This document discusses the coming AI revolution and its implications for marketing. It notes that success in creating advanced AI would be the most significant event in human history but could also pose risks. The text outlines how AI technologies like deep learning, large data sets, and increased processing power are advancing capabilities in areas like computer vision, natural language processing and decision making. It predicts that over the next 3 years, AI will increasingly be integrated into marketing functions like attribution modelling, programmatic advertising, creative optimization, analytics and CRM. The document advises marketers to prepare for these changes by upskilling talent, developing technology stacks, preparing for data-driven marketing and starting to use AI today.
25. Over the next 3 years
YEAR 1 YEAR 3
ATTRIBUTION
MODELLING
(INCLUDING
OFFLINE EFFECTS)
DSP
(DEMAND
SIDE
PLATFORM)
DCO
(DYNAMIC
CREATIVE
OPTIMISATION)
SEARCH BID
MANAGER
DMP
(DATA MANAGEMENT
PLATFORM)
WEB/SOCIAL
ANALYTICS
They will
become
ONE
35. Sentience
THE COMING AI REVOLUTION AND
THE IMPLICATIONS FOR
MARKETING
GEORGE.BETHELL@PHDMEDIA.COM
Notes de l'éditeur
<Comment about SEMPL>
There is a computer today that can conduct over 4,000 complex computations per second. It sits in a museum.
That’s because the smartphone in your pocket can perform over 3.36 billion computations per second
The human brain can perform around 39 thousand trillion computations per second
But the machine is catching up. And it’s catching up fast.
Welcome to Sentience
Hvala, mi je v veselje, da sem tu (thank you for having me, it's a pleasure to be here)
JOKE about excusing my pronunciation.
George Bethell. Global Innovations Director. PHD Global Strategy Unit across Unilever’s portfolio.
Before we start, raise your hand if you know what Artificial Intelligence is?
Raise again, if you feel like you have seen it? That may change after the presentation.
I’m here to talk about AI Revolution & imp for marketing
Sentience = Machines think, feel and do like humans.
Before we get started, imagine for a moment, we were contacted by an alien life force.
Imagine that alien life force told us they would appear on our planet in 15 years from now.
Consider for a moment
How might their arrive it change the world around us?
How might it change the way we interact with world around us?
Most importantly
What should we be doing now, today to prepare for it?
And that is how we should think about AI. It’s going to be so transformative and different, it will radically change the world we live in.
It’s also an area fraught with difficulties
Gloomy predictions, most notably this one from Stephen Hawking at the end of 2014
Elon Musk “We must be careful not to unleash the demon”
Today, it’s 11.59 on AI eve. Scrap that, it’s actually midnight. AI is already here.
But before we look at where we are today, I would like us all to fast forward 13 years to the year 2029.
Once there, we can take a retrospective look at the technologies that got us there..
And then we can start investigating the key drivers
How did you feel when you watched that? Excited? Terrified?
However you felt, get used to it.
Because it’s going to happen.
It’s going to happen fast. Here’s why.
ONE
Very large data sets. With 2.6 billion smart phones can collect more than ever.
With cloud storage systems, we can store more than ever
This is important because AI needs data to train on.
TWO
The ability of algorithms to learn. With increasing sophistications, deep learning algorithms are able to learn independently.
That’s where Artificial Intelligence comes in.
THREE
Finally the exponential increase in computational speed and relative cost.
“Everything that we formerly electrified, we will no cognitise”
PAUSE
Kevin Kelly. We worked with him on a couple of projects at PHD.
I love this quote because it gives life to things we have formerly created to help us.
Object likes self-driving cars, the home, the internet of things, we’re not just discussing the media opportunity.
It’s the data that these new objects will harvest. All connected to the cloud.
It’s only going to get bigger and bigger
90% OF THE WORLDS DATA HAS BEEN CREATED IN THE LAST TWO YEARS ALONE…
2.6 billion smart phones in 2016
6.1 billion expected by 2020
Imagine all the data we collect from Search, social, mobile, CRM, POS, online
Internet of things – fridge, car, home, etc.
Data is really important for AI to train on
Deep learning is a strand of AI
Basically an attempt to mimic the way the brain works
Manhattan – roads are neural paths. Skyscrapers are neurons.
Brain a pattern recognition machine
If we start at ground zero, the neuron it’s able to recognize edges
Then shapes
Then objects
Then Faces
Machines are now better than humans
2015 – developers at Google were successful.
Amazing when you think about it. Because what this means is that we can program machines to think in the same way humans do.
As you move to the top, more complicated patterns can be perceived.
Think of the realms of poetry, music and literature.
Here’s another example of deep learning in action
At team at DeepMind, UK based company bought by Google last year, created an algorithm to play it the Atari 2600.
It learns from experience, using only raw pixels as data input and optimises to the highest score.
After 500 attempts the algorithms are super-human players
These algorithms are getting smarter and smarter, and once reason for this is…
The reason why it’s happening is down to Moore’s Law
Every 18 months
Twice the number transistors in a chip
Which means costs reduces exponentially
Pace grows exponentially
This means really fast processing power.
30 Sequential step = 30
30 Exponential steps = over 1 billion
So, how long until computers have the same power as the human brain?
Imagine this lake represents the full capacity of the brain’s processing power
39 thousand-trillion computations per second
The water represents technology’s exponential growth.
Notice how nothing happens for a very long time and then “click” all of sudden, the lake is filled.
And that’s why we’re seeing so much investment in this area. Follow the money.
$1.7b in 2014
$55b in 2016
All this investment is further pushing AI’s development further down the line.
Read Nick Bostrom’s book called Superintelligence
All this investment is driving AI down the timeline at super fast pace.
Strong AI – the ability to think flexibly across multiple domains, like we do.
Read Nick Bostrom’s book called Superintelligence where our ability to predict the future, breaks down.
AI has become the plaything of Giants.
IBM have already invented Watson – the computer that beat Jeopardy champions Karl Jennings and Brad Rutter – in 2011 and showed the world how far AI had come. Scraping through the 40 million articles and making sense of the semantic data and knowledge base.
Today, IBM want to more than play games, it has invest $1billion into a business unit to commercialise the technology.
They are creating super intelligent assistants to support professionals in the health industry, in New York, AI is being used to help find new cancer treatments, whilst plugged into petra-bytes of medical data.
IBM has invested US$1 billion into a business unit dedicated to commercializing the technology.
Players like Amazon, Google & Apple are shaping the consumer world
This AI train is going to pass right through our backyard
The train has left the station, and there’s no-one there to pull the brakes.
It will radically reorganise us.
It will have a more dramatic impact than the internet.
It will affect not only the way we conduct marketing, but it will shape the way we live our lives.
This needs in a radical reorganization of marketing.
1. AI will shape how consumers interact with the world around them – through assistants
2. AI will impact the marketing machine – using the pools of brand and consumer data
Let’s start with the early forms of AI, here today. Chat Bots.
Chat bot is a computer program designed to simulate conversation with human users.
New way of advertising. Natural, native and conversational.
Early, nebular forms of AI, but importantly working successfully to date.
We can see this in the 40,000 chat bots on Messenger or across WeChat.
WeChat 800M users globally (mostly in China)
Messenger 1BN+ users globally (Apr 2016)
They can perform a whole range of tasks such as
Hail a taxi
Order food delivery
Buy movie tickets
Customize and order a pair of Nikes
Pay your bill
Success = $100M in transactions from chat bots in a single day.
We’ve started to experiment with brands across Unilever
This chat bot was created for a toothpaste brand,
Designed to deliver animated content to parent & child
At night time to encourage brushing at night
Results have been fantastic to date
Best thing is that we are learning from this experiences
Chat bots are early forms of assistants.
Assistants will be more powerful, they will be able to perform multiple tasks, really well.
The Virtual Personal Assistant will be the most exciting tool to shape our world.
It will control your world. It will get to know you so well it pre-emptively make suggestions for you.
It will become the bridge between the consumer and the world around them
Given the pace of change, we are not far from having a VPA is sentient,
It will spend its entire time and focus managing your life.
Through deep learning, it will understand everything. From booking to cancelling, learning your preferences, your likes and dislikes, your location and your diaries.
The infants are already here (Google Home & Amazon Echo)
Move towards more natural language. It’s able to interpret what we say, convert it into text, make sense of the text and it’s meaning, run a search through the cloud, get the correct information and relay back in speech. All in the time it takes to speak.
These systems are only going become smarter. And with suggestions, comes power.
What this means for advertisings is that messages will be created in a completely natural way.
Advertisers will compete via a Biddable advertising model with quality-relevance score and bid-price ratio
As more semantic data is collected,
Deep leaning algorithms become smarter and learn quicker
And computational speed continues to advance.
Anyone know who this chap is?
This is Jibo. Priced at under $500. You can buy this guy by the end of this year.
The important thing to notice here is that this guy is available today.
As these machines become smarter, and more useful, more people will start to use them.
So how can AI help improve the AI machine?
The next three years will blend all the marketing stacks together into one.
Attribution modelling will be determined by AI.
Who sees what, when, in which order, across which device, with which message will almost certainly be decided by algorithms, over-seen by experts.
MERGE
CRM and Media will merge together. This will improve the accuracy of who we target and when.
Improving the performance of our assets.
EMERGE
Creative and strategy will emerge together as the AI is able to conduct a series of multi-variate tests.
PURGE
Instead technology will elevate the role of the strategist
The new role will be to build the technology-stacks for clients
Set the parameters for optimisation
What practical steps can we take to prepare?
I’m going to leave you with 5 key takeaways to implement today
Upskill your talent
SEO, PPC and programmatic buying are the most transferable skills to new AI models.
Preserve them and recruit more. Build Data Analytics and Data Science teams
Optimising to the machine will be the greatest determinant of success.
Connect the critical elements of a brand’s marcoms stack
Data partners,
Agency suppliers
Off line channels
This will enable brands to quickly take advantage of the potential of AI.
For AI to make sense of everything, data needs to tagged.
Like SEO. (lncluding brand meta-data)
AI will decide for us. Data such as ingredients, materials, product source, supply chain, cost
Marketers will need to think hard about what types of product data to tag-up, and what types of offers
make the most sense in different contexts.
Example: locally sourced product offering popular in Singapore.
Positive feedback will be the important signal sentient assistants will look for recommendations.
Important signals for Sentient VPAs when looking to make a recommendation
Maintain positive brand sentiment & High quality products
Address and negative feedback immediately (likely be achieved with AI)
Advertisers will need to future-proof the way they use customer data.
Have a proper DMP strategy, gather all the data from all the points.
AI will mean there might be fewer messages seen, but the ones that do get seen will have been selected based on
extremely specific purchasing and behavioural data about you, and therefore much more influential.
Like every consumer interaction on social, offline, ads your ran, results etc.
You can train an AI bot to learn about every interaction you’ve had with you consumers which will create an AI persona that will represent your brand.
Starting using it today.
Download WeChat. Download chat bots on Messenger.
There’s a great one call “Hi Poncho”
Get used to how they work
As a marketer, try building a chat bots either through your creative agency of APIs.
Amazon Echo Dot retails at £50, but get used to the idea of your conversations in your home being recorded all the time.
Change is upon us. And if we are to prepare for his new alien life force we need to act now to take advantage.
This emerging space is going to be one of the most remarkable industries to work in
a true blend of technology, science and creativity.