But our fear of ‘robots’ isn’t likely to materialise in the way we imagine, it’s happening around us e.g. McDonald’s food ordering module
Machine learning architecture is multi-layered, the challenge is replicating the output that our eyes perceive by passing layered representations onto each stage
Humans are able to take learning from one process and apply it to another
Narrow AI we have made a lot of progress on, e.g. Siri
Strong AI, e.g. Ex Machina, we’re not there yet
The gap is being bridged at the moment by deep learning – the process of getting from narrow, task based AI, to the idea of consciousness and independent ‘thought’
There’s actually potential for strong AI to become better than humans, because it would have an infinite capacity to process and manipulate data
Machines wouldn’t make the same mistake twice, or be influenced by irrelevant external factors, in the way that humans are vulnerable to
The mapping between input and output can therefore become ‘perfect’
This can solve all kinds of world problems; science, information, manufacturing, etc
This can then become proactive, e.g. in content creation, because it now understands the ‘rules’
Bees are extremely intelligent – their learning isn’t deep, they don’t need a lot of layers between seeing and processing in order to understand what they need to do to survive
They can also adapt, taking learning from one task and applying it to another, pattern recognition etc
Support Vector Machine Processing can help us understand input/output ‘rules’
Understanding how to speed up this process, e.g. taking examples from bees, would catalyse our progress in the AI field
Brains sort input in amazing ways, e.g. the cocktail party problem – being able to tonally categorise sounds to differentiate the conversation you want to hear from the ones you don’t
Smell is the same – we can separate the difference between the smell of coffee and the smell of bread, even though they we would perceive that in a bakery we are smelling both at the same time
If we can make a computational simulation of a bee’s brain, we can use it to steer a drone, for example
Simple brains like this are a good way to start as we progress towards strong AI
As people interested in AI and its development, we have a role to play in making AI smarter – we need to understand intent, user patterns, the makeup and delivery of quality content
Insights:
There’s no correct way to build an analytics dashboard – decide what you need by what you want to achieve
Have them updating in real time and be ready to respond as things change
Create a universal analytics dashboard that includes data from all elements of the business, that way you can see how they all interact and make the most of opportunities
Have someone in charge of all of it, who is empowered to make decisions and advocate for changes in direction when needed
Use them in a gamified way to encourage improvements and optimisation internally – who can top the charts, which team can produce the best results
Encourage a shared learning culture around this
It’s worth paying for a dashboarding tool that does everything you need, because it enables this culture and saves money in the long-term
Be prepared to fail a few times before you get it right
Make your customer dashboard an internal product that everyone in the organisation can learn from
Build it up iteratively, get a few things right and go from there
There’s an element of increasing internal understanding of analytics, to drive the internal demand for acting on what’s possible
“Data Studio will become the default digital marketing tool”
Customers are complex – paths to conversion don’t go in a straight line
85bn calls are made to businesses every year
Calls convert 20% better than clicks
1.7 trillion dollars per year is phone call influenced
We need to be able to credit digital with these calls, that is often the end of a digital journey (e.g. calling an estate agent, you can’t add a house to your basket)
A single customer view should include offline methods as well
Can we get to a stage where we know what the customer is calling about before they call, because we know what they were looking at and what they’re likely to want to buy
Speech analytics helps us to make marketing decisions on how products are described by customers
Google predicts voice search will overtake text search by 2019
Agile and big brand strategy:
Aim for what’s currently impossible
Don’t take shortcuts, go for greatness
We have to make search content as good as the best piece of content we can imagine
We need to understand how to market our content better, without crossing the line into becoming a hack, or creating content for content’s sake
We’re contributing to a greater good, we all have a responsibility to provide things that are valuable, meaningful and will change the world or make it a better place
Will Google be able to measure value and contribution to the world in the future, as a quality metric, using ethics as a ranking signal
Create hero pieces that define your brand, promote your values and help you stand out from competitors
Think about advertising as a story, you can’t build a relationship by asking people to buy something before they know who you are
Sequential ads help with ad fatigue, there’s a higher conversion on the third ad
You earn the right to ask them to buy, you need to earn that within the paid social journey
Think about hooks, e.g. before a new campaign
Think about the whole package, the sequences should run at the same time as your email marketing, organic efforts, etc
Make the most of engagement audiences
According to Bing, the CPC on ‘lawyers’ is $109!
Don’t do the same work twice – use your AdWords data to plug into your Facebook ads
Information charities have the advantage of being able to produce thought leadership content
Keynote:
We don’t live in a world of perfect information and perfect trust
The quest for optimisation is difficult in that world
If you become exceptionally good at something, you’re probably sacrificing something else
You need disobedience and unpredictability to keep you adaptable to inevitable change
Nature never purely optimises along one dimension
Never stop doing the left field things, because every now and then you hit on something better
If you try to sell something in the wrong context, you fail, however ostensibly ‘perfect’ the formula
There is no objective reality
In marketing, you have the choice to solve either the reality, or the perception
E.g. TFL – they couldn’t change waiting times, so they changed the perception of waiting, by introducing arrival time updates. People don’t mind waiting if they know how long they’re waiting for
Get people to focus on something else, if you can’t change the reality
Wagamama introduce you to the idea of random food arrival by asking whether you’ve eaten there before. Everything still arrives in a funny order, but we’re ok with it because they’ve sold us on why
“The trouble with market research is that people don’t think how they feel, they don’t say what they think and they don’t do what they say”
Delayed prescriptions are post-dated so people don’t cash them in because they’re generally better before they need them (e.g. solving the antibiotic problem)
It’s easier to get fired for being illogical than unimaginative
Our brains are generating a plausible narrative from everything we perceive
McDonald’s is reliably ‘good enough’ which prevents us from wading into something that could be surprisingly bad and is therefore the biggest food company in the world
We’ll accept a lower average for the trade-off of a lower variable, it’s not optimal, but it’s not catastrophic
This is why we end up doing things ‘the same old way’
This is why big agencies can be uninspiring – they don’t fail, but neither are they likely to do anything amazing. They’re ‘safe’ and if they mess things up, it’s on them, not you
Test counter-intuitive things, because your competitors won’t