This blog post was originally published at Intel’s website. It is reprinted here with the permission of Intel.
When it comes to artificial intelligence (AI), few people can match Andrew Ng’s breadth and depth of experience, combining research in a world-leading academic environment, accelerating AI education (one of his courses really helped me when I was breaking into AI myself) with founding AI teams at some of the most successful tech companies. He isn’t predicting the future so much as building it.
Ng lead the Stanford AI Lab, was a founding leader at Google Brain where he worked with legendary engineer Jeff Dean, lead AI at Baidu, and currently serves as an adjunct professor in Computer Science at Stanford University. Among Andrew’s other pursuits: being the founder of deeplearning.ai, the founder and CEO of Landing AI, a general partner at AI Fund, and chairman and co-founder of Coursera.
In a recent episode of the Intel on AI podcast, Ng and host Intel’s Abigail Hing Wen discuss why most of the important work yet to be done with AI is in industries outside of Silicon Valley, such as manufacturing, agriculture, and healthcare. Ng sees AI as driving huge growth for successful adopters, to the point where he sees a need for society to be prepared to offer additional support to workers in disrupted industries.
From the business perspective, AI has become an all-encompassing description for both machine learning and deep learning. I’ve written before about why reducing complexity for inferencing at scale is so important to companies now. In the podcast, Ng notes that people use AI systems every day, multiple times a day, but—in general—we don’t think of those systems as AI; they simply work in the background to solve business issues. For example, AI’s first use-case at Google was to auto-correct the spelling of search terms—not as thrilling as self-driving cars, but with immediate benefits for search quality.
One saying that some people in AI have is “Once it works, it’s no longer AI.”
-Andrew Ng
AI in Manufacturing & Agriculture
Ng has a lot to say about the benefits of AI for manufacturing, and how the transition to AI has been accelerated by the need for remote working brought on by the pandemic: like other industries, the pandemic has greatly accelerated digitization.
At Intel, we’ve seen significant benefits from applying AI to our own manufacturing: last year Intel’s VP of IT Allyson Crafton outlined how AI had saved Intel an estimated 58 million dollars associated with inventory optimization and increased warehouse planning accuracy by 95 percent. Those are some remarkable numbers, even for a company at the scale of Intel, and with open-source tools like Intel Distribution of the OpenVINO toolkit, we expect companies across essentially all industries to harness the power of AI and computer vision. For example, Rosmart found production lines could reduce costs and increase accuracy and LEDA technology found it could inspect contact lens 50x faster than before.
In agriculture, AI is helping farming companies, large and small, by using a combination of connected sensors, drones, satellite imagery, and other applications to harvest more efficiently, decrease waste, and reduce the use of chemical agents.
There’s an association in many minds between “Big Data” and AI, and it is true that large amounts of training data can be hugely valuable for creating new models, and that some datasets are so large that mechanizing analysis with AI is essential. Yet one of the most important aspects of modern AI is that it lets us do much more with small data. In the podcast, Ng brings up the example of smartphone manufacturing and detecting flaws. No manufacturer is producing millions of damaged products from which they could use images to train AI systems to detect. Instead, the AI system would need to learn from only hundreds of flawed products in order to detect future defects. Use-cases like these are already addressable today because AI can take what it learns from large datasets and use “transfer learning” to bring this knowledge to small datasets. This is established best practice for machine vision, and similar approaches are now bearing fruit for natural language processing, audio and even robotics. Dealing with “small data” (very small training sets) remains a hot AI research topic, but for many use-cases, we can already draw on established machine learning engineering best practices.
AI in Health Care
As Andrew Ng points out in the interview, we can expect the quickest successes for AI in industries that are the furthest advanced with digitization; as we see it, AI’s ability to improve health care is one of the most promising application areas. Intel is working with several partners in the field to improve diagnostic and clinical practice, ranging from accelerating times to quantify heart functions to embedding AI-systems into X-Ray machines for better pneumothorax detection to assisting with precision pulmonary surgery. At the personal level, Intel’s Bryce Olson is doing extraordinary things in his fight against cancer with genomic sequencing. One way AI is literally accelerating research in this area: we’ve worked with the Broad Institute on AI-based components of their Genomic Analysis Toolkit, which uses deep learning inference on Intel CPUs to enable faster and more accurate sequencing for researchers world-wide.
Returning to the podcast, Andrew and Abigail have a great discussion around AI and ethics, and how AI can learn from healthcare, where practitioners often have to make difficult choices on how to apply their principles in their practice. This segment really resonated for me personally. My father was a doctor whose ethics strongly informed his career path, bringing patient care to very poor communities in troubled countries.
The Future of AI
One of the points Ng consistently brings up in the podcast is just how much room for expansion AI has, how we’re still only in the earliest stages of what will become a massive transformation of technology, business, and life. I’m very excited to be working in AI at a time when the number of “low-hanging-fruit” use-cases is growing faster than we can address them.
To learn more about Intel’s work in AI, visit: https://intel.com/ai
To hear more podcasts about AI, look for future episodes with host Abigail Hing Wen at: https://www.intel.com/content/www/us/en/artificial-intelligence/podcast.html
Edward Dixon
Data Scientist, Intel
