Worldwide spending on artificial intelligence (AI) systems is forecast to reach $35.8 billion in 2019, an increase of 44.0% over the amount spent in 2018. The dramatic rise of machine learning and artificial intelligence technologies not only impacted software and the Internet industry, industries such as healthcare, finance, manufacturing and supply chain also have a lot to gain from these technologies.
We recently had an in-depth interview with Alexandre Hubert, Lead Data Scientist of Dataiku. Dataiku is a computer software company develops collaborative data science software marketed for big data. As the lead data scientist, Alexandre works on several use cases for financial institutions and retailers.
What is the biggest assumption enterprises make when considering AI solutions?
The biggest assumption they make is it will be fast, simple and transform their business from top to bottom. Indeed AI has the potential to revolutionise lots of different organisations, regardless of the sector we are talking about. However, it is not an easy process, especially for organisations that are used to do business in a certain way. This type of innovation needs some redesign of internal ways of working that cannot be deployed simply with large investment and the recruitment of elite data scientists.
What's the biggest boundary to integrating AI for industry companies?
The biggest boundary to integrating AI at scale in large companies are around three main problems:
- Data: there are still lots of legacy systems, it is difficult to integrate with the newest data storage systems. Without a clear data management strategy, innovation is going to get slow down.
- People: in large scale enterprises, it is still very common to have different teams working on similar problems, only to realise it months later. Teamwork and collaboration would have ensured efficient delivery, and make a more robust product benefiting from different perspectives and ideas.
- Processes: it is necessary to make sure that a large organisation operates smoothly. However, they tend to slow down innovation especially when it's time to operationalising AI.
What do you think the biggest myth around AI and machine learning being propagated around your industry is?
AI and machine learning has certainly the potential to transform fundamentally businesses and most certainly their relationship with their customers. By offering AI designed products, they can significantly change their lives for the better. That being said, AI and machine learning are not the solutions to everything and every unsolved problem! They are powerful tools that become useful only when they are deployed in production, serving the business and by extension the customer. Building a strong neural net is not the end of the story, it is only the beginning: how to integrate it within the website, how to monitor its performance over time, how to alert the designer in case we observe a sudden shift in the data distribution in production compared to what we’ve seen in design.
How has the industry’s attitude towards machine learning and other AI changed from when you first entered the field compared to today?
AI and machine learning generated a lot of excitement when I entered the field about 5 years ago. An all-new set of possibilities was pitched to organisations that were keen to buy those big promises. I would say that people are now certainly more educated about what AI is, what it can do and how we should design, from a human-centred approach to a smooth deployment in production. However, it is only the beginning and we need to keep educating ourselves to make sure the benefits of those powerful assets are visible in our day to day job and without harming anyone.
What are you most excited about in AI right now?
I find fascinating how AI is transforming the way large corporations operate and how it redefines in-depth their relationship with their customer. I think the banking sector is a very good example where we saw Fintech and Insure-tech taking the lead on innovation forcing the major players to adapt to the new realm of the market, the behaviour, and the needs of the customers. The use cases are numerous from customer retention or acquisition or claims management with the integration of AI in portable devices and apps.
What are the main challenges to your role this year?
As a software vendor, our role is to help every type of organisation to develop strong AI-driven products and help them to operationalise them. Although it is getting better, I feel that my main challenge is still to lower the time from design to production, largely impacted right now by the silos between data, people and technology. From a personal point of view, I am transitioning from Dataiku Northern Europe to Dataiku APAC and it will also be a challenge to learn about the data culture in that part of the world. I am very excited about it.
What will you be discussing in your presentation?
Designing and validating models is only one part of a whole data science project. There is a gap to bridge to go from design to business value. Having these models running in production is the only way to create lasting measurable value for a business. And today production issues are the main reason many companies fail to see real benefits come from their data science efforts. During this presentation, we will first understand what "to go into production" means and then consider actionable steps to build production-ready data science projects.
Dataiku will demonstrate a private documentary screening at the Big Data & AI Leaders Summit. What should our audience expect?
Data scientists are in extremely high demand as organisations leverage their skills to tackle all sorts of problems. Yet the demand (and the number of job postings) outpacesthe supply. According to the Harvard Business Review, data scientists have the “Sexiest Job of the 21st Century.” This puts data scientists in a privileged position with a lot of job security, but curiously, what they actually do day-to-day remains a mystery to many. Data Scientists are at the front lines of data, privacy, and civil liberties debates, as they construct systems that take advantage of (or don’t) personal user data. It’s an incredibly challenging role, but data scientists are passionate, curious people who help generate meaning from sensors and stories from data. This documentary called DATA SCIENCE PIONEERS is for data scientists and everyone who works with them. Whether you want to be a data scientist or are one or work with them, this documentary will help you better understand data culture, history, and the variety of work experiences data scientists have. Ultimately, this documentary is for anyone interested in data, whether professionally or as an amateur.