In Events, AI

Building Smarter Products with Artificial Intelligence

By Monica Watson

Today, product professionals understand the pressing need for intuitive systems and experiences, and AI is proving to be the most efficient method of achieving this.

IMG_8220Recently, we invited NYC-based product professionals working in digital to the Clarifai offices to learn more about building products that understand with AI. Top AI leaders shared how AI is being used to solve real-world practical problems and how to build products that are intuitive and improve user experience.

Couldn't make it? Watch the full discussion below! We pulled out some of our favorite moments and quotes, which are under the video.

 

 

First, let's introduce our speakers! Raquel Ledezma-Haight is the Product Manager at x.ai, which makes scheduling meetings more efficient with AI. Amy Soyeon Kim is a Lead Product Manager at Clarifai (that's us!), and we enable computers to see - providing our customers with a computer vision engine to enable everything from moderation to retail solutions. Andrew Konya is the CEO at Remesh AI, which uses artificial intelligence to analyze, understand, and segment audience responses in real time. Swapnil Parikh is the Director of Product at HyperScience. HyperScience processes structured and semi-structured documents, helping organizations streamline complex processes. Lastly, Josh Goldenberg, Lead Product Manager at Clarifai, moderated the discussion.

 

When getting started with AI, what are questions people should consider? 

Amy: “The build vs buy discussion is one we have with almost every partner that comes through our door.” Businesses are really interested in starting with AI, but they don’t want to deal with all the elements it takes to build in house. So we view partnering with API platforms (like Clarifai) as pivotal. Learn more about the build vs buy debate here

 

IMG_8258

 

How are people using your tech to make a difference in the world?

Governments need to keep millions to billions of people content. However, understanding an entire populace's wants to reasoning has been pretty impossible to do effectively and without immense expense. Andrew spoke to one example where Remesh AI was able to make this possible for the Dutch government.

Andrew: We work with a lot of governments, and they have to make decisions on how to best spend their money in a way that's most useful to their citizens. The Dutch government [...] had noticed that their recycling numbers were going down, and they wanted to understand why. So they had a conversation with their citizens and learned that an increasing number of people were living in urban areas. They had designed the bins so they were one bin for each type of recycled item. However, all the urban apartments only had one place for a bin and not 3, so they learned that they had to develop one bin with dividers instead. 

Raquel spoke to how her company's AI was making collaboration and creation easier by taking away the pain and time of scheduling: 

Raquel: One of my favorite stories is this women who is a composer. She was putting together a piece of music from all around the world. She used x.ai to schedule calls with 50 people in the span of a couple of weeks. Our tools let her focus on the music instead of the back and forth.

Paperwork is often a huge headache both for the governments that need to read and manage all of it and for the people having to fill it out and wait on responses. More than being annoying, however, paperwork can even be detrimental to a person's health or livelihood if an agency is unable to sort through it efficiently enough. Swapnil spoke to how HyperScience has helped with one such project: 

Swapnil: Our technology can de-dupe documents. So there’s a government agency that processes disability payments in the US. This agency receives something in the order of 250 million documents a year, and they have a two year backlog to actually pay the disability because the documents need to be reviewed. The rule in the US is that you have to resubmit your medical records to this agency every six months, but there's a two year backlog. So every six months they get the same documents twice. And then a judge has to manually review every one of these documents. Greatly reducing the duplication means that judges are able to get through the troves of documents faster. 

And lastly, Amy shared how Clarifai's computer vision was used to augment doctors needing to treat and diagnose many people.

Amy: Doctors Without Borders have a lot of doctors that work in countries or areas without any internet access and have to limit the tools they can carry. They built a model to successfully diagnose ear infections. So they had just their phone and a tiny device to go inside the ear and get the image. These doctors have to look at hundreds of people, so they were able to diagnose better and more efficiently. 

 IMG_8275 

What is the question or sentiment you’re tired of hearing about AI?

Raquel: Is AI going to take over my job? The answer: most people don’t have a human assistant. And most people who are assistants aren’t only setting up meetings. AI is taking over the repetitive tasks, which frees up out time to work on more complex tasks.

Andrew: Is AI just hype? We're in 2018. Everything is hype. 

 

What are the biggest challenges to AI?

Andrew: AI can see patterns. The problems we’re just starting to get to is being able to re-use that knowledge and apply it in a whole other way.

Swapnil: How do we sell this in a controlled way? Our next big challenge is figuring out how humans look at unstructured data.

Raquel: Expectations versus reality. Trying to meet their needs while also explaining what our AI is capable of doing or not doing. There are so many nuances to balance on top of also on-boarding people to the product. It’s hard to say here’s what we don’t do, so there’s often a lot of trial and error in the first few weeks of use.

Amy: One is data. It’s tough to get good sets of data that are labeled cleanly. Open source data is great to start, but when lives are on the line, you want great data sets. Second, customers need a multi-model approach (hear, taste, smell). Even if you look at a Youtube page, you’re processing a lot of input. A lot of companies are specialized in just one domain or use case, so we have to figure out how to combine these efforts.

 

How do you generate value for the customer at the end of the day? 

Amy: In terms of how you make money, you either have to reduce costs or increase revenue for people. A lot of the use cases we see today is around reducing costs. So for example, automating a manual workflow like moderation with AI. The flip side is how do you up your revenue? That’s usually done via personalization. For example, using tracking pixels to know what a person may be more likely to buy. 

 

How do you think about an initial audience?

Raquel: For the first few years, the people who used our tech were people who were very tech-froward. They were more comfortable with mistakes in order to be at the forefront. But now more and more non-tech-forward people are aware of AI. For us, this is why education and on-boarding are so important.

Andrew: During our early stage, we had to list out what our product can actually do. [...] Our approach was that we knew that market research was probably the use case, but we know that people will think we can do all these things that we can't. So we came up with a list of like 15 value propositions like, "We allow you to talk to a group of people in real time." Then we bought a bunch of ads on LinkedIn, targeted a bunch of different positions at all the different company types we thought it would resonate with. They clicked through to a blank page because we didn't want to capture them, we just wanted to see what resonated. From there we figured out what roles were resonating with what value props. We got lucky because one role resonated with one value prop 10x more than others. 

And those were just some of the highlights that came out of our first Clarifai HQ AI meetup! We look forward to hosting more discussions in the future. Thanks to everyone who came out to this first event!

Special thanks to our amazing speakers:

Amy Soyeon Kim
Lead Product Manager at Clarifai
https://www.linkedin.com/in/amysoyeonkim/

Andrew Konya
CEO at Remesh AI
https://www.linkedin.com/in/andrewkonya

Raquel Ledezma-Haight
Product Manager at x.ai
https://www.linkedin.com/in/raquelledezmahaight/

Swapnil Parikh
Director of Product at HyperScience
https://www.linkedin.com/in/swapnil-parikh/

Moderator:
Josh Goldenberg
Lead Product Manager at Clarifai
https://www.linkedin.com/in/priorium/

 New call-to-action

Previous Next