Friday, May 25, 2018

Actions speak louder than faux rationality

Photo by taha ajmi on Unsplash
When trying to understand human behaviour our biggest mistake is to seek the rational in what we think or what we believe. 

"Rationality resides in what you do" - Nassim  Nicholas Taleb.

If you ask people why they do something, or even ask them to accurately describe what they do, the answers are clouded by self-justification and self-protection - obscuring the actual. They will tell you what they think they do (or even what they think you want to hear). They will tell you what they believe is accurate.

Observing what they do often reveals significant differences between their belief and the real.

This has obvious and applicable benefit in reducing the risk in innovation. Respond not to what people say they do or think they do, but what they actually do. A-B testing, Design Thinking and Lean Start-up methodologies are all rooted in this.

The faux rationality of conclusions drawn based on our constructs of behaviour and abstractions there-on (replete with our own cognitive biases) is where we reintroduce risk. When this doesn't fail we should far less seek to repeat - than count ourselves lucky. No-one stays lucky forever.

Mathematics does not allow for constructs or abstraction. It demands precisely defined objects and relations. Without which - no algorithm can function.

Nvidia's research into teaching robots to perform tasks by having them observe humans illustrates again how the pursuit of AI is revealing to us what is really rational about people.

Thursday, May 24, 2018

Predicting how long your job will last

If you want to predict the future, look at what has stood the test of time.

When we talk about the future of work - naturally there are going to be new roles. There may be less or different tasks, the latter more likely.

But what you can bank on is that the roles that were here 100 years ago are far more likely to be here in another 100 years than those roles that have been with us for just a few short years.

That's not to say none of the new ones will stand the test of time, but a far higher percentage of the old ones will.

The longer anything lasts - the longer it is likely to continue to last. This is one of the lessons we can draw from Antifragility and other work by Nassim Nicholas Taleb. He would say it is a lesson we can learn from the wisdom of our grandparents.

Will a teacher's job exist in 100 years time? 90% yes. Will a social media strategists? 90% no.

We are often dazzled by the new and make projections into the future on very shallow data. This fails. As AI is proving all over again.

To create value with AI the proposition needs to be reframed in terms of prediction. But unless the correct weighting of what has come before is built into the programming, researchers find they hit the problem they call 'catastrophic forgetting'. The solution is to build in virtual memories (eg Deep Mind's Differential Neural Computer).

For AI to succeed it has to factor for what you and I instinctively know - the longer something has lasted, the longer it will succeed.

The rate of change is so rapid it's difficult for one person to keep up to speed. Let's pool our thoughts, share our reactions and, who knows, even reach some shared conclusions worth arriving at?