Keynote: Not Knowing #KMWorld

Speaker: David Weinberger, Author, Everything is Miscellaneous, Too Big to Know & his latest, Everyday Chaos: Technology, Complexity, & How We’re Thriving in a New World of Possibility

Session Description: How We’re Thriving in a New World of Possibility Through stories from history, business, and technology, philosopher and technologist David Weinberger finds the unifying truths lying below the surface of the tools we take for granted–and a future in which our best strategy often requires holding back from anticipating and instead creating as many possibilities as we can. As a long-time KMWorld magazine columnist, Weinberger has often shared his views of knowledge flows and knowledge sharing as well as the technologies enabling transformation. In this talk, he helps us understand the possibilities that machine learning and other forms of AI are creating and how to harness the power of these breakthroughs to improve knowledge flows in our organizations.

[These are my notes from the KMWorld Connect 2020 Conference. Since I’m publishing them as soon as possible after the end of a session, they may contain the occasional typographical or grammatical error. Please excuse those. To the extent I’ve made any editorial comments, I’ve shown those in brackets.]

NOTES:

• Dominant technology forms us. The Net conditioned us to chaos. AI now shows a way to make sense of the chaos.
• The Internet
  • Before the Internet, most of us followed the “paleolithic” strategy: You plan for a future event and then take action.
  • The Internet Effect — For the first 19 years of production, Henry Ford sold the Model-T car without making many visible changes. He anticipated his customers’ needs so well, he did not need to change his product. By contrast, Dropbox launched with a minimal viable product and committed to make as many changes as dictated by the market. Dropbox did not anticipate the future and prepare for it the way Henry Ford did. Rather, Dropbox simply followed its customers.
    • The iPhone launched with a basic phone. The key to future sales was the App Store, which allowed others to increase the functionality of the phone.
    • Minecraft similarly allowed others to extend the game in new, and sometimes unexpected, directions.
  • The Internet is built on the principle of interoperability. So that allows an enormous number of unpredictable possibilities.
  • The final results could never have been anticipated by the originators of the tool. In fact, they didn’t want to anticipate. They wanted to watch and learn.
• Two Opposing Strategies
  • The paleolithic strategy assumes that the future has an immense number of possibilities and our best approach is to narrow our options, choose our path, plan and act.
  • By contrast, the post-Internet environment is about widening the possibilities in the future. So you start narrowly and then enable a broadening of options in unpredictable ways.
  • The impact of abandoning the paleolithic strategy is that we have organized our lives to make life more unpredictable.
• Machine Learning
  • Machine Learning — gives us a way of understanding the chaos we have created.
  • Traditional programming asks a developer to predict the relevant factors and then implement the logical relationships among these factors using code.
  • Machine learning consumes the data (the underlying factors) but is not given the logic regarding the relationships among the data. Instead, the machine iterates until it surfaces relationships. [This leads to results beyond the knowledge or planning of the developer.]
• The Black Box Problem
  • The earth itself is a black box. We don’t fully understand how it works.
  • That said, while we cannot understand how a black box system made a decision, we can test that decision to determine how correct those decisions prove to be.
  • The natural state of machine learning is to be a black box. This is a problem because machine learning systems are based on data from a culture or an organization. These data reproduce the biases in that culture or organization and then machine learning amplifies those biases. Unfortunately, this process is hidden from our view because of the black box nature of machine learning.
  • The black box nature of machine learning is resulting in a moral panic because “We don’t know how it works!” But the deeper anxiety is that “It works!” It produces results beyond anything humans can produce. Machine learning systems work in such incredibly complex ways that we cannot replicate without machines.
  • Our brains made machine learning systems but our brains are insufficient to comprehend our creatures and their work.
• Five Ways We Can Respond.
  • #1. Strategies are outdated.
    • Plato was the first to separate strategy and tactics. Tactics were comparable to what we call logistics. Strategy is more like improv or making music.
    • They require some stability for planning and execution. In an environment of unpredictability, strategy is less effective. For example, look at The Black Swan and Rita Gunther McGrath’s work that tell us that strategies are too broad and lead us away from a more productive, narrower focus. So we need to shift our thinking to a Minimum Viable Strategy, which is the minimal amount of planning necessary to allocate our resources sensibly.
  • #2. Rethink Corporate Knowledge Flows.
    • Corporate hierarchies were created to filter information until the ultimate decision maker had just the right information to make the right decision. Therefore, people lower down in the organization spend their time discarding whatever they consider to be irrelevant information. With the current explosion of information, this means that they need to ramp up their exclusion tactics and may, in the process, throw out valuable information.
  • #3. It takes a Network to Make Sense.
    • The smarter approach is to set up a series of sensors who are able to notice and report on butterflies (the key insights that are valuable). The wider and more open the network, the more effective it is. They also need to be rewarded.
  • #4. Make more possible.
    • Increase open source and access. Increase the opportunities for learning in public. Make it interoperable.
  • #5. Reify Knowledge.
    • To reify is to make something abstract more concrete or real.
    • Reify knowledge means “turn knowledge into a thing.” Although everything is constantly changing, we have believed that there are universal laws that govern and survive. This is where truth resides.
    • Now truth resides in linked open data.
    • Models are the new Body of Knowledge.
      • Traditionally, knowledge is content. But a software program is not content. AI models are not representations (in the same way) of knowledge. They learn by being used. This is new for a body of knowledge — it is a body of knowledge that can “eat” and grow.
      • Before machine learning, a body of knowledge (a domain) could only be extended by human effort.
      • There a price to this efficiency. We may end up reifying bias. Because a body of knowledge becomes a thing, it can be owned. And accessed can be limited or made expensive.
• Chaos is the Truth.