Artificial intelligence (AI) has embedded itself into a variety of markets, industries, and use cases, but for the most part, it continues to be most accessible to the big tech companies with large budgets and teams. A growing segment of companies have recognized this unequal distribution of global AI access and are making strides to combat it through AI equity, AI democratization, and self-serve AI.
Pactera Edge is a global company that focuses on providing globally accessible AI solutions for its customers. Jonas Ryberg, Chief Globalization Officer at the company, recently shared his thoughts with the CIO Insight team about how self-serve AI works and why it’s important for a more globally useful artificial intelligence.
Self-Serve Artificial Intelligence Q&A with Pactera Edge Executive
- Introduction to Jonas Ryberg and Pactera Edge
- AI Globalization Efforts
- What is Self-Serve AI?
- An Expert Perspective on the AI Market
CIO Insight: Could you tell me more about your background and your journey to becoming Chief Globalization Officer at Pactera Edge?
Ryberg: I don’t know how far back to go, but I’ve been with the company for probably 11 years now. I joined and am based in Sweden, so I sort of split my time between Sweden and the US. I joined as the first employee in Europe back then, and my first job was basically to kickstart European operations.
I’ve worked in a number of different roles, starting on the localization side of things. So language services, and other localization-related tasks, like managing delivery, managing the supply chain, and some of the client engagement as well.
Eventually, over the last 10 years, things started shifting from only language services into first something that we couldn’t quite place, but it was basically transcription/annotation type work. And over time, it became clear that this was all about AI and machine learning, and that our clients needed data to fuel these engines.
“Over time, it became clear that this was all about AI and machine learning, and that our clients needed data to fuel these engines.”
So we saw a transition from traditional software localization to data type services that help you with AI and ML models. Today, I manage the globalization practice, but in fact, our business now consists of two service lines. So it’s still the software localization and the language services, but then the other part is the AI piece.
CIO Insight: Briefly, what does Pactera Edge do for its clients, particularly in the area of AI development and deployment?
Ryberg: At the highest level, we decide, optimize, and build platforms. Our take on platforms is that they are human-centric, they’re intelligent, and they’re global. And that is sort of based on the type of capabilities that we have. We have one team that is focused on intelligent experiences. And we have a globalization team, so my team, that focuses on the global aspects of these platforms.
We can cover the end-to-end story of AI, where we provide data collection, data annotation, and data labeling to create data sets for training AI. But we also build the models that can be part of the platform, or it can be a standalone solution or tool, like an algorithm or model.
And we offer the evaluation of the models where again, we prepare test data sets, and so on, that we can run through these models to make sure that they work as well as they can. And we can optimize over time. So it’s really an end-to-end offering in the space of AI, from the database to the consumption of the data, the training, and the building of the models.
CIO Insight: What role do you play in helping global businesses roll out their AI initiatives? What are some of your key global strategies for successful AI implementation?
Ryberg: My team handles the dimension of localization language services, and back in the mid-90s when the company started, we were doing things like translation work, and so on. It was very focused on helping tech companies take their product, software, whatever it was, to a more global audience or just a few additional markets.
Our previous focus made for a very easy transition for us into AI, where companies generally build their models for one market, and in most cases, with our clients, it’s for the US. But eventually, they want to go into Europe and Asia as well.
For example, if you think of voice assistants — or Alexa, or Siri, or Google Home, whatever you use — it works well in the US and maybe a few additional markets. But then as soon as you want to have, in my case, a Swedish version of Alexa or Siri or Google Home, you will need to train that voice assistant for that particular market, and you need data in order to do that.
[Collecting] training data and test data for the AI and ML models, that can be a pretty daunting task in terms of how much data you need for that. So [gathering and managing that data] is what we’re helping companies with when it comes to the global aspect of these platforms.
CIO Insight: What recent trends have you noticed in the global AI market?
Ryberg: Accessibility, I think, is the big one. And that’s the one we’re excited about. The other one that we’re excited about is the AI localization and AI globalization aspect.
We’ve seen sort of an evolution, right? The big tech companies that started doing AI and ML five years or 10 years ago, they’ve moved into a number of other markets already. Even there, we see a lot of demand to grow into now maybe tier three or tier four markets, where it might be Sweden or Estonia or some of these smaller countries. By doing that, you’re also democratizing AI and making it more accessible globally.
And I do think that self-serve will be very important for a lot of these companies if they don’t have internal teams. There’s a shortage of talent for sure, globally.
More on AI Trends: AI Software Trends for 2021
CIO Insight: How would you describe “self-serve AI” to someone who isn’t familiar with the concept?
Ryberg: For me and generally for Pactera Edge, we’re focused on the data aspects of self-serve AI, and that might, over time, expand into pre-trained models, and so on. But the way we look at it, if I give you a little bit of context, we work with these big tech companies that I mentioned, and they are generally pretty well equipped to take AI, train AI, and build models for a global audience. They have big teams that do that.
“These big tech companies […] are generally pretty well equipped to take AI, train AI, and build models for a global audience.”
But we’ve seen over the past couple of years when looking at the second wave of companies going into AI, be it in different verticals or just smaller companies than the big tech companies over in the US, they don’t have that experience and they don’t have these big teams.
In most cases, when we work with these teams, our point of contact would be a smaller data science team, and they don’t want to spend all their time collecting the data and annotating the data. They want to work with the models and figure out how they can optimize the actual AI use case, rather than work on the data side of things.
So self-serve AI for us is really self-serve training data and test data for AI, where we build a different version of the platform that we use for these big tech companies and change it into a more self-serve version.
Someone can log in, with a premium account login, and they can either request a certain data set where we will provide a data collection with our talent pool, or they can upload their data and ask for annotation or labeling; [we provide] whatever they need with that data set in order to make that actionable for training or testing purposes.
Read more: Data Management Is Becoming Less Complex
CIO Insight: How can AI and other digital transformation efforts be made more equitable/accessible to smaller businesses?
Ryberg: The self-serve launch that we’re in the middle of now is really about reducing the need for managed services on our side. [Managed services] can get pretty expensive quickly, especially if you need a million data assets to train a voice assistant for Swedish, for example.
And if you look at training at that scale, it definitely requires a program manager, it requires a lot of quality management and engineering, and so on. And a huge talent pool internationally, across multiple markets, if it’s more than one market or locale that you want to address. The self-serve for us is about reducing the overhead essentially, the program management and engineering efforts, and providing easy access to a global talent pool that can help you create these data sets.
We have certain templates for typical use cases that people would want to annotate some type of document for computer vision purposes, or they would want to have recordings, and then they want that either transcribed or annotated to make it actionable for an AI training model. And so that’s essentially what the platform offers. So reducing that additional costs anything that is not part of core need, and
We have a globally dispersed team of about 400 to 450,000 people that work on these [template and localized data] requests through the platform for our customers. It’s really a big global pool of digital nomads, people who want to work online and make money and travel at the same time, or they work from home, they work sort of part-time, or when it suits their lifestyle.
By leveraging or partnering with a global talent pool of people, we connect those people with the needs on the client-side for this training data.
More on AI and ML: AI vs Machine Learning: What are their Differences & Impacts?
CIO Insight: How are other tech companies working toward self-serve and globalized AI models?
Ryberg: We do see other companies try to do globalized, self-serve AI more in a BPO type setup, where they would have big in-house teams in one location and try to build these data sets through that model. It doesn’t work as well because there are limitations to the variety of people that you can put in that room, right? For us, it’s about a lot of different types of personalities, and varied data sets as well.
“You want to avoid bias in the data sets, right? You need diversity.”
So you want to avoid bias in the data sets, right? You need diversity. One approach that others have taken is to have that in-house model, but it offers less diversity for cost reasons.
CIO Insight: What is the most common mistake you see companies make when they first start using AI solutions? How can they prevent those mistakes/problems?
Ryberg: A common problem is not enough data or not clean enough data. You can take that too extreme as well and have too clean of data. Data that is too clean won’t work in the real world where data isn’t necessarily clean and predictable all the time. So I guess what I’m trying to say is that not many companies don’t have the right data for their purposes.
And I also think that a lot of people and companies still underestimate the need for localization of the data or globalization of the data if they want to go into additional markets. And of course, there’s a language aspect. I think that’s pretty clear to everyone that if you want to sell in Germany or have your product work in Germany, you need it to be in German.
But everyone doesn’t realize that, even if it’s not a language-based product, you might need to look at how you adapt your AI product to the German market, for example, or any sort of market wherever you’re going.
And we see that, for example, with search algorithms or product recommendation algorithms that aren’t language-based at all, but the needs or the products that you want to recommend in Germany or China or India will be different from what you would recommend in the US, based on the local preference.
You would need to train your AI also in those cases when it’s not language-related. So that’s something else we’re trying to communicate more: that you need to localize or globalize your AI for the markets that you want to be in, even if it’s not language-based.
Read more: What Is Natural Language Processing?
CIO Insight: What do you think is the most interesting product/solution on the AI market right now? Why?
Ryberg: To me, it’s still probably voice-related, the potential of voice products. I lived in the US up until July and then I moved to Sweden this summer, and I just realized how much I missed our Alexa because it hasn’t been launched in Sweden yet for Swedish. So I had to go back to the US and pick up an Alexa just to be able to use it.
“Voice AI […] is making a huge impact on how we live and how we function.”
And it’s becoming so much more than just asking your Alexa for your latest news or the weather or whatever. I use it a lot. And so I think it’s that and how embedded it has become in our lives and how important it is now. So I think it’s that; at a very high level, but voice AI in general, I think is making a huge impact on how we live and how we function.
Read more on Datamation: Top Performing Artificial Intelligence Companies of 2021
About Jonas Ryberg
Jonas Ryberg is Chief Globalization Officer for Pactera EDGE. He and his team provide AI Globalization, AI Data Services, and Language Services. Jonas has 20+ years of industry experience in various roles, from solution-ing and linguistics to sales and management, both as an executive and as a company founder.