It is impossible for anyone to accurately predict how the next several years will unfold in relation to AI, given its complex nature and rapid rise. But, it is possible to make specific predictions about AI trends for 2018 and analyze its key implications for business, government, and society. Based on insights from AI visionaries and PwC’s own advisory experience, the firm has come up with 8 such predictions.
AI will impact employers before it impacts employment
The job market will not be hit, but work’s nature will change. Loss of manual and repetitive jobs will be offset by new jobs performed by centaurs (human-machine hybrids). To prepare, business will need to bring together teams and data from different disciplines. Many organizations will need to start retooling.
AI will ramp up its presence in the workplace
Automation of complex processes, identification of trends and forward-looking intelligence will mean less busywork for humans and better strategic decisions. Organizations will want to figure out specific problems AI can help solve; new measures to assess business value will be needed.
AI will help answer the big ROI question about data
With data now being used to solve specific business problems, development and funding of AI solutions that draw heavily on data will get easier. Firms’ data infrastructure will have to be put into order.
AI talent race will not be decided by technologists alone
An increasing need will be felt for domain experts: retail analysts, engineers, accountants, etc., who can prepare and contextualize data, and work with AI experts. These employees will need to be trained for the specific data skills needed.
AI will make cyberattacks more damaging, and cyberdefense more effective
Pattern interpretation techniques like machine learning, deep learning, and neural networks also make it easier for hackers to find and exploit vulnerabilities, forcing businesses to start thinking about AI’s security applications. Cyberdefense may be the portal through which many enterprises get their first taste of AI.
AI’s Black Box will become a priority
Growing pressure from end users, clients, and regulators, counterbalanced with the costs involved, will mean companies will need an assessment framework to determine how much information each AI application should produce about its decisions.
AI will see nations sparring over it
China may take the lead; Canada, Japan, the UK, Germany and the UAE will do well too. International collaboration is on the horizon in some areas. Governments will need to increase funding; companies will need to keep an eye on the international competition.
AI’s responsible use will not be tech companies’ responsibility alone
With an emerging global consensus around responsible AI, public and private sector institutions are likely to collaborate on AI’s societal impact. Businesses might collaborate to form self-regulatory organizations.
Competing in the Age of Artificial Intelligence
As AI finally begins to realize its long-heralded potential and change the business environment drastically in the process, companies need to develop an understanding of these fundamental changes and develop well thought out, but flexible, strategies to harness the strongest capabilities of both men and machines.
After decades of unfulfilled promise, artificial intelligence has finally begun to realize its potential, driven by massive gains in processing power and better data collection methods. Developments in natural language processing and computer vision have played particularly important roles in helping machines perform tasks traditionally reserved for men. AI has also caught the public imagination over the years thanks to well-publicized events like Deep Blue’s chess victory over Kasparov, AlphaGo’s recent victory over Lee Sedol, IBM Watson’s victory in Jeopardy or even Google’s demonstration of self-driving cars. All this has prompted massive investments in AI-related areas in industries like finance, retail, and healthcare.
Differences between human and machine thinking
Because AI systems ‘think’ and interact, they are often compared with people. Humans are fast at parallel processing (pattern recognition) and slow at sequential processing (logical reasoning); exactly opposed to this, while computers have mastered parallel processing in some narrow fields, their strength lies in superfast logical reasoning.
Human intelligence allows for different types of problem-solving capabilities. Compared to this, at current processing power growth rate, artificial general intelligence is a long-term possibility at best. AI excels at performing specific tasks, fast and thoroughly.
So while investment in AI is critical, it is useful to ask this question: How can business leaders harness AI to take advantage of the specific strengths of man and machine?
AI’s effects on traditional ways of doing business
On notions of competitive advantage
In the 1980s, a technology tool in itself (such as Wal-Mart’s logistics tracking system) could serve as a source of advantage. But, now, because of open source software, algorithms in and of themselves will not provide an edge. Other traditional sources of advantage, like positional advantage and capability, are also being reframed by AI. Companies need a more fluid and dynamic view of their strengths, instead of a focus on static aspects. These three examples show how traditional notions of competitive advantage are changing:
Data – This is the raw material for AI systems, where large and early-moving companies like Google, Facebook, and Uber have an apparent advantage. They have created massive data repositories, which helps them collect more data, and also leverage it for better ad-targeting or in self-driving cars. But other companies without these resources can still do well through collaborations, even with competitors, to create their own privileged zones. They need to figure out areas where sharing works, and where it may not.
Customer Access – Traditional notions of customer access, like well-placed physical stores, are being replaced by AI-driven customer insights that help with personalized marketing and appetite prediction, generating higher revenues at negligible extra cost.
Capabilities – AI-driven automation is encouraging the replacement of traditional segmented and discrete areas of knowledge by cross-functional capabilities and agile ways of working.
On decision making
The speed of decision making is changing rapidly.
Predictive analytics and objective data are replacing decisions based on gut feel and experience (as can be seen in stock trading, online ads, supply chain management and retail pricing).
On the role of human employees
Humans will not become obsolete but will see rapid and major dislocations into new areas of work.
They will be needed in large numbers to build the AI systems.
They will be needed to help machines with common sense, social skills, and intuition as well as for quality checks.
In such an AI-inspired world, strategic issues are enmeshed with organizational, technological and knowledge issues. Agility, flexibility and continuous education are important for winning strategies.
Getting started towards winning strategies
Companies need to identify the jobs that machines or men are better at, develop complementary roles and redesign processes accordingly. Plus, they should be willing to change and adapt strategies at short notice. This is true in general for all areas in today’s business world, but especially so for AI-enabled processes.
Instead of a brute force implementation of AI everywhere, companies could evaluate through four lenses whether AI can create a significant and durable advantage:
Customer needs – Define the fundamental needs of your customers and see if AI can better address them
Technological advances – Study if the right technology exists to address your requirements and if you can make use of work already done instead of building a system from scratch
Data sources – Create a holistic architecture that combines existing data with novel sources, even if they come from outside
Decomposition of processes – Break down processes into relatively routinized and isolated elements that can be automated, taking advantage of tech advances and data sources identified above
These steps can be challenging for most companies. Setting up a center of excellence can help keep track of current and emerging capabilities, incubate technical and business acumen and disseminate expertise through the organization. It can then collaborate with the functions that would eventually put AI to use.
Thus, the full potential of AI can be achieved only if humans and machines solve problems together and learn from each other.
The intangible nature of AI technology can make it difficult for executives to build a business case for investing in new projects. Two examples show how well-thought out, high-impact projects can help with this.
The difficulty of making a business case with AI
The excitement around artificial intelligence has reached a tipping point, with its presence across sectors making it the primary battleground for technology vendors. But, despite the desperation on the part of companies, accelerators, and VCs to find a foothold, this vibrant market remains more conceptual than one of tangible substance. This makes an investment in AI technology a step into the unknown. A business case for such projects is not easy, needing reliance on intuition rather than ROI figures. Taking risks by investing in one or two high-impact scenarios can be very rewarding. The article covers two instances of how major organizations are doing this.
Sizing up the opportunity
AI requires using large amounts of data smartly, which the global law firm Linklaters is doing by turning its 175-year-old knowledge base into a competitive advantage. AI can create more sophisticated approaches for searching through this knowledge base, helping lawyers with quick information regarding legal precedents and previous projects. Linklaters’s CIO expects the ability to digitize and search contracts for key legal themes to become commonplace very quickly. The firm has already created an AI working group to analyze services in the marketplace and to work out how these technologies might impact the business.
But this change in how lawyers work also involves a cultural challenge. Senior partners will have to start trusting computers to do the same kind of work in seconds they have traditionally relied on associates to get done after spending hours with legal documents. Among other things, it’s the reputation of the lawyer and the firm on the line.
Using data to save lives
Moorfields Eye Hospital NHS Foundation Trust is involved with DeepMind Research, a project that involves the Trust sharing a set of one million anonymized eye scans. The hope is that these historic scans will improve future care, and lead to discoveries that make early detection and reduction in preventable eye disease possible. Challenges related to data security and confidentiality make it difficult to use non-anonymised data, which is actually more useful if demographic information is to be used to inform patient care. But stakeholders trust that similar projects will eventually lead to significant change in terms of how humans look at AI.
The future with AI
These experiments show that the potential of this self-learning technology is exceptionally exciting, and should encourage everyone involved in IT to investigate its uses. What also emerges is that, contrary to reports, automation does not simply lead to job cuts, but can create a new range of data science roles.
On May 9, Netflix launched its own research website. This highlights the focus Netflix has on Deep Learning and Data Science. The site is extremely well designed showing vertical classification of the different areas that Netflix research works on along with the horizontal business areas where Data Science is deployed at Netflix. It has some great articles with everything from video encoding to A/B testing where they use Data Science. I found the website to be very comprehensive making it a go-to destination for things Netflix Data Science from different verticals to jobs.
MyAI Interview Questions articles for Microsoft, Google, Amazon, Apple, Facebook, Salesforce, Uber, LinkedIn have been very helpful to the readers. As a followup, the next couple of articles were on how to prepare for these interviews split into two parts, Part 1and Part 2. If you want to find suggestions on how to showcase your AI work please visit Acing AI Portfolios. Career Insights check out the interview I did with Jesse. Now onto the Netflix Data Science Questions article…
To maximize the impact of their research, Netflix does not centralize research into a separate organization. Instead, they have many teams that pursue research in collaboration with business teams, engineering teams, and other researchers. From our publications, we can deduce that they are focused on the applied side of the research spectrum, though they do pursue fundamental research and think that has the potential for high impacts, such as improving our understanding of causality in our data and systems.
Netflix moves quite fast. There is one phone interview with the recruiter and another detailed one with the hiring manager. There are two onsite interviews with around 4 people first time (data scientists/engineers) and 3 people (higher level execs) a second time. There is a mix of product, business, analytical and statistical questions. Statistical questions mostly revolve around A/B testing: hypothesis testing. There are a couple of SQL questions too. Analytical questions usually include a hypothetical problem to analyze and metrics to evaluate product performance. Higher level executives mostly focus on background and past experience.
- Netflix Research Blog: All Articles
- Deep Learning for Recommender Systems: Talk Slides
- Reliable ML in the Wild Workshop (ICML 2017): Making ML Reliable at Netflix
AI/Data Science Related Questions
- How would you build and test a metric to compare two user’s ranked lists of movie/tv show preferences?How best to select a representative sample of search queries from 5 million?
- Given a month’s worth of login data from Netflix such as account_id, device_id, and metadata concerning payments, how would you detect fraud? (identity theft, payment fraud, etc.)
- How would you handle NULLs when querying a data set? Are there any other ways?
- What is the use of regularization?What are the differences between L1 and L2 regularization, why don’t people use L0.5 regularization for instance?
- SQL queries to find time difference between two events given a certain condition.
- Given a single day with a large sample size and a significant test result, would you end the experiment?
- What do you know about A/B testing in the context of streaming?
- How do you prevent overfitting and complexity of a model? How do you measure and compare models?
- How do you know if one algorithm is better than other?
- Elaborate on the recent project you developed for your company.
- Why do you use XYZ method? Elaborate on how to improve content optimization?
- What technology or item that most people feel will be obsolete in the future do you not agree with?
- Why Rectified Linear Unit is a good activation function?
- How should we approach attribution modelling to measure marketing effectiveness?
- How would you determine if the price of a Netflix subscription is truly the deciding factor for a consumer?
- If Netflix is looking to expand its presence in Asia, what are some factors that you can use to evaluate the size of the Asia market, and what can Netflix do to capture this market?
- Say the CEO stops by your desk and asks you whether or not we should go into an untapped market. How would you determine the size of the addressable market and the factors the Netflix should consider before deciding to enter the market?
Reflecting on the Questions
The data around Netflix questions is sparse. The high level questions resolve around A/B testing, recommender systems and foundational knowledge questions around regularization and activation functions. This is different from the other companies we have looked at previously where focus was more foundational. All job openings are usually senior level. Good experience combined with good preparation can surely land you a job at the largest international evergreen content cinema in the world.
Consumable List: Netflix Data Science Interview Questions
This article was also featured on KDnuggets: https://www.kdnuggets.com/2018/06/netflix-data-science-interview-questions-acing-the-ai-interview.html