意見(Opinion)

Clark Stanley, also known as the ‘Rattlesnake King’, was an (in)famous entrepreneur from late 19th century. A cowboy turned medicine-man, Stanley was the first businessman to start selling ‘snake oil’ as a panacea for many ailments. In a notorious demonstration at a symposium in Chicago, Stanley took a live rattlesnake, sliced it open, dropped it in boiling water, and skimmed off the fat that rose to the surface, to show the authenticity of his snake oil. The public lapped it up. His snake oil was a hit.

克拉克·斯坦利(Clark Stanley)，也被稱為“響尾蛇之王”，是19世紀後期的著名企業家。 斯坦利是一位牛仔，後成為一名醫學家，是第一個開始銷售“蛇油”作為許多疾病的靈丹妙藥的商人。 在芝加哥的一個研討會上，臭名昭著的示威遊行中，史丹利拍了一條活的響尾蛇，將其切成薄片，將其放入沸水中，然後撇去表面上的脂肪，以證明其蛇油的真實性。 公眾對此表示讚賞。 他的蛇油很受歡迎。

But there were some issues. Big ones.

但是有一些問題。 大個子

1) Stanley’s Snake Oil did not do what it claimed to do. It did not have the efficacy against ailments that it professed.

1)斯坦利的蛇油沒有做到它聲稱的那樣。 它沒有它所聲稱的對抗疾病的功效。

2) Also, as with all marketing gimmicks, what you see is not always what you get. After over two decades of selling ‘Stanley’s Snake Oil’, authorities finally examined his magic potion. What they found in it was mineral oil, beef fat, chilli peppers and turpentine! No sign whatsoever of any snake oil was detected.

2)而且，與所有營銷頭一樣，您看到的並不總是所得到的。 在出售“斯坦利的蛇油”超過二十年之後，當局終於檢查了他的魔藥。 他們在其中發現的是礦物油，牛肉脂肪，辣椒和松節油！ 沒有發現任何蛇油的跡象。

The fraud was so well orchestrated by Stanley and went on for so long, that even to this day the euphemism ‘snake oil salesman’ denotes fraudulent salesmanship and making claims far removed from reality.

欺詐行為是斯坦利精心策劃的，並且持續了很長時間，以至於直到今天，委婉的“蛇油推銷員”仍是欺詐性推銷員，使索賠要求與現實相去甚遠。

So why am I talking about snake oil salesmen in the context of AI?

那麼，為什麼我要在人工智慧的背景下談論蛇油推銷員呢？

Someone famously said, ‘data is the new oil’. True enough. But you know what the new ‘snake oil’ is? Artificial Intelligence!

有人曾說過“資料就是新石油”。 足夠真實。 但是您知道什麼是新的“蛇油”嗎？ 人工智慧！

Allow me to explain. Everywhere you seem to look today there is someone doing something (or many things!) with Artificial Intelligence. AI is ubiquitous. It’s everywhere. While that is true in many ways — we really do use AI in many, many aspects of our lives — the problem lies in how loosely the term is interpreted. Just like all that glitters ain’t gold, all that is being done by a computer isn’t AI. Very simplistically put, AI refers to machines being made intelligent in order to replicate human actions and/or decision making to achieve set goals. (For more details you can also read my article on Decoding AI here).

請允許我解釋。 您今天所看到的每個地方都有人在用人工智慧做某事(或很多事情！)。 人工智慧無處不在。 到處都是。 儘管這在許多方面都是正確的-我們確實確實在生活的許多方面都使用了AI-但問題在於該術語的解釋程度如何。 就像所有閃閃發光的東西都不是金子一樣，由計算機完成的所有事情都不是AI。 簡而言之，人工智慧是指使機器智慧化，以便複製人類的行為和/或決策以實現既定目標。 (有關更多詳細資訊，您還可以在此處閱讀我有關解碼AI的文章)。

The fallout of this deceptive AI-led marketing is profound for the startup ecosystem.

對於初創企業生態系統而言，這種由欺騙性的AI主導營銷的後果是深遠的。

Startups around the globe, are touting many things as being AI driven, when in reality there is little to no true AI involved. And since most people barely understand what AI really is, for no fault of their own, they fall for all the gimmickry. A survey conducted in 2019 found that 40% of the nearly 2,900 AI startups in Europe did not use AI in any substantial form.

全球範圍內的初創公司都在吹捧許多事情是由AI驅動的，而實際上幾乎沒有涉及到真正的AI。 而且由於大多數人幾乎不瞭解AI的真正含義，因此，即使沒有自己的錯，他們也會為所有的all頭而屈服。 2019年進行的一項調查發現，歐洲近2900家AI初創公司中有40％沒有以任何實質性形式使用AI。

The fact that terms like AI and machine learning have become bait for VCs globally does not help the cause. Investors line up for anything that remotely claims to be AI-adjacent. In fact, the same survey quoted above, concluded that startups that claim to work in AI attract between 15 to 50 percent more funding compared to other software companies. Startups therefore have an almost legitimate reason to try to position themselves, by hook or crook, in the AI ecosystem.

AI和機器學習等術語已成為全球VC的誘餌，這一事實無濟於事。 投資者排隊尋找任何聲稱與人工智慧相鄰的東西。 實際上，上面引用的同一項調查得出的結論是，聲稱在AI中工作的初創公司比其他軟體公司吸引了15％至50％的資金。 因此，初創企業幾乎有合理的理由試圖通過鉤子或騙子在AI生態系統中定位自己。

If there is one stakeholder group that pays a heavy price for the fast and loose usage of AI jargon by startups, it is potential employees. There are tens of thousands of young professionals, some just out of college, looking to fulfill their dreams of working in the emerging domain of data science and AI. They are lured by the promise of building the next AI unicorn and of getting the opportunity to work on the most cutting-edge analytics problems. The reality though could often be far from this, which would lead to disappointment and regret across the board. The employee is left feeling cheated, and the company has to deal with problems of high employee turnover necessitating knowledge transfer and even rebooting development efforts (always an expensive proposition!).

如果有一個利益相關者群體為初創企業快速而寬鬆地使用AI行話付出了沉重的代價，那麼它就是潛在的員工。 有成千上萬的年輕專業人員，其中一些剛剛大學畢業，他們正在努力實現他們在資料科學和AI新興領域工作的夢想。 他們被建立下一個AI獨角獸的希望所吸引，並有機會從事最前沿的分析問題。 儘管現實常常離這一目標相去甚遠，這將導致失望和全面的遺憾。 員工感到被騙了，公司必須處理員工離職率高的問題，這需要知識轉移，甚至重新啟動開發工作(總是很昂貴的提議！)。

Therefore, as a potential employee of an AI startup, you must ask the right questions and assess the depth of what the company is doing or plans to do. And as the founder of an AI or ML startup, you are always better off being upfront with potential team members about what the data science team would be expected to do.

因此，作為AI創業公司的潛在員工，您必須提出正確的問題，並評估公司正在做或計劃做的事情的深度。 作為AI或ML初創公司的創始人，與潛在的團隊成員一起提前做好關於資料科學團隊的期望總是更好。

Having been in the domain of data science, AI and machine learning for the last 15 odd years, I come across many disillusioned young people who thought they were signing up for a glorious career, but instead found out that most of their time was spent in cleaning or labeling datasets. Following is a list of questions I always advise them to ask when they interview for a job with an AI startup. These not only help a potential employee better appreciate what he/she is getting into, it also demonstrates to the founders that you understand the nuances involved.

在過去的15多年裡，我一直處於資料科學，人工智慧和機器學習領域，我遇到了許多幻滅的年輕人，他們以為自己報名參加了光榮的職業，但他們發現大部分時間都花在了清洗或標記資料集。 以下是我始終建議他們在面試AI初創公司工作時要問的問題列表。 這些不僅可以幫助潛在員工更好地理解他/她正在從事的工作，還可以向創始人表明您瞭解所涉及的細微差別。

1.服務還是產品？ (1. Services or Product?)

This is an important distinction. The work one does in an AI services company will be very different from that in a products company. While services would usually provide more variety, building a product would go into far greater depth.

這是一個重要的區別。 在AI服務公司中所做的工作將與在產品公司中所做的工作截然不同。 雖然服務通常會提供更多的種類，但是構建產品的深度會更大。

2.如果提供服務，那麼什麼樣的服務？ (2. If services, then what kind of services?)

Services work would in most cases be like consulting, i.e. delivering AI solutions or models to clients to solve specific problem statements. Therefore, it is vital to ask what sort of problem statements does the startup work on for its clients? Is there a wide variety in the problem statements or is there a well-defined set they address?

在大多數情況下，服務工作就像諮詢，即向客戶提供AI解決方案或模型以解決特定的問題陳述。 因此，至關重要的是要問啟動為客戶處理什麼樣的問題陳述？ 問題陳述中是否存在各種各樣的內容，或者它們解決的定義是否明確？

Also, what part of the services delivery would you be expected to get involved in — data aggregation/cleansing, model building, manual training set creation etc.? It is important to realize that there will always be grunt work involved but having some sense of the proportion of time expected to be spent on actual model development, research on techniques etc. is always helpful.

另外，您希望服務提供的哪一部分參與其中-資料聚合/清理，模型構建，手動培訓集建立等？ 重要的是要意識到總是會涉及艱鉅的工作，但是對預期用於實際模型開發，技術研究等方面的時間比例有所瞭解總是有幫助的。

3.如果是產品，那是什麼樣的產品？ (3. If product, then what kind of product?)

The vast majority of AI startups would fall in the products bucket. But not all products are the same. Being able to understand a little more about the type of product a startup is building can provide insights into the kind of data science or machine learning work that may be involved. Provided below is a broad guide for types of products that startups could typically be involved with.

絕大多數AI初創公司都屬於產品類別。 但並非所有產品都是相同的。 能夠對創業公司正在構建的產品型別有更多瞭解，可以提供對可能涉及的資料科學或機器學習工作的見解。 以下提供的是初創企業通常會涉及的產品型別的廣泛指南。

Once you understand the type of AI product being built, dig deeper to find out what specific aspect of product development would you be involved in? It’s also often helpful to understand the medium and long term product vision — startup founders would usually (hopefully!) have a fair idea about how they envision their product to evolve.

一旦瞭解了要構建的AI產品的型別，就可以更深入地瞭解您將參與產品開發的哪些特定方面？ 瞭解中長期產品願景通常也很有幫助-初創公司的創始人通常(希望！)對他們如何設想產品的發展有一個清晰的想法。

4.正在使用什麼技術堆疊？ (4. What is the tech stack being used?)

This question is important for two reasons. Firstly, it will help you get some insight into product vision. Secondly, it will also help you assess if there are specific skills or languages that you may need to pick up or brush up on if you join the company.

這個問題很重要，有兩個原因。 首先，它將幫助您深入瞭解產品願景。 其次，它也將幫助您評估如果加入公司，是否需要掌握或掌握某些特定技能或語言。

5.他們通常將使用哪種型別的資料來構建和測試模型？ (5. What type of data would they typically use to build and test models?)

The reason this question is important is because it helps separate the wheat from the chaff. Startups that are deeply committed to an AI-first strategy, will have a very good understanding of the data they would use for product development. For example, a company that wants to build a product that uses computer vision technology to have machines predict cancer occurrence by analyzing CT scan images, needs to have access to a large repository of images to train and build models.

這個問題很重要的原因是因為它有助於將小麥與穀殼分離。 致力於AI優先策略的初創公司將對用於產品開發的資料有很好的瞭解。 例如，一家公司想要開發一種使用計算機視覺技術的產品，以使機器通過分析CT掃描影象來預測癌症的發生，則需要訪問大型影象庫以訓練和構建模型。

If you get a solid answer to this question, you can usually rest assured that the startup is on relatively strong footing.

如果您對這個問題有一個肯定的答案，通常可以放心，創業公司的地位相對較強。

6.創始團隊的背景？ (6. Founding team’s background?)

You may not need to explicitly ask this question, and instead could just do your own research to find this out. This question is important because AI is a largely technical domain. It is also a dynamic, rapidly evolving domain. Founders (or part of the founding team) having prior experience in the AI space would be critical to ensuring a higher probability of success because they would intimately know the challenges, would be able to guide the teams and shepherd the startup in the right direction.

您可能不需要明確地問這個問題，而是可以自己做研究以找出答案。 這個問題很重要，因為AI在很大程度上是技術領域。 它也是一個動態，快速發展的領域。 擁有AI領域經驗的創始人(或創始團隊的一部分)對於確保更高的成功概率至關重要，因為他們會充分了解挑戰，能夠指導團隊並向正確的方向發展。

The field of AI and data science is undoubtedly one of the most exciting fields today. We have collectively just about started scratching the surface on what’s possible — the horizons are limitless. The impact that the proliferation of AI will have on the world, is expected to be even more dramatic than that of the industrial revolution.

人工智慧和資料科學領域無疑是當今最令人興奮的領域之一。 我們已經集體開始在可能的範圍內開始摸索—視野是無限的。 預計AI的擴散將對世界產生比工業革命更大的影響。

Therefore, anyone who gets into this domain early enough, will have tremendous opportunities as the field grows even further. Make a wise choice, make an informed choice. Don’t get sucked into a pipe-dream when countless real and exciting opportunities with startups exist across the globe.

因此，隨著該領域的進一步發展，任何人儘早進入這一領域將擁有巨大的機會。 做出明智的選擇，做出明智的選擇。 當全球範圍內的初創企業存在無數真實而激動人心的機會時，請不要陷入夢pipe以求的事情。