Cutting Through the Hype: Making AI and Machine Learning Simple for the Future
Tamara: Can you share a story that inspired you to get involved in AI?
Josh: I became interested in AI for its potential to transform healthcare. My mother passed away from an incurable disease before the birth of my second child and it drove me to want to make a difference. I don’t have an MD and I’m not a medical researcher so getting involved in AI was a way I could use my background in computer science to make an impact. My mother’s passing was also part of the inspiration for working with Kaggle to create the Data Science Bowl, an annual data science competition for social good that has largely focused on using AI to combat health issues, like using AI to improve heart health and make earlier diagnosis of lung cancer possible.
Tamara: Describe your company and the AI/predictive analytics/data analytics products/services you offer.
Josh: At Booz Allen, which has one of the largest data science and analytics teams in the world, we:
- Build. We help our clients understand how they can use machine learning to tackle the most complex challenges of today and tomorrow, and then build integrated solutions using machine learning, automation, and modern digital platforms.
- Train. To help build the future workforce, we are working with NVIDIA to train government employees to apply machine learning techniques to key challenges in healthcare, defense and cybersecurity.
- Apply for social good. We apply advanced machine learning capabilities to employ new approaches to issues like the clinical treatment and research of traumatic brain injuries among US service members.
Tamara: How do you see the AI/data analytics/predictive analysis industry evolving in the future?
Josh: One day, Machine Learning (ML) will be as ubiquitous and important to businesses and government as Labor and Capital. We need ingenious people who can imagine, build and use these evolving technologies to automate repetitive work and solve complex problems at the speed of the Internet. As Angela Zutavern and I described in our book, The Mathematical Corporation, ML has already invented a malaria vaccine more effective than anything humans have created; and will drive innovation in every industry imaginable. For example, during the last census in 2010, the collective number of miles census workers drove clocked in at five times the length of all the roads in the US. Since they planned their own routes and travel time, they routinely arrived when no one was home, leading to repeat visits. For the 2020 census, census workers will follow algorithm-delivered advice on tablets to optimize their travel routes and visits times, which is expected to eliminate tens of thousands of miles of travel.
Tamara: What is the biggest challenge facing the industry today in your opinion?
Josh: Bias and failing to consider ethical implications from the beginning when designing and building machine learning systems. ML has great potential, but if we don’t adopt it responsibly, we may do more harm than good. For example, leaders that rush to implement ML could end up inadvertently automating bad decisions they are already making which escalates the potential for negative outcomes.
Tamara: How do you see your products/services evolving going forward?
Josh: We see ML already fundamentally changing the nature of business and mission operations which is why it’s a major area of investment for Booz Allen today. As the technology becomes more common, people won’t stop to think about whether the solution they’re using is powered by ML. So it is important for us to continue to help our clients cut through the hype to better understand and responsibly adopt it, and develop practical applications. To do so, ML will be embedded in all the services we provide. We will continue to grow our team of technologists and domain experts to stay a step ahead of technology advances, while ensuring all of our consultants and analysts are trained in advanced principles of machine learning, deep learning and machine intelligence.
Tamara: What is your favorite AI movie and why?
Josh: Short Circuit from 1985. I like how it humanizes AI (in this case the robot protagonist).
Tamara: What type of advice would you give my readers about AI?
Josh: Don’t put stock in the narrative about ‘killer robots’ that will take over the world. The technology is still very much in a nascent stage and too much focus has been given to this trope at the expense of developing good, functional technology that can help solve business problems. Before diving in, we recommend organizations take five steps. First, consider the goals or value you want to capture through the investment. Organizational aspirations may be as simple as creating efficiencies in internal operations or as audacious as transforming the organization’s mission. Second, consider your appetite for risk. How much are you willing to tolerate? Third, evaluate the state of your data assets. ML is still highly dependent on its ability to learn from vast quantities of labeled and well-organized data. Open data is good, but exclusive access is a must. Fourth, evaluate the state of your ML talent. Talent is scarce so consider partnering with research institutions and academic organizations to have access to necessary talent. Fifth, understand your organizational values and how ML could potentially threaten those values. We believe organizations should prioritize respect, transparency, privacy and equity.
Tamara: How does AI, particularly your product/service, bring goodness to the world? Can you explain how you help people?
Josh: Given the nature of our client base, much of our work is focused on applying ML across industries to deliver positive social impact. For example, we worked with MedStar’s Institution for Innovation to develop a technology called Dictation Lens which is being used in D.C. hospitals today to improve patient care. The technology automates the assessment of relevant facts in a patient’s complete medical history so clinicians can quickly find critical data in the moments of urgent care. We are also partnering with companies like Microsoft and Samsung to reimagine military training, which is often static and predictable, to be more innovative. Combining ML with immersive technologies like virtual reality, augmented reality, and mixed reality can generate dynamic scenarios to train soldiers in ways similar to what happens in real missions.
Tamara: What are the 3–5 things that most excite you about AI? Why? (industry specific)
Josh:
- ML has only just begun to demonstrate its potential to diagnose and treat diseases. In the near future, its ability to parse large troves of patient data could accelerate life-saving medical research and develop treatments for previously incurable illnesses. For this year’s Data Science Bowl, thousands of participants are training ML to automate nuclei detection, a critical step to unlock cures to deadly diseases.
- As the opportunity for data collection expands, it becomes impossible for analysts to effectively review and analyze all the information gathered. In the current state, ML technologies can help alleviate the information overload burden, allowing human analysts to focus their efforts on issues of greater complexity and enable more thorough security efforts of Department of Defense forces, for example.
- Cyberattacks have become more sophisticated than ever before and traditional tools can no longer scale to protect today’s complex organizational networks. ML is particularly adept at anomaly detection, particularly in cyber security and fraud detection. As we recently demonstrated at NVIDIA’s GTC AI conference, we have been testing some of the latest GPU driven machine learning techniques to identify Domain Generation Algorithms (DGAs) which have been extremely successful in helping hackers circumvent traditional cyber defenses to steal data from an organization’s network.
Tamara: What are the 3–5 things worry you about AI? Why? (industry specific)
Josh:
- Privacy considerations — ML depends on access to large data sets and when companies fail to protect access to sensitive personal information, there are serious privacy implications for individuals. Facebook’s recent data misuse showed that the company allowed a third party app to access more users’ information than those who downloaded the app, and potentially led to targeted political advertising. Until regulators and policymakers are able to develop a cohesive and comprehensive set of rules regulating the use of sensitive personal information, we will see companies monetizing your personal information in very different ways.
- Opacity of algorithms — We are approaching a future where ML will be used to suggest whether you are stopped by police, whether you’re hired for a job, the mortgage loan rate you’re offered, or whether you’re admitted to the school of your choice — decisions with real consequences. Transparency of how ML systems work, such as the variables used in algorithms, how training data is selected, and how verification testing is done should be standard practice for ML practitioners.
- Equity and social inclusion — ML systems are created using data produced by humans, which often reflect our biases, thereby perpetuating and even amplifying existing societal biases. ML’s efficiency and ability to scale can quickly lead to the systematization of bias, which make applications across all industries potentially threatening.
Tamara: Over the next 3 years, name at least one thing that we can expect in the future related to AI?
Josh: In the not-too-distant future, I expect we’ll see an industrial level ML-related accident which will cause us to re-think regulation and safety implications. An incident like this could provoke the U.S. to develop a national ML strategy, a step other countries have already taken. In a recent paper we co-wrote with CSIS titled “Shaping MI’s Impact on U.S. Economy, Security, and Society”, we advocate for a national ML strategy that promotes the safe and responsible development of ML.