How was the book?

Do you remember your first AI moment? Was it with Siri? Or was it predictive text input? Or even with Tesla? Anyways those are the AI technologies that we currently see used and is commercialized.

AI is a prediction technology that will make decision making more efficient. Or as the writers define ”Our first key insight is that the new wave of artificial intelligence does not actually bring us intelligence but instead a critical component of intelligence—prediction.”

Somehow the name artificial intelligence is misleading, because ”the breakthrough that will give rise to general artificial intelligence remains undiscovered.” Maybe we should be talking about prediction technology than AI. But let’s remain true to the book and I will also talk about AI.

Key topic of the book is that Agrawal, Gans & Goldfard will provide a framework ”for identifying the trade-offs associated with each AI-related decisions”. The writers want to ease the activities that any organization has regarding artificial intelligence.

What are the key learnings of the book? 

The reason why AI is happening now is pure economics. This technology is cheap enough for large scale deployments. The components that make AI affordable are getting cheaper and cheaper all the time as we have learned from the Moore’s law. Components that make AI possible are for example cloud computing, data analytics capacity, etc.

First key learning is that efficient predictions to support decision making are available already. Soon predictive applications will emerge into new market places and quicker than ever.

Short definition of prediction is that it is ”the process of filling in missing information. Decisions usually occurs under conditions of uncertainty”, but a prediction is not a decision. So the decisions requires applying ”judgment to a prediction and then acting.” By the way humans have always performed prediction and judgment together and especially we Finns are very good on this. Goes without saying that it’s important to realise that AI don’t make any judgment and only humans do, because currently only humans can understand and express the trade-offs and ”the relative rewards from taking different actions.” Marching order will remain that ”a machine predicts what is likely to happen and humans will still decide what action to take”.

CASE: ”For every bomber that returned from bombing raids over Germany, the engineers could see where they had been hit by antiaircraft fire. The bullet holes in the planes were their data. But were these the obvious places to better protect the plane? They asked statistician Abraham Wald to assess the problem. With this insight, the air force engineers increased the armor in the places without bullet holes, and the planes were better protected.”

Second key learning is that how AI will change strategy. The writers gives a very concrete example from Amazon. They have a new business model which is called ”from shopping-then-shipping to shipping-then-shopping”. I.e. Amazon might send you the next item before you order it. You must evaluate your business model and current applications against the AI potential. If and when you see the time is right for the AI development work in your organisation you could start using startup methods. Methods such as Lean startup method could help you developing your own way of working in the AI era. One of the end-results will be the increase of the value of judgment.

Prediction is always based on data. ”More and better data leads to better predictions. In economic terms, data is a key complement to prediction. It becomes more valuable as prediction becomes cheaper.” But data collection is not cheap – it should be considered as an investment. Three points about data:

1. Input data is needed to feed to the algorithm and used to produce a prediction.

2. Training data is needed to ”generate the algorithm in the first place”

3. Feedback data is needed to ”improve the algorithm’s performance with experience”.

Data collection will be a expensive part of the AI business development. Here are the scenarios where the data is needed:

1. ”Known knowns. With rich data, machine prediction can work well. And we know the prediction is good. This is the sweet spot for the current generation of machine intelligence.”

2. ”Known Unknowns. We know our predictions will be relatively poor in situations where we do not have much data. We know that we don’t know: known unknowns. Predicting a presidential election outcome a few years out is nearly impossible.”

3. ”Unknown Unknowns. If something has never happened before, a machine cannot predict it. For example the Nassim Nicholas Taleb’s ”The Black Swan” or eighteen-year-old Shawn Fanning developing Napster. Both were unknown unknowns.”

4. ”Unknown Knowns. Prediction machines appear to provide a very precise answer, but that answer can be very wrong. You also need to know what would have happened if you hadn’t read this book. You don’t have that data.”

”Shifting to an AI-first strategy means downgrading the previous top priority. In other words, AI-first is not a buzz word — it represents a real tradeoff. An AI-first strategy places maximizing prediction accuracy as the central goal of the organization, even if that means compromising on other goals such as maximizing revenue, user numbers, or user experience.”

CASE: The Tesco’s Clubcard could be a good case example and especially Clive Humby. He has stated that “Data is the new oil. It’s valuable, but if unrefined it cannot really be used. It has to be changed into gas, plastic, chemicals, etc to create a valuable entity that drives profitable activity; so must data be broken down, analyzed for it to have value.”

The third key learning is that prediction machines and humans will work as a team. Those both are needed. A great example would be school bus drivers. Their job might change, because the driving will be automated (in future), but they are still needed as teachers on the bus. ”The point is that automation that eliminates a human from a task does not necessarily eliminate them from a job. Sometimes, the combination of humans and machines generates the best predictions, each complementing the other’s weaknesses. This is a classic division of labor, but not physically as Adam Smith described.”

”Humans have three types of data that machines don’t:

1. Human senses are powerful. In many ways, human eyes, ears, nose, and skin still surpass machine capabilities.

2. Humans are the ultimate arbiters of our own preferences. Consumer data is extremely valuable because it gives prediction machines data about these preferences.

3. Privacy concerns restrict the data available to machines.”

Will Humans Be Pushed Out? No, because in our lifetime ”humans will have a role in prediction and judgment when unusual situations arise.”

The fourth key learning is processes. In order to implement AI technology into different businesses requires rethinking of processes. ”The distinction between AI and automation is muddy. Automation arises when a machine undertakes an entire task, not just prediction.”

CASE: ”One of Hammer and Champy’s favorite examples was the dilemma Ford faced in the 1980s, In North America, its accounts payable department employed five hundred people, and Ford hoped that by spending big on computers, Once a new system was put in place, Ford’s accounts payable department was 75 percent smaller, and the whole process was significantly faster and more accurate. The rise of prediction machines motivates thinking about how to redesign and automate entire processes, or what we term here “work flows,” effectively removing humans from such tasks altogether.” This is a classical robotic process automation case (RPA).

Fifth key learning is AI tools. Steve Jobs have stated that, “one of the things that really separates us from the high primates is that we’re tool builders.” Currently ”Google is developing more than a thousand different AI tools to help with a wide variety of tasks, from email to translation to driving. AI tools can change work flows in two ways.

1. AI tools can render tasks obsolete and therefore remove them from work flows.

2. AI tools can add new tasks. This may be different for every business and every work flow.”

CASE:” March 2016 when Microsoft launched an AI-based Twitter chatbot named Tay. Precisely how Tay evolved so quickly is not entirely clear. Most likely, interactions with Twitter users taught Tay this behavior.”

Sixth learning is about security risks. ”Three classes of data have an impact on prediction machines: input, training, and feedback. All three have potential security risks. For example, you can trick an AI into misclassifying a video of a zoo by inserting images of cars for such a short time that a human would never see the cars, but the computer could. The implications are clear. Your competitors or detractors may deliberately try to train your prediction machine to make bad predictions.

CASE: ”Your competitors may be able to reverse-engineer your algorithms, or at least have their own prediction machines use the output of your algorithms as training data. Google’s team showed that Microsoft uses its toolbar to copy Google’s search engine.”

Minor, but interesting notions:

• Tesla aggregates and uses data from cars to upgrade Autopilot. Learning takes place in the cloud. Only then does it roll out a new version of Autopilot. This standard approach has the advantage of shielding users from undertrained versions. The downside, however, is that the common AI that resides on devices cannot take into account rapidly changing local conditions or, at the very least, can only do so when that data is built into a new generation. Thus, from the perspective of a user, improvements come in jumps.

• You are the product…. A few years ago, users of Internet services began to realize that when an online service is free, you’re not the customer.

• Privacy as a competative advantage… Apple.

• Experience Is the New Scarce Resource.

How should we change according to the book?

Ask yourself that what does it mean to your position when more and more AI tools are rolled out? Elon Musk, Daniel Kahneman, Bill Gates and Stephen Hawking (R.I.P.) are against AI. Or at least they are very doubtful. What’s your stand?

If there would be like a Robotlandia where robots compete head to head ”with humans for some tasks, so wages for those tasks fall. If you understand the benefits of free trade, then you should appreciate the gains from prediction machines. The key policy question isn’t about whether AI will bring benefits but about how those benefits will be distributed. If the competition is with human labor, then wages fall. A second trend leading to increased inequality is that technology is often skill-biased. It disproportionately increases the wages of highly educated people and might even decrease the wages of the less educated.”

Another thing is that new competition will emerge. ”This is not the first time that a new technology raises the possibility of breeding large companies.” From China for example. ”With AI, there is a benefit to being big because of scale economies. Technology-based monopolies are temporary due to a process that economist Joseph Schumpeter called “the gale of creative destruction.” Few facts about China and AI:

1. China’s share of papers at the biggest AI research conference grew from 10 percent in 2012 to 23 percent in 2017. Over the same period, the US share fell from 41 percent to 34 percent. 13

2. One city—China’s eighth largest—has allocated more resources to AI than all of Canada.

3. China has a second advantage: scale. Prediction machines need data, and China has more people to provide that data than anywhere else in the world.

4. Data access is China’s third source of advantage. The country’s lack of privacy protection for its citizens may give the government and private-sector companies

What should I personally do? 

Is this a race to the bottom or next dot-com boom?

Summary

The book in six words – Who will be the shawn fanning of AI?