vlambda博客
学习文章列表

【区块链+人工智能】人工智能+区块链:双倍的炒作还是双倍的价值?


译者按: 过去的几天,是“区块链”成功“出圈”,引起极大关注和广泛讨论的几天。 把区块链作为核心技术自主创新的重要突破口和主攻方向的讲话,无疑成为了推动国内区块链行业快速发展的强心剂。


不过在积极拥抱区块链的同时,不少专业、行业内人士也表示,区块链不是用来“炒”的,应净化市场的浮躁。 诚如他们所言,在冷静之后,我们仍需关注的是真切的行业发展,是实实在在的应用落地。


下面这篇文章,对于关注区块链行业应用、关注区块链技术与其他新兴技术融合的人们来说也许能提供一些新的思路。


来源 | Forbes(转载请注明来源)

作者 | Ron Schmelzer

译者 | 王婕

编辑 | 蒲蒲


人工智能(AI)作为一个充满炒作的市场,供应商、消费者和媒体都在高谈阔论人工智能的大体功能和它们的具体产品。无独有偶,区块链也是一个被广泛炒作的市场,技术供应商和其客户都声称区块链具有各种各样的功能,无论它们存在与否。


那么,人工智能与区块链的融合是不是就自然而然地成为了双重炒作?


而从另一方面来看,人工智能正以我们每天谈论的无数种形式提供真实的、有形的价值。同样的,区块链也正在通过一系列应用程序及行业应用彰显其价值。因此,将AI和区块链相结合也许反而会得到双重效果。


 区块链在人工智能环境中的作用 


区块链是一个分散的、分布式的交易分类账本,它包含透明性、可信任、可验证性和智能合约等特性。分散和分布式意味着信息以这样一种方式存储在网络上后,其中的每个节点都可以访问数据,而不需要获得中心服务器的许可。网络也是分布式的,因为交易发生在每个终端,不需要集中协调。分类账本是交易的记录。区块链记录了两个独立的交易方之间的互动记录,无论是金融交易所,还是监管链,都显示了随着时间的推移,事情何时发生了变化。


由于区块链中的每个区块都包含不同的经过加密或编码的信息,因此区块链可以帮助确保数据的可信任性和可验证性。区块链中的“链”和“区块”的概念是,每一个“区块”都有自己的信息,这些信息包含了与其前面的“区块”的链接,这些链接延伸“链”并形成了可验证的监管链。


任何单个参与者都不可能在不使特定“区块”中的整个信息链失效的情况下更改“区块”中的信息,从而破坏这个“链”。由于该链分布在无数其他地方,并且在其他各方之间,一旦要对该链进行更改,就需要所有各方的一致同意。


谈到区块链概念,就势必会谈到智能合约。它是分散的代码片段,可以在满足特定的操作链时触发。通过这种方式,当你有两个参与方希望执行安全、可信的交易而不需要使用中介时,这将成为区块链的理想用途。


那么区块链该如何赋能人工智能?


第一个好处就是你可以在无中介的情况下,在所有各方之间共享机器学习模型。面部识别软件就是一个很好的例子:如果一个设备知道一个人的相貌并将其上链,那么其他连接的设备也能知道这个人的相貌。


当其他设备上传自己的面部识别数据时,其他设备将获得将其用于面部识别模型的能力。由于这发生在区块链上,没有对面部识别的中央控制,因此没有一家公司能够拥有或存储这些数据。这种方法还允许每个人都能通过使用集成的人工智能和区块链来更快地参与学习和协作。


人工智能系统还可以通过使用区块链来促进跨多个模型的数据共享。一个很好的例子是在在线零售中使用机器学习模型进行产品推荐。如果在线商店知道某个购物者的偏好,之后若该客户转到了另一个商店的网站,则可以通过区块链连接这两个站点,以共享可信的个性化信息。


这可能会成为那些希望彼此间共享个性化信息的小型电子商务网站的必争之地。不同于每家公司收集他们自己的个性化数据,他们可以在区块链中互相分享这些数据,这无疑加剧了亚马逊和沃尔玛这样的网站之间的竞争,尤其是当这些网站已经开发了自己的数据系统来收集客户的这些信息。而这对客户的好处是,作为信息被分享的交换,他们可以从中得到更好的价格、定制化的购物体验等服务。如果所有信息和支付系统都通过区块链存储和共享而不是集中的数据服务器,这还可以防止数据被泄露。


 区块链如何赋能人工智能 


区块链不仅可以被用来共享模型和数据,还可以以一种跨多个人工智能系统共享的方式,帮助扮演“智多星”的角色。如果我们能够将所有共享学习的优势和区块链、人工智能结合起来,那么就能将机器学习结合起来,然后使这些学习结果与网络上所有的人工智能系统共享。这样一来的好处就是没有人能独握它,即使是政府也无法控制机器人或共享的大脑,它可以是不偏不倚的,因为大量的信息都来自不同的领域和不同的角度。


另一个应用是解决可释AI的挑战。深度学习领域的一个更重要的问题是,对于什么输入导致什么输出,以及它们如何影响整个序列尚没有一个清晰的概念。如果在深度学习神经网络中出现了什么问题,我们就无法知道如何识别和纠正这些问题。这实际上就是神经网络的“黑盒子”问题,毫无真正的透明性或可解释性。


然而,如果我们使用区块链,我们就可以记录个人行为如何以一种不规范的方式导致最终决定,并可以回头看看哪里出了问题,然后解决它。区块链将被用来记录事件,且在之后不会被修改,例如自动驾驶汽车的决策和行动;同时这也可以增加信任,因为区块链元素是客观的,只是用于存储和分析,任何人都可以进入其中查看发生了什么。


最后,人工智能系统通常可以用来改进区块链。机器学习系统可以持续关注区块链中发生的事情、查找存储数据类型中的模式和异常,以及观察在特定服务器上执行的操作,并用于在可能发生紧急事件时向用户发出警报。同时,机器学习系统还可以寻找正常的行为,并标记出什么是不寻常的。人工智能系统可以帮助区块链更安全、更可靠、更高效。


虽然AI与区块链融合的世界很有可能充满了不切实际的炒作,但也的确存在着实际、真切且现实的方法,可以使这两个新兴技术实现彼此受益,并为那些希望在其工作平台中实施这些技术的人提供真实的成效。


原文如下


AI and Blockchain: Double the Hype or Double the Value?

                                

Artificial Intelligence (AI) as a market is full of hype, with vendors, customers, and press all speaking breathlessly about the capabilities for AI in general and their offerings specifically. Likewise, blockchain is also a widely hyped market, with technology providers and customers claiming all sorts of capabilities that may or may not be possible. Combining AI and blockchain then must be double the hype? On the other hand, AI is providing real, tangible value in many myriad ways we talk about every day. Likewise, blockchain is starting to show value across a range of applications and industry. So, perhaps combining AI and blockchain will also show twice the value combined together. 


The role of blockchain in the context of AI


Blockchain is a decentralized, distributed ledger of transactions that has elements of transparency, trust, verifiability, and something called smart contracts. Decentralized and distributed means that information that is stored across the network in such a way that each end point has access to the data without requiring access to a central server. The network is also distributed because the transactions happen at each end point without requiring centralized coordination. A ledger is a record of transactions. Blockchain records a ledger of interactions between two separate parties whether it be a financial exchange or even a chain of custody showing when things have changed hands over time.


Since every block in the blockchain contains a different piece of information that is encrypted or encoded, the blockchain can help guarantee trust and verifiability of data. The concept of the chain and block in blockchain is that each block has its own information and that information contains a link to the block before it, which develop the chain and provides a verifiable chain of custody. No individual actor can change the information in a block without invalidating the entire chain of information in a particular block and thus messing up the chain. Since the chain is distributed in myriad of other places and between other parties, to make changes to the chain, a consensus of all the parties would need to agree to the changes being made. 


Adding to the concept of the blockchain are smart contracts, which are decentralized pieces of code that can be triggered when a specific chain of actions has been met. In this way, when you have two parties wanting to execute a secure, trusted transaction without the use of an intermediary this is what blockchain ideally would be used for. 


So how can blockchain help with AI? The first benefit is that you can share machine learning models among all parties without an intermediary. A good example is with facial recognition software. If one device knows what a person looks like and uploads it to the chain, other devices hooked up also know what that person looks like. When other devices upload their own facial recognition data, the other devices will gain the ability to use that for the facial recognition model. Since this happens on the blockchain, there is no central control over facial recognition, and as such no one company owns or stores the data. This approach also allows for everyone to learn faster and collaboratively through the use of integrated AI with blockchain.


AI systems can also use blockchain to facilitate the sharing of data used across multiple models. A great example is the use of machine learning models for product recommendations in online retail. If an online store knows the preferences of one shopper, and then that customer goes to a different store website, these two sites can be connected through a blockchain to share trusted personalization information. This could potentially become a place of competition for smaller ecommerce sites that want to share with each other personalization information. Instead of each company gathering their own personalization data, they can share it amongst each other in a blockchain, providing competition to sites like Amazon and Walmart who have already developed their own data systems to gather this information about their customers. The benefit for this to the customer is that in exchange for sharing their information, they can get everything from better pricing to customized shopping experiences. This could also prevent data breaches if all information and payment systems are stored and shared through a blockchain rather than a centralized data server.


How blockchain can benefit AI


Not only can blockchains be used to share models and data, but blockchains can help serve a role as a “master brain” in a manner shared across multiple AI systems. If we can put all of these shared learning benefits and blockchain and AI together, the possibility may also be there to combine all of these things that can learn from their surroundings and then share that learning with all the AI systems on the network. A major benefit would be that no one owns it and there’s no government control over the bot or the shared brain. It could potentially be unbiased because of the sheer amount of information coming in from different areas and different angles.


Another application is addressing the challenge of explainable AI.One of the more significant problems with deep learning is that there isn’t a clear idea about what inputs result in what output and how that affected the whole sequence. If something goes wrong in a deep learning neural network, we don’t have a clear idea of how to identify the problem and correct it. This is the problem of neural networks in effect being a “black box” without any real transparency or explainability. However, if we use blockchains, we can record how individual actions result in a final decision in a non-reputable manner, which allows us to go back and see where things went wrong and then fix the problem. The blockchain would be used to record events, such as autonomous vehicle decisions and actions in a way that will not be modified later. This can also increase trust since the blockchain element is unbiased and is just for storage and analysis, anyone can go in and see what has happened.


Finally, AI systems can be used to improve blockchains in general. Machine learning systems can keep an eye patterns and anomalies in the types of data being stored and actions being performed on a particular server and be used to alert users when something may be happening. The systems can look for normal behavior and flag what seems to be unusual. The AI systems can help keep blockchain more secure, more reliable, and more efficient.


While it’s quite possible that the worlds of AI and blockchain are full of hype, there are actual, tangible, realistic ways in which the two emerging technologies can be used in ways that benefit each other and provide real outcomes for those looking to implement the technologies today in their environments.


先进制造业+工业互联网




产业智能官  AI-CPS


加入知识星球“产业智能研究院”:先进制造业OT(自动化+机器人+工艺+精益)和工业互联网IT(云计算+大数据+物联网+区块链+人工智能)产业智能化技术深度融合,在场景中构建状态感知-实时分析-自主决策-精准执行-学习提升的产业智能化平台;实现产业转型升级、DT驱动业务、价值创新创造的产业互联生态链。


产业智能化平台作为第四次工业革命的核心驱动力,将进一步释放历次科技革命和产业变革积蓄的巨大能量,并创造新的强大引擎; 重构设计、生产、物流、服务等经济活动各环节,形成从宏观到微观各领域的智能化新需求,催生 新技术、新产品、新产业、新业态和新模式; 引发经济结构重大变革,深刻改变人类生产生活方式和思维模式,实现社会生产力的整体跃升。

产业智能化技术分支用来的今天,制造业者必须了解如何将“智能技术”全面渗入整个公司、产品、业务等商业场景中, 利用工业互联网形成数字化、网络化和智能化力量,实现行业的重新布局、企业的重新构建和焕然新生。

版权声明产业智能官(ID:AI-CPS推荐的文章,除非确实无法确认,我们都会注明作者和来源,涉权烦请联系协商解决,联系、投稿邮箱:[email protected]