Gaining an understanding of deep learning is the best place to start for executives interested in keeping up with the rapid developments in AI. An exciting and powerful subfield of AI, deep learning focuses on how computers learn as opposed to how they are explicitly programmed by us.
In this method, researchers place concepts into a hierarchy. At each layer, a machine learns a concept and passes it to the next layer, which in turn uses it to build a more sophisticated concept. The more layers these models have – or the ‘deeper’ they are – the more concepts they can learn, leading to ever-expanding possibilities of applying AI to business problems.
Here are four deep learning breakthroughs business leaders should be aware of, from the immediately applicable to the most cutting edge. Drastically different applications prove that these can be used innovatively to introduce the next great product or service based on the right set of data.
1. Image understanding – Deep learning algorithms called convolutional neural networks, which can be trained to identify objects in an image, already do better in image classification than humans.
· Google Image Search
· Self-driving cars
· Disease diagnoses
2. Sequence prediction – Recurrent neural networks can be trained to look at huge amounts of past sequences of characters or data, learn their patterns and generate future sequences.
· Producing human-like handwriting
· Prediction of user demand by Uber
3. Language translation – Discovered in 2014, this technology uses sequence-to-sequence architecture and recurrent neural networks to make machine translation almost as good as human translation. So far used for narrowly defined domains, this area holds significant promise.
· Google Translate and Apple’s Siri
4. Generative models – Creation of models that generate complex data, like images that resemble faces but aren’t actual faces. This is possible due to architectures called generative adversarial networks, which use convolutional neural nets. Though their business applications are limited as of now, research is on to find exciting ways to deploy their power to solve practical challenges.
· Aiding image classification models to distinguish real images from fake ones
Open source implementations of the above breakthroughs make it possible to download pre-trained models to apply to your data. For example, pre-trained image classifiers can be purchased to feed your data through to classify new images.
This overview brings you closer to companies like Uber and Google that put deep learning models to good use. The next generation of business applications is yet to come, and will be driven by new ways discovered by you to apply these techniques to your own data.
To capture value in this market, suppliers need to clearly understand the key trends in the industry, and devise a go-to-market strategy accordingly.
With better algorithms, increased stores of data and improved hardware performance, AI is finally beginning to achieve its potential. Sophisticated technologies like Deep Learning further increase the need for investments in leading-edge hardware and software. While several companies have taken early steps to win in this market, the industry is still nascent, leaving the field open for supply side companies to capture value and see a return on their huge AI investments.
The supply side landscape
Before moving on to a strategy, it is important to clearly understand the chaotic supply landscape in AI. The offerings in AI can be seen as a nine-layered technology stack:
Positioning to win in AI
To capture value in the growing AI market, companies on the supply side need to heed the following six points:
1. Value capture in the consumer sector will be limited initially – Early AI offerings are product and service enhancements that appeal to consumers and increase their engagement level, but do not contribute to the bottom line, e.g. online translation, digital voice assistants. These are typically offered by large tech companies, who have bigger pockets and access to more data, leaving little opportunities for smaller companies initially.
But fee-based offerings, like home assistants and self-driving cars, will create more opportunities in the next wave of innovations.
2. Winners will focus on microverticals in promising industries – Instead of getting confused by the hundreds of opportunities available, suppliers should focus on specific industries.
To select the right industries, following criteria can be helpful:
· Size of the industry
· Potential for disruption within the industry
· Maturity (or readiness to embrace new solutions)
Based on these criteria, the strongest opportunities appear to be available in the public sector, banking, retail and automotive industry.
Once an industry is identified, suppliers should identify specific verticals where solutions can result in high ROI. The value proposition is not compelling for broad, horizontal, industry-wide solutions.
3. Suppliers will need to provide end-to-end solutions – Customers look for solutions across all nine layers of the technology stack from the same supplier. This saves effort spent in getting different components to work together, and also provides suppliers with strategic foothold with customers. Such wide-ranging capabilities could be completely in-house, through acquisitions or through collaborations. Nvidia, for example, offers its Drive PX platform for cars as a module, not just a chip, combining processors, software, cameras, sensors and other components.
4. Most value will come from services and hardware – Bucking the general trend of commoditization of hardware, about 40-50% of value to AI vendors will come from this part of the technology stack. This is because, in AI, every use case has slightly different requirements, needing partially customized hardware (head nodes, inference accelerators, training accelerators).
Another 40-50% value will come from services (solutions and use cases). System integrators, who often have direct access to customers, will capture most of these gains by bringing solutions together across all layers of the stack.
Software, surprisingly, is unlikely to be a differentiator. Since data – another important component of AI – comes from the customer itself in most cases, it will not provide much value either. Though a market for third-party data might emerge in the future.
5. Specific hardware architectures will be critical differentiators – Cloud will continue to be the favored option for many applications, given its scale advantage. Within cloud hardware, customers and suppliers vary in their preference for application-specific integrated circuit (ASIC) technology over graphics processing units (GPUs), and the market is likely to remain fragmented.
There will be a growing role for inference at the edge, where low latency or privacy concerns are critical, or when connectivity is problematic. At the edge, ASICs will win in the consumer space because they provide a more optimized user experience, including lower power consumption and higher processing, for many applications. Enterprise edge will see healthy competition among field programmable gate arrays, GPUs, and ASIC technology. However, ASICs may have an advantage because of their superior performance per watt, which is critical on the edge.
6. The need to act is now – To win in AI, reliance on status quo won’t work. Strategy revision and big bets to develop solid offerings right now is the key, instead of waiting two to three years. Unconventional strategies are already seeing high returns. For instance, Nvidia is increasing its AI R&D expenditure and creating an end-to-end product ecosystem, gaining market share and seeing exceptional returns from AI offerings.
Keeping in mind the above points, these three questions could be useful in devising a strategy for the AI market:
· Where to compete – Look at industries and microverticals, select a use case that suits your capabilities and address the customers’ most pressing needs
· How to compete – Look for partners to provide end-to-end solutions; Be open to creative pricing options
· When to compete – Instead of striving for perfection, focus on solutions that help you establish a presence now
The advent of artificial intelligence is throwing up challenges for many sections of the industry. This is a revolution with many possibilities, but it also means that the human workforce will have to adapt smartly to extract its full potential.
One of the more obvious changes we are seeing is how robotic systems are beginning to handle many transaction services, like taking food orders at a drive-thru food outlet or handing out money at a bank ATM. Not just low-paying entry-level jobs, but even mid-level positions will see increasing presence of AI as the technology develops, especially in sales.
AI vs. Human intelligence
Does this mean that all sales workforces will be replaced by robots completely? Before answering this, let’s consider the key arguments that AI has over a human worker. One, salespersons take home salaries, which an employer saves on when working with a machine (even after considering the maintenance and programming costs). Two, automation allows customers access to services at any time – think waking up an admin guy at 3 am against switching on a robot. Three, even the most ardent opponents of AI will concede that there are some ways in which artificial intelligence scores over human intelligence. So, in a competitive world, employers are not completely unjustified in giving AI a chance.
What history teaches us?
But it’s not all doom and gloom. Consider the automobile industry as a comparison. Despite early beliefs in the role of the human element, the machine revolution of the 80s and 90s did put a large number of workers out of business. At the same time, those who learned to work the robots stayed behind. And other opportunities opened up – gas boom picked up; so did the mortgage industry – absorbing workers from the auto industry and creating new jobs too.
The way forward
Thus, the challenge for salespeople also brings up an opportunity in its wake. The future of sales is learning automation, the tech side and artificial intelligence so that you can work with the robots. And as other sectors of the industry open up, you need to keep your eyes out for them. It is all about embracing the new technology, so that instead of getting washed away by the wave, you can ride it successfully.
What is Reinforcement Learning?
It is an approach to artificial intelligence that gets computers to learn like people by doing things incrementally well, without explicit instruction. The author gets a taste of this at a Barcelona AI conference where self-driving cars in a simulation learn to maneuver through a complicated four-lane highway by practicing moves over and over. This is also the approach that helped AlphaGo, designed by Alphabet subsidiary DeepMind beat the World No. 1 Go player in 2017 – Go is a notoriously complex game, making this achievement particularly remarkable.
Reinforcement Learning – A Historical Perspective
Psychologist Edward Thorndike first documented this principle over 100 years ago with an experiment involving cats who learned to escape from a box after stepping on a lever by chance. An early reproduction in machines was in 1951, where a machine simulating a rat escaping from a virtual maze saw some synaptic connections strengthen to reinforce successful behavior. Following decades saw intermittent successes, but using this for more complex tasks became computationally impractical.
The Breakthrough of Deep Learning
In recent years, reinforcement learning has become so formidable because of breakthrough in deep learning, which uses a large simulated neural network to recognize data patterns. This makes storing values for every move and updating them – the basis of reinforcement learning – easier.
This breakthrough is also what made DeepMind first excel phenomenally in Atari video games in 2013, resulting in its acquisition by Google. Today, Google is using these capabilities to make its data centers more energy efficient.
Reinforcement Learning in Self-Driving Cars
This approach is particularly well-suited for simulating humanlike behavior in self-driving cars. Today’s driverless cars often falter in complex situations that involve interacting with human drivers, such as traffic circles. To prevent extreme behavior – taking unnecessary risks or being too hesitant – these cars will have to acquire nuanced skills.
Mobileye, the Israeli firm that developed the Barcelona demo, plans to test the software on a fleet of vehicles with BMW and Intel this year. Google and Uber are already testing reinforcement learning for their vehicles.
While reinforcement learning helps automated driving by enabling good sequences of decisions, which is more efficient than pre-programming all such decisions, there are challenges too. The huge amount of data, the time needed to practice simulations, multiple objectives (avoiding accidents vs. keeping roads safe for other cars) – all make this an extremely complex area.
Artificial Intelligence has developed enough in the last few years to offer applications beyond relatively simple tasks. But a general AI system, which can solve many types of problems and is self-aware, is still not available. Even within the scope of narrow AI, where machines use abilities like deep learning and natural language processing to solve very specific problems, what new developments mean from a manager’s perspective are slightly different from what they do for a general user. The business aspect of AI can be considered with the help of three important questions:
1. What is AI?
As computers lowered the cost of arithmetic, AI has the ability to lower prediction costs, i.e. generating new insights from existing information. But, judgment will remain a human forte for now, e.g. AI predicts the best fit candidate for a job position, while a manager handles mentoring and taking ethical decisions. In view of this, not just managerial skill sets will have to adjust, but other changes will occur too.
2. How will AI influence business strategy?
- · Increasingly more complex prediction problems will be handed over to AI systems.
- · Managers will have to train workers in judgment-related tasks.
- · To properly time the shifting of workforce training, rate and direction of adoption of AI technologies will have to be continually assessed.
- · Processes that combine the strengths of human workforce and AI will have to be developed.
- · Strategic alliances to access effective data will have to be formed.
3. What are the major management risks from AI?
- · Replacement threat – Some jobs may no longer be necessary as reliance on AI systems grows.
- · Dependence threat – Prediction dependence on algorithms could create new vulnerabilities and potential inefficiencies, especially if errors and biases creep into these algorithms.
- · Security threat – Protecting sensitive corporate information will not just be an IT problem, but a responsibility of the management too.
- · Privacy threat – As more data is collected from the labor force, new questions of autonomy and freedom will have to be addressed.
The net effects of the strategic implications of AI on management, along with related risks, are uncertain – what is certain is that AI will change business. Understanding this evolution will help managers be better prepared for the business world of tomorrow.