for millions of patients involving allsorts of diseases based on much more limited data than that fed to neuralnetworks. No one has ever presented any conclusive evidence to show that morepatients are dying from misdiagnosis compared to those whose health and liveshave been improved as a result of treatment based on their physician’s expertopinion. However, this is a very different matter to having swarms of complex,non-ethical and amoral systems spread across the globe that are capable ofsimultaneously and immediately inflicting potentially serious harm to hundredsof thousands if not millions of people.

例如，医生一直在基于比提供给神经网络更有限的数据为数百万涉及各种疾病的患者提供诊断。没有人提供任何确凿的证据表明，与根据医生的专家意见治疗后健康和生活得到改善的患者相比，更多的患者因误诊而死亡。然而，这是一个非常不同的问题，因为成群的复杂、非伦理和非道德的系统遍布全球，能够同时并立即对数十万甚至数百万人造成潜在的严重伤害。

For this reason AIresearchers should always question whether the AI systems that they are creatingwill at least solve more problems than they generate.

出于这个原因，AI研究人员应该始终质疑他们正在创建的AI系统能解决的问题是否至少多于所产生的问题。

One possiblesolution might be to implement an equivalent of the drug approval processthrough clinical trials. Requiring AI systems to undergo initial testing usinga progressively increasing number of test systems or scenarios might be aneffective way to uncover lurking flaws that could cause an AI system to behaveunpredictably with potentially disastrous consequences.

一种可能的解决方案可能是实施相当于通过临床试验的药物批准程序。要求AI系统使用逐渐增大数量的测试系统或场景进行初步测试可能是揭示潜伏的缺陷会导致AI系统的不可预知行为造成潜在灾难性后果的有效途径。

Black-box AI and abstract ideas

黑箱AI和抽象概念

If one is dealingwith a black-box AI, one can argue that it should not be considered patentable.According to one of the definitions provided by the *Merriam-WebsterDictionary*, ‘abstract’ means “difficult to understand.”

如果一个人正在处理一个黑箱AI，可以争辩说它不应获得专利。根据《韦氏词典》提供的一个定义，“抽象”的意思是“难以理解”。

However, when adeep-learning algorithm is characterised as a black box, it usually means thatno one understands how it arrives at the decisions it uses during theprocessing of an input to generate an output. Based on these definitions of theterms ‘abstract’ and ‘black box’, a black-box AI would be considered anabstract thing. Yet according to US patent law, an abstract thing is notpatentable.

然而，当深度学习算法被表征为黑箱时，通常意味着没有人能够理解它在处理输入以产生输出过程中所使用的决策。根据“抽象”和“黑箱”的定义，黑箱AI将被视为抽象的东西。然而根据美国专利法，抽象的东西不可专利。

In fact, an AIwith inner workings that are not well understood, is much more conclusively anabstract thing compared to some business method patent claims that have beenadjudicated by the courts to be directed towards patent-ineligible abstractideas. The reason for this is that when AI researchers characterise adeep-learning algorithm as a black box, there is typically no disagreement thatthe AI’s black-box component is not well understood. In its present stage ofdevelopment, it is probably correct to say that an AI is either not wellunderstood or it is. All that is needed is for someone to provide conclusiveproof that there is no mystery at all as to how it works.

事实上，与一些已被法院裁定为不合格抽象概念的商业方法专利权利要求相比，其内部运作方式尚未得到充分理解的AI更是一种抽象的东西。这方面的原因是，当AI研究人员将深度学习算法描述为黑箱时，通常没有人会质疑AI的黑箱组件不好理解。在目前的发展阶段，可能正确地说AI要么不是很好理解，要么就是那样。所需要的只是让某人提供确凿的证据，证明它的工作原理并不神秘。

On the other hand,a compelling counter-argument to the non-patentability of a black-box AI isthat patents have been granted for many inventions whose inner workings areonly poorly understood, if at all. For example, many patented drugs’mechanisms-of-action are discovered only after the patents have issued. The factis, provided that a claimed invention satisfies the patentability criteriapertaining at least to utility, novelty, non-obviousness, written descriptionand enablement, the patent application claims stand a chance of being allowed.

另一方面，对于黑箱AI不可专利的一个引人注目的反驳论点是，许多内部运作方式很难理解（就算真有）的发明已被授予专利权。例如，许多专利药的作用机制在授权之后才被发现。事实是，如果要求保护的发明满足至少包括实用性、新颖性、创造性、书面说明书和可实施的可专利标准，则专利申请的权利要求应被允许。

Unfortunately, themeaning of the term ‘abstract idea’ as used in the patent field has itselfbecome something of a black box. The courts do not appear to have spent muchtime dwelling on its precise meaning nor have they provided a definition thatcould meaningfully guide patent practitioners. The more recent definition orinterpretation of the term appears to have been more a device of conveniencethat has merely forced any existing hints about its meaning deeper into arabbit hole – although it should be stated that trying to come up with a moreprecise and useful meaning of the term is no trivial task and could inadvertentlycreate more problems than it solves.

不幸的是，专利领域中使用“抽象概念”的含义本身就是黑箱。法院似乎没有花太多时间来研究它的确切含义，也没有提供可以有意义地指导专利从业者的定义。最近对这个术语的定义或解释似乎更像是一种方便的手段，只是将任何关于其含义的现有暗示深陷其中无法自拔——尽管应该说明试图提出该术语的更精确有用的含义不是一项微不足道的任务，可能会无意间产生比它解决的问题更多的问题。

Inventorship

发明人

AI is now beingused to analyse vast amounts of information from scientific journals,newspapers, magazines, Twitter and Reddit posts, Facebook accounts, Snapchat,patents and other references in order to:

AI现在被用于分析来自科学期刊、报纸、杂志、Twitter和Reddit的帖子、Facebook的账户、Snapchat、专利和其他参考文献的大量信息，以便：

·       extract trends,patterns and connections;

·       test hypothesesbased on existing data;

·       calculate thestatistical probability that a technique will work for a certain applicationunder certain conditions;

·       provide a rankedlist of recommendations on how to proceed to the next step of a research ordevelopment process; and

·       determine whethera combination of certain components, techniques or steps will produce a productor process that is more reliable, accurate, stable or fast.

*提取趋势、模式和联系；

*根据现有数据检验假设；

*计算某种技术在某些条件下适用于特定应用的统计概率；

*提供关于如何进行研发过程的下一步的建议的排序列表；以及

*确定某些组件、技术或步骤的组合是否会产生更可靠、准确、稳定或快速的产品或方法。

Not only has AIbeen used for many different applications, but it can carry them out at a scaleand speed impossible for humans.

AI不仅用于许多不同的应用，而且它可以以人类无法达到的规模和速度进行这些应用。

However, the waysthat AI has been used for those various applications essentially describes thecritical steps in the inventive process (or discovery). In many cases, thehuman participants merely input the relevant questions, any required initialdata or information inputs, or else the preliminary instructions regarding whatthe AI should look for and analyse. Thus, one can argue that in those cases thesubstantive act of inventing originates mainly, if not exclusively, from the AIitself.

然而，AI已经用于各种应用的方式基本上描述了创造过程（或发现）中的关键步骤。在许多情况下，人类参与者仅输入相关问题、任何所需的初始数据或信息输入，或者关于AI应该寻找和分析的初步指令。因此，人们可以争辩说，在这些情况下，发明的实质性行为主要来源于AI本身，如果不是唯一的话。

The question thenbecomes: should an entity (eg, an individual inventor or company) who reliesmostly on an AI to come up with an invention be entitled to any resultingpatent?

那么问题就变成了：主要依靠人工智能来提出发明的实体（例如，个人发明人或公司）是否应该有权获得任何最终的专利？

Obviousness

创造性

The legal standardused by the courts to determine whether a claimed invention is not patentablebased on obviousness is whether a hypothetical person of ordinary skill in theart would have considered the invention obvious as of the effective filing dateof the patent application.

法院基于创造性确定所要求保护的发明是否不具有可专利性的法律标准是本领域普通技术人员在专利申请的有效申请日之前是否会认为该发明是显而易见的。

If the courtsdecide that who or what did the inventing does not matter and the only thingthat counts is whether the owner was first to file a patent application, howthen do we deal with the obviousness criterion when it comes to an AI-inventedinvention?

如果法院判定发明的人或者发明什么并不重要，唯一重要的是所有者是否首先提交专利申请，那么当涉及到AI的发明时，我们如何处理创造性标准呢？

When addressingthe issue of obviousness as it relates to an AI invention, we can no longerrefer to the hypothetical person of ordinary skill in the art as presentlydefined by patent law. One possibility is that the courts will decide to expandthe definition of the hypothetical ‘person’ of ordinary skill in the art toinclude an AI.

当处理与AI发明有关的创造性问题时，我们不再能够参考目前由专利法定义的假设的本领域普通技术人员。一种可能性是法院将决定本领域普通技术的假设“人”的定义扩展到包括AI。

But in that case,we are faced with another dilemma: how do we determine if an invention wouldhave been considered obvious by an AI of ordinary skill in the art? Should aninvention by an AI be judged only according to whether a hypothetical AI,rather than a hypothetical human, of ordinary skill in the art would haveconsidered the claimed invention obvious? And what is ‘ordinary skill’ whenreferring to an AI?

但在这种情况下，我们面临另一个困境：我们如何确定一项发明是否会被本领域普通技术人员认为是显而易见的？是否仅根据作为本领域普通技术人员假设的AI而不是假设的人来判断AI的发明是否是显而易见的发明？在提到AI时，什么是“普通技术”？

One emerging areais focusing on how an AI arrives at a decision, given several possible optionsor scenarios. Because we do not yet know the limits of an AI’s capabilities, ifthey exist at all, and because the field itself will likely keep evolving formany years to come, we will probably have to wait before we can answer thequestion regarding what is obvious to a hypothetical AI of ordinary skill inthe art.

一个新兴领域正在关注AI如何做出决策、给出几种可能的选择或方案。因为我们还不知道AI能力的极限（如果存在的话），并且因为领域本身可能会在未来许多年内不断发展，我们可能必须等待，直到我们能够回答对于本领域普通技术人员的假设AI什么才是显而易见的问题。

However, if acertain functionality or capability has already been developed to maturity,perhaps we could reasonably assert by then that the ability of an AI to, say,distinguish a coyote from a jackal would have been obvious to an AI of ordinaryskill in the art as of the effective filing date of a patent application. Inthat case, an examiner might be within the bounds of USPTO rules when rejectinga claim based on obviousness by asserting that it would have been no stretch ofthe AI’s “imagination” and would thus be a routine matter to apply an AI’swell-documented feline identification skills to other animal genus or species.

然而，如果某个功能或能力已经发展到成熟，或许我们可以合理地断言，AI的能力，例如，将土狼与郊狼区分开的能力对于专利申请的有效申请日时本领域普通技术人员的AI来说是显而易见的。在这种情况下，审查员可能在美国专利商标局规则的范围内，基于创造性驳回权利要求，声称它不会扩展AI的“想象力”，因此是将AI有充分记录的猫科动物识别技能应用到其他动物属或物种的常规手段。

Utility and possession of theinvention

发明的实用性和拥有

Let us assume thatthe algorithm described in the patent application is generally acknowledged bypractitioners in the field to be substantially a black-box AI, which generatesunpredictable or inconsistent outputs owing to its sensitivity to, for example,the quality or amount of certain input data types (eg, the labelled examplesused in image recognition through deep learning). In this case, one can arguethat a patent applicant would not have been in possession of the claimedinvention as of the effective filing date of the invention. Otherwise how canone say that a patent applicant is in possession of the claimed invention – orif there is any invention at all – if the output of the claimed inventioncannot be predicted with reasonable certainty?

假设专利申请中描述的算法通常被本领域从业者承认为基本上是黑箱AI，因为其对例如某些输入数据类型（例如，通过深度学习在图像识别中使用的标记示例）的质量或数量的敏感性而产生不可预测或不一致的输出。在这种情况下，人们可以争辩说，在发明的有效申请日之前，专利申请人将不会拥有要求保护的发明。否则，如果不能合理确定地预测要求保护的发明的输出，怎么能说专利申请人拥有要求保护的发明——或者没有任何发明呢？

Further, would apatent application be enabling if the claimed AI algorithms are known toproduce only unpredictable outcomes? A rationale behind a patent application’senablement requirement is that any member of the public should be able topractice the invention when the patent term ends. However, if the most thepublic can expect to get from the patented invention are unpredictable outcomesis that a fair bargain? Also, since algorithms are not necessarily described ina detailed way in patent applications, who is to say that the members of thepublic who attempt to reproduce the invention would not end up with mostlyuseless results from the invention as described?

此外，如果已知要求保护的AI算法只能产生不可预测的结果，那么专利申请是否可以实施？专利申请可实施要求背后的理由是，任何公众成员都应该能够在专利期限结束时实施发明。但是，公众可以预见到从专利发明中获得的大多是不可预测的结果是否公平呢？而且，由于算法不一定在专利申请中详细描述，也就是说，试图重复发明的公众成员最终将不会得到所述发明的大部分无用结果？

Also, would an AIalgorithm that produces only unpredictable or inconsistent outputs satisfy theutility requirement, even if some of the outputs it generates are genuinelyuseful, although most of the times they are not? One possible, although perhapssnarky, answer is as follows: patent applications directed to gambling machineshave been previously granted. AI algorithms that yield unpredictable resultsare not that much different from slot machines. Therefore, AI algorithms thatproduce unpredictable results only should also be patentable.

另外，只产生不可预测或不一致输出的AI算法是否满足实用性要求，即使它产生的某些输出真的有用，尽管它们大部分时间不是？一个可能的，虽然可能是讽刺的答案如下：针对赌博机的专利申请已经被授权。产生不可预测结果的AI算法与老虎机没有太大区别。因此，产生不可预测结果的AI算法也应该是可专利的。

On the other hand,if some of the generated outputs of the claimed invention can be potentiallydangerous for certain applications under certain conditions because of theirunderlying unpredictability, it would seem unwise to reward these by issuingpatents for them.

另一方面，如果要求保护的发明产生的某些输出由于其潜在的不可预测性而在某些条件下对某些应用具有潜在危险性，通过授予它们专利权来对其进行奖励似乎是不明智的。

AI and availability of data

AI和数据的可用性

The other side ofthe equation that people do not talk about as much is data. The fact is,algorithms such as those used in neural networks require lots of training databefore they can become useful for accomplishing certain types of narrowlydefined tasks. For real-world applications (eg, drug discovery) one needs lotsof data much more sophisticated than the highest resolution images of all theknown cat species in the world.

人们不会谈论的另一方面是数据。事实上，诸如神经网络中使用的算法需要大量的训练数据才能完成某些类型的狭义任务。对于现实世界的应用（例如，药物发现），人们需要比世界上所有已知猫科动物的最高分辨率图像更复杂的数据。

Thus, one might bein possession of the most advanced deep-learning algorithms in the world, butwithout access to large amounts of high-quality data needed for drug discoverytraining (eg, data relating to structure-function relationships, results fromanimal models, x-ray diffraction data and human clinical trials), the output ofthe neural network algorithm would most likely be unreliable, if not unusable.Garbage in, garbage out.

因此，人们可能拥有世界上最先进的深度学习算法，但无法获得药物发现训练（例如，与结构-功能关系相关的数据、动物模型的结果、X射线衍射数据和人体临床试验）所需的大量高质量数据，神经网络算法的输出很可能是不可靠的，如果不是不可用的话。无用输入，无用输出。

The problem iswhile there are huge amounts of chemical, biological and physical data frompublished scientific journals, the most important data is that which is mostrelevant to the specific drug or class of drugs covered by the drug discoverystudy. Unfortunately, this critical data is not usually publicly accessible.According to IBM’s CEO Ginny Rommety, only 20% of the world’s data issearchable, while the rest is in the hands of established businesses.

问题是虽然已发表的科学期刊有大量的化学、生物和物理数据，但最重要的数据是与药物发现研究所涵盖的特定药物或药物类别最相关的数据。不幸的是，这些关键数据通常不能公开获得。根据IBM首席执行官Ginny Rommety的说法，世界上只有20％的数据是可搜索的，而其余数据则掌握在已建立的企业中。

So, AI-basedinventions that require huge amounts of data to be useful will likely matteronly in terms of real-world utility and reliability to those who are already inpossession of the required data. Given this, there will probably be very fewpotential licensees for those patented inventions and these will likely belarge companies. In this case, the public, or even the companies themselves whofile the patent applications, would probably not benefit much from patentsgranted for those types of inventions. 

因此，需要大量数据才有用的基于AI的发明可能只对已经拥有所需数据的人提供实际效用和可靠性。鉴于此，这些专利发明的潜在被许可人可能很少，而这些可能是大公司。在这种情况下，提交专利申请的公众、甚至公司本身可能不会从授予这类发明的专利中获益。

Patent AI or keep as tradesecret?

对AI进行专利保护还是作为商业秘密？

One importantquestion is when would it make sense to protect AI-based inventions as tradesecrets, rather than via patents? If a company has been developing, forexample, its own neural network algorithms for its drug-discovery studies, it mightbe better off protecting any inventions or discoveries it generates relating tothe drug-discovery studies as trade secrets, including the neural networkalgorithms that otherwise might be patentable.

一个重要的问题是，什么时候以商业秘密而不是通过专利来保护基于AI的发明好？例如，如果一家公司为其药物发现研究开发了自己的神经网络算法，最好以商业秘密保护所产生的与药物发现研究有关的任何发明或发现，包括其实可以取得专利的神经网络算法。

A potentialdownside to having an invention published through patents or published patentapplications is that the disclosure in the patent could help a patentee’scompetitors progress more quickly with their own drug-discovery efforts for acompeting drug or class of drugs. Also, if a patentee grants a licence to itscompetitors for the use of its patented neural network algorithms, thecompetitor could use the patented invention as a basis for creating andperfecting its own neural network algorithms, which could end up being far moresophisticated than those covered by the patent.

通过专利或公开的专利申请公布发明的潜在缺点是，专利中的公开可以帮助专利权人的竞争对手通过他们自己的药物发现努力更快地开发出竞争药物或竞争类药物。此外，如果专利权人授予其竞争对手使用其专利神经网络算法的许可，竞争对手可以使用该专利发明作为创建和完善其自身神经网络算法的基础，这可能最终比专利所涵盖的那些算法更先进。

Keeping a neuralnetwork algorithm a trade secret has its benefits. For one thing it can be usedexclusively for far longer than would be possible if it was patented. Also, itmay be relatively easy to keep it a trade secret because many of theimprovements on the algorithm could be highly dependent on the type, source,amount and quality of data, all of which are exclusively available to thecompany that developed the patentable algorithm.

将神经网络算法作为商业秘密进行保护有其好处。一方面，它可以排他使用的时间比获得专利的时间长得多。此外，保持商业秘密可能相对容易，因为算法的许多改进可能高度依赖于数据的类型、来源、数量和质量，所有这些都是仅开发可专利算法的公司才能获得的。

Action plan

行动计划

In developing anAI IP protection strategy there are some key points to bear in mind:

在制定AI知识产权保护战略时，需要牢记以下几个要点：

·       Weigh the pros andcons of filing patent applications versus protecting AI-related inventions astrade secrets.

·       Consider otheroptions such as cross-licensing if it would take too much time and resources todevelop certain AI-based technology in-house.

·       Test your AIlearning algorithms for potential reproducibility and bias issues.

·       Ensure as much aspossible that you are using reliable data as inputs for your AI learningalgorithms, especially for critical applications.

·       *权衡提交专利申请与将AI相关发明作为商业秘密进行保护的利弊。

·       *如果在内部开发某些基于AI的技术需要花费太多时间和资源，请考虑其他选项，例如交叉许可。

·       *测试您的AI学习算法是否存在潜在的重复性和偏差问题。

·       *尽可能确保使用可靠的数据作为AI学习算法的输入，特别是对于关键应用。 

Source:https://www.iam-media.com/patents/building-effective-ip-strategy-ai

Each article is copyrighted to their original authors. The news is for informational purposes only and does not provide legal advice.

--End--

往期相关内容：

---

P.S. 我开设了一个专利实务知识学习平台，我会带领大家共同学习两本中文教程、三本英文教程、中国、美国和EPO的专利审查指南，以及近十年最高院发布的专利指导案例，以及美国、EPO的重要判例，共计100余件。我还会指导大家进行法律分析和写作。可以随时加入，时间为加入后的一年，按照自己的进度完成每周任务即可，目前已经有140+伙伴加入，具体情况见：专利实务能力提升计划

加入计划，扫描下图中的二维码