Huma Abidi, Senior Director of AI Software Products at Intel – Interview Series

Huma Abidi is a Senior Director of AI Software Products at Intel, answerable for technique, roadmaps, necessities, machine learning and analytics software program merchandise. She leads a globally numerous crew of engineers and technologists answerable for delivering world-class merchandise that allow clients to create AI options. Huma joined Intel as a software program engineer and has since labored in a wide range of engineering, validation and administration roles within the space of compilers, binary translation, and AI and deep learning. She is obsessed with ladies’s training, supporting a number of organizations world wide for this trigger, and was a finalist for VentureBeat’s 2019 Women in AI award within the mentorship class.

What initially sparked your curiosity in AI?

I’ve at all times discovered it fascinating to think about what might occur if machines might communicate, or see, or work together intelligently with people. Because of some huge technical breakthroughs within the final decade, together with deep studying gaining recognition due to the provision of knowledge, compute energy, and algorithms, AI has now moved from science fiction to actual world functions. Solutions we had imagined beforehand at the moment are inside attain. It is really an thrilling time!

In my earlier job, I used to be main a Binary Translation engineering crew, targeted on optimizing software program for Intel {hardware} platforms. At Intel, we acknowledged that the developments in AI would result in enormous trade transformations, demanding super progress in compute energy from units to Edge to cloud and we sharpened our focus to grow to be a data-centric firm.

Realizing the necessity for highly effective software program to make AI a actuality, the primary problem I took on was to guide the crew in creating AI software program to run effectively on Intel Xeon CPUs by optimizing deep studying frameworks like Caffe and TensorFlow. We had been capable of reveal greater than 200-fold efficiency will increase as a consequence of a mixture of Intel {hardware} and software program improvements.

We are working to make all of our buyer workloads in varied domains run sooner and higher on Intel know-how.

 

What can we do as a society to draw ladies to AI?

It’s a precedence for me and for Intel to get extra ladies in STEM and pc science basically, as a result of numerous teams will construct higher merchandise for a various inhabitants. It’s particularly necessary to get extra ladies and underrepresented minorities in AI, due to potential biases lack of illustration may cause when creating AI options.

In order to draw ladies, we have to do a greater job explaining to women and younger ladies how AI is related on the earth, and the way they are often a part of creating thrilling and impactful options. We want to point out them that AI spans so many various areas of life, and so they can use AI know-how of their area of curiosity, whether or not it’s artwork or robotics or knowledge journalism or tv. Exciting functions of AI they’ll simply see making an affect e.g. digital assistants like Alexa, self-driving vehicles, social media, how Netflix is aware of which motion pictures they wish to watch, and many others.

Another key a part of attracting ladies is illustration. Fortunately, there are numerous ladies leaders in AI who can function glorious position fashions, together with Fei-Fei Li, who’s main human-centered AI at Stanford, and Meredith Whittaker, who’s engaged on social implications via the AI Now Institute at NYU.

We have to work collectively to undertake inclusive enterprise practices and increase entry of know-how abilities to ladies and underrepresented minorities. At Intel, our 2030 purpose is to extend ladies in technical roles to 40% and we will solely obtain that by working with different corporations, institutes, and communities.

 

How can ladies greatest break into the trade?  

There are just a few choices if you wish to break into AI particularly. There are quite a few on-line programs in AI, together with UDACITY’s free Intel Edge AI Fundamentals course. Or you might return to high school, for instance at considered one of Maricopa County’s community colleges for an AI associate degree – and research for a profession in AI e.g. Data Scientist, Data Engineer, ML/DL developer, SW Engineer and many others.

If you already work at a tech firm, there are possible already AI groups. You might take a look at the choice to spend a part of your time on an AI crew that you simply’re interested by.

You may work on AI for those who don’t work at a tech firm. AI is extraordinarily interdisciplinary, so you’ll be able to apply AI to nearly any area you’re concerned in. As AI frameworks and instruments evolve and grow to be extra user-friendly, it turns into simpler to make use of AI in numerous settings. Joining on-line occasions like Kaggle competitions is a good way to work on real-world machine studying issues that contain knowledge units you discover fascinating.

The tech trade additionally must put in time, effort, and cash to achieve out to and assist ladies, together with ladies who’re additionally underrepresented ethnic minorities. On a private observe, I’m concerned in organizations like Girls Who Code and Girl Geek X, which join and encourage younger ladies.

 

With Deep studying and reinforcement learning just lately gaining probably the most traction, what different types of machine studying ought to ladies take note of?

AI and machine studying are nonetheless evolving, and thrilling new analysis papers are being revealed often. Some areas to give attention to proper now embrace:

  1. Classical ML strategies that proceed to be necessary and are broadly used.
  2. Responsible/Explainable AI, that has grow to be a important a part of AI lifecycle and from a deployability of deep studying and reinforcement studying point-of-view.
  3. Graph Neural Networks and multi-modal studying that get insights by studying from wealthy relation data amongst graph knowledge.

 

AI bias is a large societal problem with regards to bias in direction of ladies and minorities. What are some methods of fixing these points?

When it involves AI, biases in coaching samples, human labelers and groups might be compounded to discriminate in opposition to numerous people, with severe penalties.

It is important that range is prioritized at each step of the method. If ladies and different minorities from the neighborhood are a part of the groups creating these instruments, they are going to be extra conscious of what can go fallacious.

It can be necessary to ensure to incorporate leaders throughout a number of disciplines corresponding to social scientists, medical doctors, philosophers and human rights specialists to assist outline what is moral and what’s not.

 

Can you clarify the AI blackbox downside, and why AI explainability is necessary?

In AI, fashions are educated on large quantities of knowledge earlier than they make selections. In most AI techniques, we don’t know the way these selections had been made — the decision-making course of is a black field, even to its creators. And it is probably not attainable to essentially perceive how a educated AI program is arriving at its particular choice. An issue arises once we suspect that the system isn’t working. If we suspect the system of algorithmic biases, it’s tough to test and proper for them if the system is unable to clarify its choice making.

There is at the moment a serious analysis give attention to eXplainable AI (XAI) that intends to equip AI fashions with transparency, explainability and accountability, which can hopefully result in Responsible AI.

 

In your keynote address during MITEF Arab Startup Competition closing award ceremony and convention you mentioned Intel’s AI for Social Good initiatives. Which of those Social Good tasks has caught your consideration and why is it so necessary?

I proceed to be very enthusiastic about all of Intel’s AI for Social Good initiatives, as a result of breakthroughs in AI can result in transformative adjustments in the best way we sort out downside fixing.

One that I particularly care about is the Wheelie, an AI-powered wheelchair inbuilt partnership with HOOBOX Robotics. The Wheelie permits excessive paraplegics to regain mobility by utilizing facial expressions to drive. Another superb initiative is TrailGuard AI, which makes use of Intel AI know-how to combat unlawful poaching and shield animals from extinction and species loss.

As a part of Intel’s Pandemic Response Initiative, we’ve many on-going tasks with our companions utilizing AI. One key initiative is contactless fever detection or COVID-19 detection via chest radiography with Darwin AI. We’re additionally engaged on bots that may reply queries to improve consciousness utilizing natural language processing in regional languages.

 

For ladies who’re interested by getting concerned, are there books, web sites, or different assets that you’d advocate?  

There are many nice assets on-line, for all expertise ranges and areas of curiosity. Coursera and Udacity supply glorious on-line programs on machine studying and seep studying, most of which might be audited totally free. MIT’s OpenCourseWare is one other nice, free technique to study from among the world’s greatest professors.

Companies corresponding to Intel have AI portals that comprise loads of details about AI together with supplied options. There are many nice books on AI: foundational pc science texts like Artificial Intelligence: A Modern Approach by Peter Norvig and Stuart Russell, and fashionable, philosophical books like Homo Deus by historian Yuval Hararri. I’d additionally advocate Lex Fridman’s AI podcast on nice conversations from a variety of views and specialists from completely different fields.

 

Do you’ve gotten any final phrases for ladies who’re inquisitive about AI however will not be but able to leap in?

AI is the long run, and can change our society — in truth, it already has. It’s important that we’ve trustworthy, moral individuals engaged on it. Whether in a technical position, or at a broader social degree, now is an ideal time to get entangled!

Thank you for the interview, you’re actually an inspiration for ladies the world over. Readers who want to study extra in regards to the software program options at Intel ought to go to AI Software Products at Intel.

LEAVE A REPLY

Please enter your comment!
Please enter your name here