AI取代收银员?亚马逊超市AI系统背后的人工真相

in STEEM CN/中文13 days ago (edited)

image.png

最近看到了一个有趣的新闻,目前AI技术非常火爆,被各行各业所挖掘。AI在本行业中的应用甚至在一些上班族中引起了恐慌,他们害怕自己的工作被AI取代。然而,这个新闻可能会让人们的担忧有所缓解。这则新闻是关于亚马逊在其线下超市启用了一个AI系统,该系统能够识别顾客从货架上拿走的商品,并自动记录到账单上,然后直接从顾客的信用卡扣款。使用这个系统后,亚马逊超市就不再需要设置收银台,顾客也不必在收银台前排长队,可以拿起商品就走,提供了一种非常流畅的购物体验。
但是,这样的系统可能会让收银员这样的岗位受到威胁,因为这是许多低学历人群能够从事的工作之一。如果这种系统普及,这些工作岗位的需求将会大大减少。不过,新闻很快出现了反转。原来这些所谓的“拿起就走”的超市并非完全依赖AI自动运行。实际上,亚马逊以非常低的工资,远程雇佣了一大批印度员工,通过远程摄像头监控并记录顾客的购物行为。所以,亚马逊的这个人工智能系统实际上是在使用人工来让它看起来更智能。
据说在自动取款机刚推出时,许多早期的ATM机里面实际上坐着一个工作人员,只是顾客看不到他们。如今亚马逊的这个超市AI系统可以说是完美重现了这个都市传说。当然,这不能完全说亚马逊是在作假,他们确实开发了这样一个能够识别用户购物行为的人工智能系统。按照最初的设想,人工识别只是作为AI识别的一个补充,在所有的识别场景中,只有不到30%需要人工处理。但实际应用中,发现顾客的购物行为非常复杂,AI无法准确识别的情况占比超过了一半,也就是说,在大部分情况下,AI无法进行准确识别,人工识别的比例接近60%。
这其实说明,在某些场景下,使用AI可能并不是最佳选择。AI的特点在于发现数据中的规律和模式,而在消费者购物这样具有高度随机性的场景中,AI能够发挥的作用非常有限,因为其中并没有特别明显的规律和模式。另一方面,这也表明,当前世界人口过剩,在许多第三世界国家存在大量低收入人群,利用这些低成本的劳动力可能是一种成本更低、效率更高的方式。因此,人们不应该盲目追求AI技术,也应该考虑将研究资源投入到其他领域中去。


这几天看到一个好玩的新闻,现在AI确实很火,各行各业都在挖掘Ai在本行业中的应用,甚至在一些上班族中都引起了恐慌,害怕自己的工作被AI所取代。但是看到这个新闻之后,也许人们就没有那么担心。

原来是亚马逊在他的线下超市里面启用了一个AI系统,可以识别顾客从货架上拿了哪些商品,然后记录到账单上面,直接从顾客的信用马上扣款,使用这个系统的亚马逊超市就不必设置收银台,顾客也不必在收银台前大牌长龙,可以拿起商品就走。在这样的超市,购物体验可以说那是相当的丝滑。

但是收银员的岗位也是很多低学历人士是能够从事的岗位之一。这样一来,这些人就要失去工作,对他们来说这真是一个悲商的故事。这个世界对这种要求和门槛较低的工作岗位的需求又少了好多。不过很快这个新闻就发生了神奇的反转。

原来这些所谓拿起就走的超市背后,并非完全依赖AI自动运行。实际上,亚马逊用非常低的工资,远程雇佣了一大批印度人,让他们通过远程摄像头来监控并记录顾客的购物行为。所以亚马逊的这个人工智能还真的就是使用人工,让它看起来更智能。

据说许多银行推出的早期自动取款机,其实里面就坐着一个工作人员,只是你看不到他而已。如今亚马逊的这个超市人工智能才真是完美重现了这个都市传说。当然也不能完全说亚马逊是在作假,他们也确实开发了这样一个识别用户购物行为的人工智能。按照最初的设想,人工识别仅仅是作为人工智能识别的一个补充,大约在所有的识别场景中,只有不到百分之三十需要人工处理。但实际应用的时候,发现顾客的购物行为五花八门。比如有的会拿起商品放进购物篮,过一会儿又改变主意,随便找个地方放下之类的。在实际使用中,AI无法识别的情况占比实际上已经超过一半了。也就是说,大部分的情况下,AI是无法进行准确识别的,而人工识别的比例接近百分之六十。

其实仔细想一想,这样的场景是不太适合使用AI的。AI的特点。通俗的说,就是发现数据中的规律和模式,而像用消费者购物这样的随机比较大的场景,AI能够发挥的作用是非常有限的,因为其中并没有特别明显的规律和模式。

不过另一方面也说明,现在世界人口过剩,在很多第三世界国家存在大量的低收入人群,将这些低成本的劳动力充分利用起来。也许是一件比使用AI技术成本更低、效率更高的方式。人们不应该为了AI而AI。也应该将研究AI的资资源投到其他的一些领域中去。


Recently I saw an interesting news that AI technology is currently very hot and is being mined by all walks of life. The application of AI in the industry has even caused panic among some office workers, who fear that their jobs will be replaced by AI. However, the news may ease some of those fears. The news is about Amazon's rollout of an AI system in its offline supermarkets that recognizes what customers take off the shelves, automatically records it on the bill, and then charges it directly to the customer's credit card. After using this system, Amazon supermarkets no longer need to set up checkout counters, customers do not have to queue in front of the checkout line, can pick up goods and go, providing a very smooth shopping experience.
However, such a system could put jobs like cashiers at risk, as this is one of the many jobs available to people with low education levels. If such a system becomes widespread, the demand for these jobs will be greatly reduced. But the news soon took a turn. It turns out that these so-called "pick up and go" supermarkets don't rely entirely on AI to run autonomously. In fact, Amazon hired a large number of Indian employees remotely at very low wages to monitor and record customers' shopping behavior via remote cameras. So, Amazon's AI system is actually using humans to make it look smarter.
It is said that when ATMs were first introduced, many of the early ATMs actually had a staff member sitting inside, but customers could not see them. Now Amazon's supermarket AI system is the perfect recreation of this urban legend. Of course, this is not entirely to say that Amazon is cheating, they did develop such an artificial intelligence system that can identify users' shopping behavior. According to the original vision, human recognition is only a supplement to AI recognition, and less than 30% of all recognition scenarios require manual processing. However, in practical applications, it is found that the shopping behavior of customers is very complex, and the proportion of cases that AI can not accurately identify is more than half, that is, in most cases, AI can not accurately identify, and the proportion of artificial recognition is close to 60%.
This actually shows that in some scenarios, using AI may not be the best choice. AI is characterized by finding regularities and patterns in data, and in highly random scenarios such as consumer shopping, AI can play a very limited role, because there is no particularly obvious regularities and patterns. On the other hand, it also suggests that the current world is overpopulated and there are large low-income populations in many third World countries, and utilizing these low-cost labor forces may be a less costly and more efficient way. Therefore, people should not blindly pursue AI technology, but also consider investing research resources in other fields.

Recently there seems to be a pupil jumping off the news, or happened in Hubei? However, it is difficult to investigate the responsible person for suicide, which is related to many factors such as everyone's values and psychological endurance. For others, the same thing may not even bother. Well, for some people it seems like a big deal, too big to pass.

The specific reason for this primary school student's suicide is said to be mostly related to the teacher's punishment. Seems to be punished to copy wrong questions ten thousand times, of course, this is hearsay, do not know the true or false, should not be so outrageous. But it should be common to copy hundreds of times. Of course people will of course many people will attribute this matter to the poor psychological bearing capacity of the parties, but also think that strict teachers are good teachers thank you, and this is an accidental incident, there does not seem to be much discussion value. But if you think about it, there's a lot to think about.

Since the German iron-blood chancellor Bismarck carried out social reform and established the compulsory education system, this system has basically become the object of copying homework in the world. Education has shifted from private consumption to state ownership. Of course, Bismarck's purpose was to mold the people into soldiers and workers who met the needs of the military state system, while instilling the concept of the state and the nation and cultivating the spirit of obedience. Because these are the two qualities that the army and the factories need most.

The nationalization of education also made the profession of teacher part of the system. Today, teachers are probably the most popular system in China after civil servants. Why do people like working in the system? A very classic metaphor is that the system is an iron rice bowl - in Mises's words, that is, the bureaucracy does not accept the test of profit accounting, there is no pressure to make profits, and drought and flood are guaranteed. Working in such a system environment, even idealists, after a period of polishing, will enter a state of lying flat, only seeking no fault and no merit.

However, in primary and secondary schools in China, everyone should have this feeling, that is, primary and secondary school teachers are particularly responsible, especially for poor students with bad temper teachers really like to use corporal punishment, humiliation and other means. Since it is already an iron rice bowl, why are you so serious and responsible? The academic performance of poor students does not seem to have any effect on the work of teachers. If, like this teacher, the students are forced to do something similar to suicide, it is likely to bear a great responsibility. And why this phenomenon of excessive responsibility only appears in the primary and secondary school stage, read the domestic university all know that the university teacher does not care whether you learn lessons, is scripted students doze off below, he is generally non-interference, so is the gourd bear is too big, not um bye drop drop it seems that the university teacher this state is the normal state. His scientific research funds and salaries are all issued by the government, and the students' grades are good or bad, and they can find a job, which has nothing to do with him. We don't say. Well, even for many professors, lecturing is not the main job at all - and the reason is that there is still a standard of evaluation for primary and secondary schools. Well, the person in charge of the school is the principal, and the quality of the principal's management of the school is still a very clear evaluation standard, that is, the scores and enrollment rates of the high school entrance examination and the college entrance examination. The school management is good, giving him a personal future is very meaningful and decisive, and ah pro score and correct rate will form a good word-of-mouth effect in the society. School teachers will also get the attention of students' parents, such as giving small gifts and other extra benefits and interests. I know that in this way, both schools and principals and parents will give teachers, primary and secondary school teachers pressure box power, find ways to improve, thank the students' performance. But this model is completely different from a fully market-oriented education or an educational institution with profit and loss accounting as an educational enterprise. Mm-hmm, no, is it? On what grounds? Because education companies, private education companies, it will be consumer-centric. If it is, it will value the feelings of consumers, and consumers are students and students' parents. Well, public school teachers are also responsible for raising students' grades. But in his eyes is not your poor academic performance of students, is to affect their own work performance stumbling block, so naturally there is no good temper, encounter extreme temper, bad teachers, naturally is a variety of humiliation corporal punishment, no longer care about the feelings of students oh. So we can see from this analysis that this is to get a good education. Only by fully exercising the rights of consumers, it is possible that education is handed over to the domestic government as a national welfare, which is mostly counterproductive. Well, good night. Who's that? Are you right?

Sort:  

Upvoted! Thank you for supporting witness @jswit.

学习了

Coin Marketplace

STEEM 0.28
TRX 0.12
JST 0.032
BTC 61672.72
ETH 2996.85
USDT 1.00
SBD 3.78