What does the creation of new artificial intelligence products look like today, and what do experts in this field foresee realistically happening in the near future? One thing’s for sure, the way we work and function in life will change as a result of growth in this field. Listen and find out more.

Below is a partial transcript. For the full interview, listen to the podcast episode by selecting the Play button above or by selecting this link or you can also listen to the podcast through Apple Podcasts, Google Play, Stitcher, and Overcast.

Transcript

Irmak Sirer: “It’s kind of like a Where’s Waldo of finding an expert in this entire giant ocean of people.”  

Ginette: “I’m Ginette.”

Curtis: And I’m Curtis.”

Ginette: “And you are listening to Data Crunch.”

Curtis: A podcast about how data and prediction shape our world.”

Ginette: A Vault Analytics production.”

Curtis: “Brought to you by data.world, the social network for data people. Discover and share cool data, connect with interesting people, and work together to solve problems faster at data.world. A complex dataset with a ton of files can quickly become scary and unwieldy, but you need not fear! Now you can use file labels and descriptions to manage and organize your many files on data.world. With file labels and descriptions, you can quickly see what type of file it is, view a short description, and also filter down by file type. Wanna see an example of how data.world users are using file labels and descriptions to keep their dataset organized? Search “data4democracy/drug-spending” on data.world.

Ginette: “Today we’re taking a closer look at something that is starting to seep into our daily lives. In one of its forms, it’s something Stephen Hawking, Bill Gates, and Elon Musk are concerned will eventually be a threat to mankind. In another form, though, you’re probably already using it, and it’s becoming a major game changer, kind of like the early days of the desktop computer. We’re talking about artificial intelligence. You use AI when you talk to Siri or your in-home assistant, Alexa or Echo, and some people are using it in the form of a self-driving car.

“So daily applications of artificial intelligence are on the rise, becoming much more of a staple in our society, but AI’s definition shifts according to the source. Popular movies depict AI as having a consciousness, emotions, and exhibiting human-like characteristics. Usually it’s involved in some sort of world-domination plot to kill all the humans. Although most experts agree that artificial intelligence will never actually think and feel like a human, the existential threat still exists. This kind of apocalyptic AI is known as ‘general AI.’ But that’s a topic for another episode. Today, we’re focusing on the kind of AI that currently exists, otherwise known as narrow AI.”

Curtis: “A narrow AI is called narrow because it’s usually focused on one specific task, where as a general AI would be able to be good pretty much any task thrown its way. The Google search bar is probably the most ubiquitous example of a narrow AI that most people use on a daily basis. The process usually goes like this: you give it an input like ‘How to own a llama as a pet.’ It does its processing. It gives you an output in the form of the 10 most relevant web pages to answer your questions (along, of course, with some paid advertisers who are trying to sell you a pet llama).

“The simplicity of the interaction belies the complexity of the cognitive work that’s going on behind the scenes. Imagine if you had to do the same cognitive task without the help of Google. What would that actually entail? You would have to individually look at and read every single website, and there are over 1 billion, and peruse them to see if they have anything to do about llamas, not to mention then find the individual pages on those websites that actually answer your questions.

“This is a really big task! It’s basically intractable for a human to accomplish because you just don’t have the cognitive abilities to do it. If it took you three seconds on average to scan a website and see if it’s relevant, then it would take you 10 and a half years to do the same task, supposing you didn’t eat, sleep, or do anything else for those 10 years. In context of an eight-hour work day, it would take the entire career of the average person to accomplish this task. Google just took the entire life’s work of someone and performed it in a fraction of a second to answer your question about a llama.”

Ginette: “We spoke with someone who works at Stanford’s computer science department to get her take on AI.”

Swati Dube-Batra: “My name is Swati Dube-Batra, and I’ve been working at the Stanford University for the last four years now, and here, I’ve had the privilege to work with one of the best machine learning teams in the world.”

Ginette: “Swati sees AI saving us lots of time in the future and gives examples of what it’s doing for us now.”

Swati: “Self-driving cars will replace a human having to drive a car, so that will free up hundreds of hours that we actually spend sitting behind the steering.”

AI is already transforming our present with technologies like the wearable devices that we wear on a day-to-day basis, like the Fitbit and conversations with chatbots and various business tools, which a lot of us are using on a day-to-day basis. This also means better productivity and not having to do a task which can be done by a machine.”

Curtis: “Narrow AI has the potential to drastically change how we live and how we work. For one thing, it could alleviate some cognitive load in our jobs. Let’s take a look into one business that developed an incredibly helpful, narrow AI tool to help solve a huge business headache for a multinational company. Building out an AI to help reduce our workload starts with finding a cognitive problem that needs a solution.”

Brian Lange: “My name is Brian Lange. I’m a partner and a data scientist at DataScope. We’re a data science consultancy based in Chicago, Illinois. What that really means is that we provide a number of different data-related services to our clients, ranging from just helping them figure out what they could or should be doing with data, sort of identifying opportunities, whether they be really innovative ways they could use it or ways they could improve their current processes up through testing the viability of some of those ideas, and ultimately implementing custom software to help them do that.

“So I guess if I’m going to tell the story of this one, it starts with a contact that we had with Procter and Gamble for a previous project. This person was tasked with sort of trying to help researchers and engineers and people within Procter and Gamble work together more effectively and more frequently.”

Ginette: “If you don’t know much about Procter and Gamble, know this: it’s a massive global company. While DataScope was working with P&G, it had operations in 70 countries and marketed its products in 180 countries. And their research and development arm? It had 20,000 researchers, and over 1,000 of their employees have PhDs.”

Brian: “Our client here, Kathy, would get a bunch of emails every day basically saying, ‘I’m wondering if you know somebody who knows something about blank—insert obscure technical topic here—and I need to know if there’s somebody with some expertise in this who can help me out.’ Because, you know, these people are researchers combining different technologies and trying to determine if they can come up with something that’s unique and innovative and can benefit Procter and Gamble in some way. So Kathy gets all these emails, and she responds to them, right?”

Curtis: “At first, she could handle the problem.”

Brian: “It started out that she had a small enough group of people she was doing this for that she was kind of a rockstar and the way that she did this was just by being a really amazing people person and keeping tabs on what people knew about and paying attention to what was going on and talking to people a lot and then sort of fostering those collaborations just using what was in her head.

“But as the size of her group that she was in charge of grew by orders of magnitude, this wouldn’t scale.”

Ginette: “This is where Dunbar’s number comes in. British anthropologist Robin Dunbar came up with a number of how many people we can maintain a stable social relationship with, and that number is around 150 people, so it makes sense that this Procter and Gamble employee could only keep working like this for so long as more and more people needed her to connect them with experts.

“For this project, Bryan and the DataScope crew built a tool that could do all this work for Kathy, significantly reducing her workload.”

Brian: “So it was basically an expertise search engine that would allow her to search people within the organization based on what the person emailing is asking about and find people that she might recommend.”

Curtis: “But how did they build this tool? Another Datascope worker, Irmak, explains:”

Irmak: “The problem we were facing was, how are we going to define expertise? Like when you say you are an expert, you don’t say you are an expert in specifically the title of the research report you wrote. That’s way too specific, and when people search for experts, they don’t have a very, very specific problem that a single research paper has already solved. They have problems like, ‘We are now going into perfumes, and I need to know about aerosols. Do we have any experts on aerosols?’ And nobody has written just a research paper on aerosols. They have just solved specific problems, but people that have worked on problems related to aerosols know a lot about it, so we would call them experts. So we had to somehow tell the computer to look at all these reports and when we have a question like, ‘Who’s an expert on aerosols?’ find out who that is.”

Curtis: “So how do you train a computer to find experts that aren’t labelled as experts?”

Above is a partial transcript. For the full interview, listen to the podcast episode by selecting the Play button above or by selecting this link or you can also listen to the podcast through Apple Podcasts, Google Play, Stitcher, and Overcast.

Some Sources

Music

“Night Owl” by Broke For Free is licensed under a Creative Commons Attribution License

Articles

http://www.mckinsey.com/global-themes/digital-disruption/harnessing-automation-for-a-future-that-works
http://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/where-machines-could-replace-humans-and-where-they-cant-yet
https://blogs.nvidia.com/blog/2016/07/29/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/
https://www.quora.com/What-are-the-main-differences-between-artificial-intelligence-and-machine-learning-Is-machine-learning-a-part-of-artificial-intelligence
http://camelsmouth.com/2015/07/28/artificial-intelligence-and-the-end-of-the-world/
https://medium.com/mmc-writes/the-fourth-industrial-revolution-a-primer-on-artificial-intelligence-ai-ff5e7fffcae1