AI's 'long tail' prevents mature adoption, says Andrew Ng

AI’s ‘long tail’ prevents mature adoption, says Andrew Ng

Andre Ng

Andrew Ng is one of the biggest names in artificial intelligence and machine learning, having founded and led a team at Google Brain, Baidu and elsewhere, and as founder of Coursera and Landing AI. Her online courses have attracted millions of views.

Artificial intelligence has huge potential outside of consumer software and internet applications, he believes. But as he walks around user premises talking to project managers about their work, he discovers an industry that is in desperate need of guidance and leadership:

I think the greatest potential of AI is still ahead of us, to use it for all other industries other than consumer software and the internet. Everything from retail, travel, transportation and logistics, automotive and assembly, and more.

But frankly, when I walk around everywhere, from factories to hospitals, they’re just looking for mentors. I think the adoption of AI in all of these industries is still very nascent. A McKinsey study estimated the annual value at $13 trillion by 2030. But I think there’s still a lot of work ahead of us to create that value. AI cannot reach its full potential until it is accessible to everyone.

So why isn’t AI being more widely or maturely adopted in many industries? According to Ng, this is a long tail or customization issue. He explains:

Here’s what I mean. If you want to take all potential and actual AI projects and sort them in descending order of value, you get a curve. We are on the left and perhaps the most valuable trading AI system in the world is an online advertising system, and perhaps the second is a web search engine that shows the most relevant results. And then maybe with self-driving vehicles, we will get there one day in the future.

So in the world of AI, we figured out how to hire dozens or hundreds of machine learning engineers to build a giant monolithic system to serve millions, hundreds of millions, even billions of users. But once we go into other industries, I see a lot of projects that are worth maybe $1 million to $5 million each, whether it’s a pizza chain wanting to do full demand forecasting , a t-shirt manufacturer wanting to improve product placements, or how to do better quality control in automotive manufacturing.

[The problem is that] Anywhere outside of consumer software and the internet, we don’t have these databases of 100 million users to apply a single AI system to. Instead, I see tens of thousands of projects that aren’t running efficiently right now, due to high customization costs. I can’t hire 10,000 machine learning engineers to build 10,000 of these projects. »


So what is the solution ? Ng supports:

Fortunately, an emerging technology called data-centric AI now makes many of these projects possible. When you think about what a team needs to do today to build an AI system, they have to write a lot of code. And while I hope everyone will learn to code, realistically it’s hard to get everyone to make good cutting-edge AI software on their own.

Most people get a dataset somewhere and then have a team write code and focus on improving the software, but it turns out to be difficult. But with the data-centric approach to AI, we’ve turned that recipe on its head. We have observed that for many AI applications, the code is already a solved problem. There might be an open source implementation of an AI model that you can get from a vendor that works just fine. Instead, it’s more fruitful to provide the tools for your teams to work with data. »

Using the example of an in-vehicle AI system to identify if a driver or passenger has left their wallet/purse on the car seat, Ng demonstrates the challenge of teaching a computer vision system which pixel layout means “portfolio” when different designs are viewed from multiple angles. Similar challenges are faced by computer scientists teaching robots to pick up tools, for example, or driverless cars to recognize pedestrians.

The problem is not the hardware or the coding of the application, but the quality and consistency of data labeling, he says:

With data-centric AI, you can increase access to AI for teams because rather than needing the expertise to write the code, you really need the expertise to know what’s going on. is a wallet in a car. What you’re really trying to do is provide and describe data in a very consistent way. So the key part of this journey to democratize access to AI is not just to let the 100 million or billion dollar systems get built, but to let all of those 1 to 5 billion dollar projects in the tail build.

With data-centric AI, it’s about providing training for more people, but for subject matter experts rather than machine learning engineers. This is what my team is doing and what many others are trying to do in different application areas. Hopefully these will provide a foundation with which we can give many more people access and many more people the ability to create custom AI systems.

my catch

An obvious point, perhaps, but important. With so many sophisticated AI tools, the real challenge is accurate and consistent labeling in industry-specific domains, even in relatively small datasets (as opposed to those with millions of items).

Hopefully Ng’s profile will ensure that data-centric tools are more widely adopted, rather than business leaders spending millions of dollars recreating the wheel when more data is needed on the wheel itself.

#AIs #long #tail #prevents #mature #adoption #Andrew

Leave a Comment

Your email address will not be published. Required fields are marked *