Book: Human + Machine

human_machine“Human + Machine: Reimagining Work in the Age of AI” By Paul R. Daugherty and H. James Wilson
  • “When Amazon acquired Kiva Robots in 2012, it signalsed that the mobile bots zipping around Amazon’s warehouses were a key to their fulfillment advantage. not only do the robots help lift and stack plastic bins filled with different products, they also do the legwork of autonomously ransporting items aroudn the facility to human ‘pickers,’ who then slect the right products to fulfill different orders. Thanks to such increased efficiencies, the company has been able to offere same-day shipping for customers.”
  • “proctor & Gamble, whose CEO reently stated his goal of cuting supply-chain coss by a whopping $1 billion a year. Part of those savings will come from near-term efforts like the use of AI and the internet of things (IoT) technologies to automate warehouses and distribution centers. And other savings will come from longer-term projects, including the customized automation of product deliveries of up to seven thousand different stock-keepingn units.”
  • “Precision agriculture–which leverages AI and fine-grain data about the state of crops–promises to significantly improve yield, reduce the waste of resources like water and fertilizer, and increase overall efficiency.”
  • “At one such facility in Newark, New Jersey, run by AeroFarms, data is continuously collected on temerature, humidity, carbon-dioxide levels, and other variables, and machine-learning softare analyzes that information in real time to grow the crops (including kale, arugula, and mizuna plants) as efficiently as poissible. According to the ccompany, the Newark facility is expected to use 95 percent less water and 50 percent less fertilizer than traditional farms and, because the crops are grown indoors, pesticides aren’t needed.”
  • “Akshaya Patra, an Indian nonprofit with the vision that ‘no child in india shall be deprived of education because of hunger,’ combines the power of AI with blockchain (a digital, decentralized, public ledger) and IoT technologies. To achieve its vision, the company’s midday meal program provides one wholesome lunchtime meal to keep children suffiiently motivated and nurished to pursue their education.”
  • “Now feedback is digitized where once it was manually input, and blockchain is driving efficiencies in audit, attendance recording, and invoice processing. AI is used to accurately forecast demand, and IoT sensors monitor and sequence cooking processes to minimize waste and ensure consistent food quality.”
  • “Machine Learning (or ML). The field of computer science that deals with algorithms that learn from and make predictions on data without needing to be explicitly programmed.”
  • “Supervised learning. A type of ML in which an algorithm is presented with preclassified and sorted data (known in the field as ‘labeled data’) consisting of example inputs and desired outputs.”
  • “Unsupervised learning. No labels are given to the learning algorithm, leavingit to find the structures and patterns of the inputs on its own.”
  • “Semi-supervised learning. Uses both labeled and unlabeled data for training–typically more unlabeled data than labeled.”
  • “Reinforcement learning. A kind of training in which an algorithm is given a specific goal, such as operating a robot arm or playing the game Go. each move the algorithm makes toward the goal is either rewarded or punished. The feedback allows the algorithm to build the most efficient path toward the goal.”
  • “Neural network. A type of machine learning in which an algorithm, learning from observational data, processes information in a way similar to a biological nervous system.”
  • “Deep learning and subsets: deep neural networks (DNN), recurrent neural networks (RNN), and feedforward neural networks (FNN). A set of techniques to train a multilayered neural network. In a DNN, ‘sensed’ data is processed through multiple layers; each layer uses the output of the previous layer as it’s input. RNNs allow the data to flow back and forth through the layers, in contrast to RNNs, in which data may only flow one way.”
  • “In 2016, Tesla announced that every new vehicle would be equipped with all the hardware it needs to drive autonomously, including a bevy of sensors and an onboard computer running a neural network. The kicker: the autonomous AI software won’t be fully deployed. As it turnes out, Tesla will test drivers against software simulations runnnin in the background on the car’s computer. Only when the bakcground program consistently simulates moves more safely than the driver does will the automous software be ready for prime time. At that point, Tesla will release the program through remote sofware update. What this all menas is that Tesla drivers will, in aggregate, be teaching the fleet of cars how to drive.”
  • Using an Einstein-powered app, in pilot with select Coca Cola retailers, an employee on-site could take a cellphone photo of the cooler cabinet, and Einstein’s image-recognition services will analyze the photo to identify and count the different coca-cola bottles in it. Einstein would then predict and recommend a restockingorder, using CRM data and other information, including weather forecasts, promotional offers, inventory levels, and historical data to account for seasonal fluctuations and various other factors. The automation of the count and restocking order could save employees paperwork and time, and the added intelligence of the system has the potetnial to improve sales and increase customer satisfaction.”
  • “Take, for example, the global fasion coumpany Ralph Lauren, which has partnered with Oak Labs, a San Francisco-based startup, to develop an integrated commerce experience for shoppers. A key part of the technology is the connected fitting room, equipped with a smart mirrror that relies on RFID to automatically recognize the items that a shopper brings into the room. The mirror, which can translate six languages, anc then display details about an item. It can also change the lighting (bright natural light, sunset, club setting, and so on) so schoppers can see how they look in different settings. And the mirror can indicate whether items are available in additional colors or sizes, which a sales associate delivers to the dressing room.”
  • “Of course, the smart mirror also collects data bout the customer–the length of a fitting-room session, the convertion ratio (items bought versus those tried on), and other information–that a store can then analyze in aggregate to gain valuable insights.”
  • “The AI system can identify you by your gender, approximate age, and race. Boutique shops and fashion brands like Benetton have been deploying the high-tech mannequins to learn more about their customers. One retail outlet, for example, discovered that men who shopped during the first few days of a sale tend to spend more than women, promoting that business to modify its window displays accordingly. Another store reportedly learned that Chinese shoppers made up one-third of customers using a particular entrance after 4 p.m., so Chinese-speaking staffers were located there during these hours.”
  • “The simple truth is that companies can achieve the largest boost in performance when humans and machines work together as allies, not adversaries, in order to take advantage of each other’s complementary strengths. What’s easy for us (folding a towl, for instance) can be devilishly tricky for mahcines. And what’s easy for machines (spotting  hidden patterns in huge datasets, for instance) can be extremely difficult for us to do. Indeed, humands can thrive in situations where there is little or no data, whereas machines excell in situations where there is lots of data. Business require both kinds of capabilities, and it’s in the missing middle where that type of collaborative teamwork occurs.”
  • “Moreover, machine learning and other AI technologies can often function like ‘black boxes,’ resulting in decisions that might not be explainable. That might be aceptable for certain types of systems but other applications (for example, tose in the medical and legal fields) generally require humans in the loop.”
  • “The six-year-old company Stitch Fix is a prime example of the missing middle and process reimagination in action. Its main service is personal shopping, but with a twist: the company picks out new clothes and sends them straight to your door, based on data you provide, such as a style survey, measurements, and a Pinterest board. If you don’t like any of the items, you just send them back. Gone are the days of sepnding hours at the store and trying on dozens of outfits, only to find (if you’re lucky) a few that work.”
  • “Stitch Fix wouldn’t be possible without machine learning. But the company also knows a human touch is crucial to its success. Because Stitch Fix lives and dies by the quality of its clothing suggestions, its recommendation system–composed of both people and machines–is at the core of its service. The structured data, like surveys, measurements, and brand preferences, is managed by machines. Human stylists pay more attention to the unstructured data, such as images from Pinterest boards and notes from customers about why they’re looking for new clothes.”
  • “That’s why the role of data hygienist is crucial. Not only do the algorithms themselves need to be unbiased, but th edata used to train them must also be free from any slanted perspective.”
  • “Three Laws of Robotics”
    • “A robot may not injure a human being or, through inaction, allow a human being to come to harm.”
    • “A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.”
    • “A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws.”
  • “The technologies that power Amazon Go–computer vision, sensor fusion, and deep learning–are systems very much under development.”
  • “So, as a stop-gap to ensure quality ocntrol, Amazon employs people to watch video and scan images to make sure the cameras are tracking items and charging customers appropriately (sounds like trainer and sustainers, doesn’t it?). The store is an example of human-in-the-loop automated processes, with the goal of improving a system to perform more accurately and autonomously before deploying to a broad customer base.”
  • “[T]he human in a highly complex and automated system may become simply a compeonent–accidentally or intentionally–that bears the brunt of the moral and legal responsibliites when the overall system malfunctions. The metaphor of the moral crumple zone isn’t just about scapegoating. The term is meant to call attention to the ways in which automated and autonomous system deflect responsibility in unique, systematic ways. While the crumple zone in a car is meant to protect the human driver, the moral crumple zone protects the integrity of the technological system, itself.”
  • “Here are some ways to address the current shortcomings. First, create ways for algorithms to be accountable and identify root causes so that they can be fixed. Accountability isn’t just for human workers. Second, give human workers in the system the ability to second-guess the AI. trust that workers having judgements and provide valuable context, and that they can provide quality assurance for the service. Theird, allow rating systems to be used for algorithms or machines, not just for humans. Fourth, continually find where misalignments between control and responsibility are emerging. to fully address the problems that arise from developing system that lead to moral crumple zones and liability sponges, companies ened to spend significant effort realigning cultural values and norms.”
  • “Today, about 90 percent of the time of people who train AI applications is spent on data preparation and feature engineering, rather than on writing algorithms.”
  • See:

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s