What I Learned as a Data Science Researcher turned AI Leader in a Year

栏目: IT技术 · 发布时间: 4年前

内容简介:Last week, I was recognized and nominated asHere’s what I learned as a newcomer, leader in AI.As a newcomer to the industry — fresh out of university and a year in the industry, I’ve been able to shift and mold very rapidly. Not long ago, I was just a Data

Last week, I was recognized and nominated as VentureBeat’s AI Rising Star , one of the AI Leadership Awards awarded annually:

This award will honor someone in the beginning stages of her AI career who has demonstrated exemplary leadership traits.

Here’s what I learned as a newcomer, leader in AI.

As a newcomer to the industry — fresh out of university and a year in the industry, I’ve been able to shift and mold very rapidly. Not long ago, I was just a Data Science Researcher in my own little lab bench corner of a Particle Accelerator Research Lab while pursuing my undergraduate studies. I shouldn’t say “just”, because data science is a lot of hard work! Little did I know that a year later, I’ll be recognized in the industry as a Leader.

I went from 0–60 mph in under two seconds!

What I Learned as a Data Science Researcher turned AI Leader in a Year

Photo by Dylan Calluy on Unsplash

Most importantly, I’ve also been able to grasp and observe the industry with a newcomer perspective. Here’s what I learned in a year of going from a data science researcher to an industry leader.

AI requires a completely different perspective

Most AI leaders today are “ transcended data scientists ”. They typically have pursued formal training in science, engineering or maths, and then woke up one day and decided they were more interested in leading people.

In all too common situations they find themselves in, the double mastery of technical and leadership skills was earned in series not in conjunction with each other.

They usually fall in one of two buckets:

  1. A leader who starts to lose the technical side of things and can only talk about products at a high-level
  2. A leader who is so entrenched in the technology and usually gets too into the weeds of the development

And typically, AI companies today fall in one of two buckets:

  1. A company that pursues abstraction of technology and focuses on the real-world problem at hand (i.e., domain experts)
  2. A company that gets too in the weeds of technology and loses sight of the real-world problem they were trying to solve in the first place (i.e., technology providers)
The industry requires all of you — the technical and visionary leader, and the abstracted and technology-focused organizations. But, there is a silo problem.

The industry is silo-ed. In the ecosystem, there are organizations that develop the underlying, necessary technologies, like frameworks, pre-trained models, and libraries. On the other hand, there are organizations that are specialized in certain domains — they know specific industries really well and understand the underlying problems the industry faces best. But these organizations don’t typically partner closely with each other. They build technologies independent of one another.

Within those organizations, there are AI leaders that are either too abstracted or too entrenched in the technology. There are leaders that only know how to spell AI, and leaders that understand the maths behind it all. They are typically not the same person.

What is needed in order to optimize AI technologies and deliver to its promise are leaders that can be both abstracted and entrenched in the technology, and organizations that work closely with each other and in harmony with one another.

But it is not an easy feat. If it was, then we would be doing it already, right?

There is anattitude problem. Do we really value these intersections and the inter-weaving of skills and organizations?

The most valuable leadership trait required in AI

In the age of information, ironically, leaders are typically shielded from new information, because of the sheer volume of information and the rate at which new information is generated every day.

On July 16, 2020 alone, there were about 70 papers submitted to arXiv.org that have to do with machine learning with each paper ranging from 15–50 pages of information. That doesn’t even include other research publishers. That is a lot of new information and knowledge generated every day.

The most necessary leadership trait required of an AI leader today is humility.

In the sea of rapid change and everyday shifts in the industry, it is impossible to purely rely on 20-year-old wisdom of how the industry works. AI leaders need to be willing to learn and be open to learning from someone who 20 years younger or multiple levels of the hierarchy down. Organizations need to empower more Rising Stars .

The most effective leaders I’ve come to follow are those that are unafraid to learn from interns and direct reports.

Companies typically hire newcomers to lead as product managers or project leads, because the situation warrants an outside-in perspective and a fountain of knowledge full of new and yet-to-be primed ideas. Organizations require things to be shaken up and rapidly adapt to new information every day.

AI needs to be democratized

Gone are the days when AI was only accessible to PhD researchers and organizations that could afford a research residency program. For the industry to realize the full potential of AI, there needs to be democratization. Making the technology accessible expands the possibilities and the realm of what we can do with it.

More than any other industry, AI requires a large community to collaborate and datasets, insights, and models that can be shared among each other, which is why most industry leaders today open-source code. This creates and engenders trust.

Not only do datasets and models need to be democratized, knowledge and know-how also need to be accessible, especially to industry experts who best understand the real-world problem.

AI is not a black box nor is it magic. Simply put, it is a construct that allows us to generate instructions out of examples, as opposed to relying on programmers to outline step-by-step instructions.

Once more of the industry understands this, the more we could better understand the use cases where AI is needed and when it is not (and yes! Some use cases do not require AI — sometimes, it only requires data analytics and statistics) . The more industry leaders understand this, the better we could decipher the possibilities where AI can enhance existing solutions we have today and create new solutions to unsolved issues, and the less we get frustrated at the data scientists in our teams who spend full-day business hours reading papers on arXiv.

To summarize, leaders in AI need to be open-minded, have humility, and work to democratizing AI. AI needs to be an accessible technology that is available to everyone who sees and wants to solve problems.

Successful real-world implementation of AI needs a new breed of decision-makers that could adapt to the ever-so-changing industry.

This is not to say we only need young, newcomer leaders in the industry. I have a minuscule wisdom dataset compared to someone who has been in the industry for more than 20 years! It is a “yes, and.” We need both leaders that have acquired a foundation of wisdom (after all, there is no compression framework to experience!) and new leaders who could shake up the industry.

It is no easy feat to change the world, there is more to be gained if we learn from each other.

以上所述就是小编给大家介绍的《What I Learned as a Data Science Researcher turned AI Leader in a Year》,希望对大家有所帮助,如果大家有任何疑问请给我留言,小编会及时回复大家的。在此也非常感谢大家对 码农网 的支持!

查看所有标签

猜你喜欢:

本站部分资源来源于网络,本站转载出于传递更多信息之目的,版权归原作者或者来源机构所有,如转载稿涉及版权问题,请联系我们

格蠹汇编

格蠹汇编

张银奎 / 电子工业出版社 / 2013-3-1 / 66.00元

《格蠹汇编——软件调试案例集锦》以案例形式讨论了使用调试技术解决复杂软件问题的工具和方法。全书共36章,分为四篇。前两篇每章讲述一个有代表性的真实案例,包括从堆里抢救丢失的博客,修复因误杀而瘫痪的系统,徒手战木马,拯救“发疯”的windows7,经典阅读器的经典死锁,拯救挂死的powerpoint,转储分析之双误谜团,是谁动了我的句柄,寻找系统中的“耗电大王”,解救即将被断网的系统,转储分析之系统......一起来看看 《格蠹汇编》 这本书的介绍吧!

MD5 加密
MD5 加密

MD5 加密工具

Markdown 在线编辑器
Markdown 在线编辑器

Markdown 在线编辑器

正则表达式在线测试
正则表达式在线测试

正则表达式在线测试