意见 (Opinion)

Disclaimer: This is an opinion piece. I’d love to hear your thoughts and counter arguments in the comments.

免责声明：这是一个观点。 我很想听听您的想法，并在评论中提出反对意见。

There’s a lot of doom and gloom in the field.

这个领域充满了厄运和忧郁。

Hiring is frozen.

招聘被冻结。

…some hypothesize that investors will lose hope in AI altogether. Google has freezed hiring for ML researchers. Uber laid off the research half of their AI team…there will be far more people with ML skills than ML jobs.- Chip Huyen

…一些假设，投资者将完全对人工智能失去希望。 Google已冻结ML研究人员的招聘。 Uber裁员了一半的AI团队...拥有ML技能的人将比ML工作多得多。- Chip Huyen

We have a recession.

我们有经济衰退。

People are talking about an AI Winter.

人们正在谈论 AI冬季 。

It makes sense that artificial intelligence (AI), machine learning (ML) and data science (DS) are the first to go in a crunch. They’re luxuries for most businesses.

人工智能(AI)，机器学习(ML)和数据科学(DS)最先出现危机是有道理的。 它们是大多数企业的奢侈品。

But that doesn’t mean the future isn’t bright.

但这并不意味着未来并不光明。

If you create value.

如果您创造价值。

AI Winter不会影响大多数AI / ML / DS工作 (AI Winter will not affect most AI/ML/DS jobs)

An AI winter is a period of decreased funding and interest in AI research.

人工智能冬季是一个减少资金投入和对人工智能研究兴趣的时期。

But most of us don’t do research. We read papers, get ideas and innovate... but we use existing techniques.

但是我们大多数人都不做研究。 我们阅读论文，获得想法并进行创新……但是我们使用现有技术。

Additionally, the popularity of building ML-powered products doesn’t necessarily correlate with the volume of research coming out.

此外，构建基于ML的产品的普及并不一定与即将开展的研究数量相关。

If anything, there’s an increasing amount of research that’s not being applied. Anecdotally, industry is still catching up in its implementation of machine learning invented decades ago.

如果有的话，还有越来越多的未应用的研究。 有趣的是，行业仍在追赶数十年前发明的机器学习的实施。

“AI-powered” products are more popular now because ML is more approachable, not because of new research.

现在，“人工智能驱动”产品变得更加流行，因为机器学习更加容易实现，而不是因为新的研究。

您不需要前沿AI即可解决问题 (You do not need bleeding edge AI to solve problems)

The opposite is true.

反之亦然。

Classic algorithms + domain knowledge + niche datasets are going to solve most real problems, not deep neural nets. Most of us aren’t working on self-driving cars.

经典算法+领域知识+生态位数据集将解决大多数实际问题，而不是深度神经网络。 我们大多数人都不会在自动驾驶汽车上工作。

I wrote about this in “Democratizing AI is irrelevant, Data is siloed, And how to build an AI company anyway”.

我在“ 民主化AI无关紧要，数据孤岛，以及无论如何如何建立AI公司 ”中 写道 。

In my opinion, focusing on extreme technical competency is overrated outside large tech companies, in contrast to a problem-solving mentality and general development skills.

我认为，与大型企业相比，专注于极端技术能力的问题被高估了，这与解决问题的心态和一般的开发技能形成了鲜明的对比。

Outside of tech, there’s still a tonne of boring/manual work that should have been automated a long time ago. And it doesn’t require breakthroughs.

在技​​术之外，仍然有很多无聊的/手工工作应该在很长时间之前就已经自动化了。 而且不需要突破。

使用机器学习，但专注于在改变世界中创造价值 (Use machine learning, but focus on creating value over changing the world)

When you solve a problem (any problem), everyone wins.

当您解决问题(任何问题)时，每个人都会赢。

Silicon Valley has deluded us to believe we should be taking moonshots rather than improving our local communities and the lives of people we know.

硅谷使我们相信，我们应该大开眼界，而不是改善我们的当地社区和我们认识的人们的生活。

I love Uber, and it has changed the world. But if keeping Uber alive costs $5 billion per quarter, maybe something is wrong.

我爱Uber，它改变了世界。 但是，如果维持优步每季度的生存成本为50亿美元 ，那么也许有些问题。

Yes, some companies are long term plays and will affect 7 billion people. But simpler improvements like reducing data entry mistakes in a “boring” industry also creates value.

是的，有些公司是长期公司，将影响70亿人。 但是，减少“无聊”行业中的数据输入错误等更简单的改进也可以创造价值。

学习ML是对抗AI恐惧的最佳方法 (Learning ML Is the best way to fight fear of AI)

We hear of automation killing jobs, because nothing sells like fear. Not because technological unemployment is around the corner.

我们听说自动化会杀死工作 ，因为没有什么能卖得像恐惧。 不是因为技术失业迫在眉睫。

Pick up machine learning. Then try conceptualizing, training and deploying a model to solve a real problem. You’ll quickly see how hard this still is, and secondly, how far we must be away from an AGI takeover.

掌握机器学习。 然后尝试概念化，训练和部署模型以解决实际问题。 您将很快看到这仍然很困难，其次，我们离AGI收购还有多远。

Infrastructure is super underdeveloped and real data is messy.

基础设施超级落后，真实数据混乱。

When you download a CSV from Kaggle to train a model for a specific problem, 99% of the work has been done for you.

当您从Kaggle下载CSV来为特定问题训练模型时，已经为您完成了99％的工作。

If more people did this, they’d sleep better at night.

如果更多的人这样做，他们晚上会睡得更好。

使ML容易的工具之间存在差距 (There is a gap in tools to make ML easy)

Ease-of-use has done more for ML adoption than any algorithm breakthroughs over the last 10 years.

在过去10年中，易用性为ML的采用做出了比任何算法突破都更多的事情。

We’re almost at the point where a software engineer can cobble together an ML solution using out-of-the-box components, but it’s not easy enough yet.

我们几乎可以说软件工程师可以使用现成的组件将ML解决方案拼凑在一起，但这还不容易。

As tools progress, we’ll see less pure ML jobs but a huge increase in software engineers using ML to solve all sorts of problems. And more companies outside of “tech” benefiting.

随着工具的进步，我们将看到纯净的ML工作减少，但是使用ML解决各种问题的软件工程师的数量将大大增加。 除了“技术”之外，还有更多公司从中受益。

If you become an engineer who builds great tools for ML, I’d forever be in your debt….- Chip Huyen

如果您成为为ML开发出色工具的工程师，我将永远背负着您的重担....-Chip Huyen

ML is driving value around the world, but I think we’ve barely scratched the surface. Wait until the right tools are here.

机器学习正在世界范围内推动价值增长，但我认为我们几乎还没有触及表面。 等到正确的工具在这里。

首先进行软件工程 (Do software engineering first)

Unless you have an advanced degree in an AI related subject, do yourself a favour and learn software engineering. Then move into AI.

除非您具有AI相关主题的高级学位，否则请帮自己一个忙，并学习软件工程 。 然后进入AI。

Learning software engineering is like getting an MBA in technology (when MBAs were valuable). You’ll learn fundamentals, create full stack solutions, and understand the code that facilitates ML.

学习软件工程就像获得技术MBA一样(当MBA很有价值时)。 您将学习基础知识，创建完整的堆栈解决方案，并了解有助于ML的代码。

There are also more jobs and it will be easier to change careers as the industry landscape shifts.

还有更多的工作机会，随着行业格局的变化，更容易改变职业。

Many software engineers go on to successful careers in ML/DS. But you rarely see the opposite.

许多软件工程师继续在ML / DS中取得成功。 但是您很少看到相反的情况。

结论 (Conclusion)

There is mega hype around AI. And with any rise comes a “fall”, but that doesn’t have to be a bad thing if we’re prepared.

围绕AI进行了大肆宣传。 随之而来的是“下跌”，但是如果我们准备好了，那不一定是一件坏事。

If we focus on developing a general skillset (including ML), solving real problems, and creating value, there will always be something for us to do.

如果我们专注于开发通用技能(包括ML)，解决实际问题并创造价值，那么总会有事情要做。