A Deep Dive into Knowledge Automation

The Opposite of Observation

A common mistake that scientists make is to believe that knowledge comes from observation. Startups that struggle are often led by a CEO with the opposite thinking.

On the Shoulders of Giants

Terence Tao, a Fields Medallist, says that AI does not replace the human creativity required to generate original ideas. AI is a co-pilot with great potential to accelerate cooperation, testing and the advancement of knowledge. But it does not create.

This echoes the thinking of Karl Popper and David Deutsch. For Popper, all knowledge starts with conjecture, before being subject to criticism and testing. If a theory is robust enough to survive and improves on previous understanding, it is accepted.

Deutsch developed many of the proofs for the quantum theory of computation. The brain is a computer and subject to the same physical processes and laws as any other.

This means there is no reason why artificial general intelligence won’t surpass the human brain. The obstacle is understanding conjecture.

We do not know how we come up with new ideas. Hence we cannot yet make models generate them.

From Scientific Method to Process

The Scientific Method was developed to throw off the doctrines of the monarchy and church. Knowledge must be based on observation, not precedent or scripture. That was fine in the 18^th century.

Nowadays we have knowledge of plenty we cannot see, from sub-atomic particles to far distant galaxies. The CERN physics laboratory searches for particles we know exist but have not yet found. It discovered the Higgs Boson in 2012 after a 40-year search.

The Scientific Process is to conjecture a theory, subject it to criticism and test. This is how businesses should build.

Attractive and Affordable Tools

The mistake founders make is to stop at conjecture. They believe they are able to will an idea to be successful. Legends have grown around founders who dreamed products that sold themselves. But these are just legends.

A prime example is Apple’s iPhone. Steve Jobs released this when the mobile was established, the iPod music player successful and internet access in demand. His skill was combining these ideas in a beautiful product. Nokia, Motorola and Blackberry did not see past their own ideas and collapsed.

The lesson is listen to criticism and test the market. Start with few features and build on them. In 1985, Jobs said of the use case for personal computers,

“I can only begin to speculate. We see that a lot in our industry: You don’t know exactly what’s going to result, but you know it’s something very big and very good.”

Jobs knew handing people attractive and affordable tools that are easy to use unleashes creativity. He didn’t know what and we don’t know how.

Use Your Own Understanding

This understanding of knowledge creation is important when adopting AI. A quarter of SME leaders say AI is too complex or they don’t understand it. That’s just the beginning.

If you buy third party tools you purchase someone else’s understanding of what AI should do. You can ask interesting questions but the responses depend on the training data and method. The answers are determined by statistics and hence subject to error.

If you train models on your own data, you have control. You teach AI to repeat processes and scale your business. The models research, report and reconcile the way you do. Just much faster and cheaper.

Automating processes may be accelerated by selecting an open source model pre-trained in a particular field. For example, on the rules of clearing houses or stock exchanges. This model is then fine-tuned with your workflows that use those rules.

The result is a small model with expert knowledge that is cheap to run. If it makes a mistake it’s because of probability, not because it’s gone rogue. If it’s trained on internal methods and data, the risk is low.

Opening up Opportunities

AI opens up opportunities to simulate product launches, estimate market demand and train staff to handle objections and difficult customers. There is even less reason to trust intuition alone.

I’ve started working with a business supplying AI as a service. The offer is a self contained series of models, secure on clients’ servers and trained in specific tasks.

If you’d like to know more reply to this email or contact me via simonmaughan.com.

I’m Simon Maughan and I write The Profit Elevator as a guide for B2B firms seeking faster growth. The P.R.O.F.I.T. Through Process Planner is my programme to design, develop and automate 30 processes that scale small and medium sized businesses.

If you found this letter valuable, please share it with a colleague.