I really enjoyed this talk from Benedict Evans of Andreessen Horowitz, sharing some long views of innovation and discussing an s-shaped curve for the adoption of new technologies. For a technology (mobile, for example, or machine learning) he describes an initial phase where people are just trying to make it work, a period of fast development and uptake, followed by commoditisation where more and more uses are found for that technology. He talks about several examples, and places AI / machine learning, crypto-currency, mixed reality (augmented reality is a part of this) and other technologies on the curve to help us understand where they are in their lifecycle.
The key growth technology he is calling out now is machine learning. The technology works, and is available. The number of potential applications is proliferating and there are lots of examples of value being delivered. He talks about 3 aspects of this:
- Automating existing activities, the most discussed but maybe least interesting
- Discovering new things from existing data
- Understanding new sorts of data, such as speech or video
It isn’t a long talk and I would recommend anyone looking for some interesting discussion of innovation, especially in connection with AI, machine learning or robotics, to watch it soon.
OODA is a concept that originated in air warfare. It stands for Observe, Orient, Decide, Act: the process is a powerful one for many classes of problem. The more frequently you can complete an OODA loop, the more disruptive you will be to your competitors. This is called ‘getting inside your opponents OODA loop’ and some have suggested this is the reason Trump succeeded in the 2016 Presidential Election.
In business, it means keeping your plans light and your ability to execute agile and flexible. Amazon should never have beaten Barnes & Noble, but their flexibility, ability to change direction quickly and to execute without long programme and project cycles was enough to disrupt anything the incumbents could do.
I came across the concept in Tim Harford’s book, Messy. There are lots of interesting and disruptive ideas here – recommended reading.
The BBC’s Seriously, Close to the Edit was a fantastic listen on the history of editing, created by Mike Figgis. Given the subject matter, it had to sound great, with some lovely examples of good and bad editing.
Walter Murch, who created the sound design for Apocalypse Now, told a good story about editing pre-digital. Talking about Apocalypse Now, he related how the film’s raw material was spread over 236 hours of film, weighing 7 tonnes, and sometimes he was looking for a single frame (weighing grams) to use in his edit process. Much easier, I imagine, now everything is on digital storage.
The key point the programme makes is that almost everything we experience has been edited before we read, see or hear it: it always pays to question what is really going on.
This article https://www.theguardian.com/money/2017/oct/21/couple-lose-120000-email-hacking-fraud-legal-sector tells the store of a couple who lost £120,000 after they relied on banking details in an email which appeared to be from their solicitors. In fact, their solicitors’ email was a fake, and the couple had sent a tax payment to the bank account of a fraudster. The Guardian reports the couple may never get their money back.
The most important lesson from the story is not ever to trust banking details received in emails. Emails can easily be faked today, though the technology to defeat this has been available for some time. Always confirm bank details through a trusted channel, for example face to face with the intended recipient or in a phone call with someone whose identity you know for sure.
Email non-repudiation technology, which can confirm reliably that the sender is who they say they are, is widely available and proven. It hasn’t gained widespread acceptance because the big technology companies most of us rely on haven’t adopted it and rolled it out. If Apple, Google and Microsoft decided to implement it by default (as we have for many years with SSL / TLS encryption on websites) it would become normal very quickly, and the risk of this sort of crime would fall as a result.
Secure bank transfers would also be quite simple to implement – banks don’t today verify that the recipient’s name on a bank transfer is the same as that of the account the money is destined for. Given that banks have high ‘know your customer’ standards for opening an account, this simple check would reduce the risk considerably. If the transfer says ‘Steed & Steed Solicitors’ and the account is in the name of ‘Graceak Ltd’ it could (and should be returned to sender.
Ultimately, blockchain technologies offer considerable benefits in money transmission for businesses and customers. Traceability of transactions (because of the open ledger technologies used), coupled with strong cryptographic identification of sender and recipient, and robust delivery mechanisms, would greatly reduce the risk we all take when we send money through digital channels.
This is an interesting and very useful talk by Michael Ross of https://twitter.com/DynamicAction
He has captured some strong ideas around the practical use of AI, focusing on Amazon’s successes in applying it, and his conclusions are very actionable.
I was fascinated to hear that Amazon are training managers to manage teams of people and AIs. I wonder what the AIs’ performance reviews are like?
If you’re concerned about how much time you are spending playing with your phone, this article explains how to change the screen from full colour to grey scale. It works on a very fundamental level to reduce the stickiness of the interface, which should reduce the addictive qualities of the device.
Forecasting is hard, whether it’s about the weather, economic performance, sales volumes, or anything else. This article by Dan Gardner, originally in the RSA Journal, makes some very helpful observations on the psychology of forecasting, and on how to get better at it.
We all have unconscious biases. A lot of people remember the things they forecast correctly, but not the rest – this is called the outcome bias. Looking back with a view that everything that happened could have been predicted is also common – this is the hindsight bias. Wikipedia has this excellent list of cognitive biases.
Hindsight bias makes it very hard for people to compare today with the past reliably. In his article, Dan Gardner talks about the recurrence of people looking gloomily at the present and nostalgically at the past, because the things we were anxious about in the past are so readily forgotten. He writes:
“Dig into the contemporary records of almost any year and you will find people worrying about the future and looking back to more certain times.”
A sensible, practical thing we can all do to improve our forecasting, and address these biases, is to track forecasts and their outcome, and to learn from the good and bad outcomes. This is common in science, but a lot less common in business.