The document discusses artificial intelligence (AI), including its current state, applications, and future potential. It notes that AI allows machines to learn from experience and perform human-like tasks. Examples of current AI applications mentioned include voice assistants, agriculture, healthcare, and personalized recommendations. The document also discusses how access to vast amounts of data has fueled recent AI advances and how AI can help automate tasks that would be too time-consuming for humans alone.
2. Agenda
What it is and state of play
Looking forward
Applying AI tools to your
organisation
3. “AI is probably the most important
thing humanity has ever worked on. I
think of it as something more
profound than electricity or fire”
Sundar Pichai
4. What Is It? AI 101
Artificial intelligence makes it possible for machines to learn from experience, adjust to
new inputs and perform human-like tasks.
Supervised LearningReinforced Learning Unsupervised Learning
16. Over 500 hours of video
uploaded to Youtube -
every minute
17. This highlights how machine
learning can take the extra
step that humans cannot: by
using data science elements,
Netflix could identify over
76,000 genre types to describe
user tastes: a process that
would have been extremely
lengthy if completed only by
humans.
80%of streaming time
on Netflix is driven
purely through
recommendations
powered by AI.
18. Thread.com use AI and
personalisation to deliver
a bespoke shopping
experience to their
650,000 customers with
a team of just 10 stylists.
36. Not For Everyone, Not For Everything
Although 25% of people like receiving personalized entertainment
recommendations for things like TV shows or music playlists, 39% say they do not.
Do you like receiving personalized
entertainment recommendations?
Great to be with you today
I really want to spend the next few minutes demystifying the world of AI, explaining the power that it will have in shaping the future of business and sharing some practical use cases that you can implement right now, whatever size of business you operate.
We’ll then wrap up, with Q&A, welcoming Prakah from Google onto the stage to also share his insights
I’ve spent the vast majority of time, helping companies leverage technology to unlock new opportunities
First as head of Commerce at O2 bringing new digital solutions to over 2 million businesses,
Then at Google helping companies navigate and leverage Google’s solutions
And now at Adzooma where we’re developing marketing solutions to provide every business with the opportunity to thrive in the digital world
In 2016 Sundar Pichai, CEO of Google proclaimed,
“AI is probably the most important thing humanity has ever worked on. I think of it as something more profound than electricity or fire”
For me, this was the watershed moment, that made governments and companies really sit up and take seriously the defining power of Artificial Intelligence
But let’s back up a little. Before getting carried away, let’s give a run down of what AI is
The term artificial intelligence was coined in 1956, but AI has become more popular today thanks to increased data volumes, advanced algorithms
Artificial intelligence (AI) makes it possible for machines to learn from experience, adjust to new inputs and perform human-like tasks.
There’s a wide range of AI subsets, that we won’t go into, but it’s worth making a short reference to 3 category’s.
Reinforcement learning is a subfield of machine learning in which systems are trained by receiving virtual “rewards” or “punishments”, essentially learning by trial and error. Google DeepMind has used reinforcement learning to develop systems that can play games, including video games and board games such as Go, better than human champions.
Supervised AI - Turning your expertise into an algorithm - this is where the vast amount of AI is being done - labelled data – great way of bottling up knowledge of your organisation
Unsupervised AI - Point a computer at data without a human curating it - unlabelled data - doesn’t measure accuracy - gets an unlabelled data set and will attempt to learn some sort of structure from it
Neural networks are a subset of machine learning techniques. Essentially, they are AI systems based on simulating connected “neural units,” loosely modeling the way that neurons interact in the brain. AI practitioners refer to these techniques as “deep learning,” since neural networks have many (“deep”) layers of simulated interconnected neurons.
This was AI then - stories that most of you are very familiar with
Gary Kasparov, the former world chess champion was beaten by IBM’s Deep Blue in 1997
Just over 20 years later, we’ve got technology that can do this…
This tool built out of Nashville by creative agency redpepper and leverages Google’s open source ML platform to identify Waldo, pointing to anything with a 95% confidence level
But let’s look at some more practical use cases
Voice
71% of AI technology is employed for the purposes of voice recognition, ML and personal assistants.
Think we all have a different Google Home, depending on our queries and demands
Agriculture
Agriculture is no longer as simple as a farmer harvesting their crops. Data is gathered from sources like weather stations and infrared cameras, which allow for key insights on production. Farmers now know exactly how much to feed livestock and can even control their tractors and harvesters remotely.
Healthcare
Some of these cases involve disease diagnosis and improved care through the discovery of new drugs
We could for as long talk about education, marketing and so on and so forth
So what does future of AI look like
Joking aside, this isn’t what this talk is about. Let’s put aside the questions of will it overtake humans and be slightly more optimistic
Mc Kinsey estimate that AI could contribute an additional global economic activity worth around $13 trillion by 2030
First off, it isn’t just AI - it’s a wider cocktail of technological developments
Alongside AI, there are 3 major shifts occurring that are driving technological capabilities
IOT - Estimates show they’ll be up to 30bn connected devices by 2020. And will only accelerate with the advent of smart cities and driverless vehicles. Robotics can also fall into this category
5G - We’re on the edge of 5G rollout, which will be 20 times faster than 4G.
Data - The amount of data we have to play with due to both usage and storage enhancements is mind-blowing
And make no mistake about it, in AI data is Gold
This handsome gentleman is Kai Fu Lee, developer of the first independent speech recognition system whilst doing his PHD in 1988 at Carneigie Melon with around 100mb of data - equivalent to around 5 songs on an mp3 device
Whilst working at Apple in 1990 the data sets he has access to were around 1gb
Today most researchers will train AI systems with a data set of 100tb - an order of 1m larger
Therefore, most of these enhancements have been down to the advancements in processing power.
This is a transistor count chart. The textbook chart for monitoring advancements in computer hardware by looking at he number of transistors engineers have squeezed onto chips
Most people tend to look at a graph like this and see impressive, but steady progress
But this isn’t a typical scale. Notice how the numbers on the Y axix increase exponentially
On a linear scale, the numbers look more like this
Looking at this, it’s hard to imagine massive disruption not on the cusp
Which now makes AI done right, like being in possession of a magical knob
One in which you can make slight adjustments to, to unearth groundbreaking new discoveries
Which has allowed some companies to achieve previously unconceivable progress
Youtube - for instance, uploads over 500m hours of content every minute. Reviewing all that content would be inconceivable without the use of AI in performing the first level of filtration
Netflix - Uses AI based recommendations to drive over 80% of streaming time - and their AI was able to categorise films into over 76,000 genre types
Thread - A fashion start up was able to provide personalised shopping advice to over 650,000 customers with just 10 stylists
But maybe you’re still sitting here thinking why bother…. What relevance does it have to me?
It was first businesses who failed to go online that fell
Next it’s businesses that refuse to adopt AI will struggle
The customer doesn’t care that you’re small! Expectations don’t discriminate
“69% expect businesses of any size to deliver an Amazon-like experience, the bar has been set incredibly high for businesses not already leveraging AI and personalisation to deliver at scale.”
Depending on the size of business and how central AI is to your strategy, there’s really 3 main ways to sprinkle AI onto your organisation:
Build your own
Have someone build it for you
Leverage a platform
In house
Establishing your own team
Might be expensive
Hard to find developers or compete for them
Great for automating lengthy processes in a large organisation
Great if you’re a start up focused on taking on a big competitor
There are open sourced tools you can use to give you a head start
Outsourced
Come with the expertise
Don’t have to find and hire staff
May have worked on a similar project
This is Google’s Cloud API open sourced tool. The backbone of the Where’s Waldo example I showed earlier
Platform
The trick isn’t that you have to build your own platform, but rather leverage existing platforms in smart ways
There’s a cumulative learning you can leverage from using a larger product - hence why platforms tend to be monopolistic
This is something we’re dedicated to building for customers at Adzooma
So with Google, we take in the billions of signals to analyse how best your campaign is performing and provide optimisations that are data driven.
There will be a time where any sort of ad buying will be too complex for a human to partake in
Rather than you being left behind, we want to build an easy to use platform that every business can use to perform at the same level as the worlds largest companies
So we’ve decided we want AI in our business. We now know our options to implement it. Lastly we need to figure out where to put it.
I’d argue there’s 3 key steps you’d want to implement AI into your business .
Marketing
Sales
Customer Service
Marketing
This is really about helping you to scour the world to find the most appropriate customers and deliver to them the right message
Similar audiences
Tailored messaging
Dynamic keywords
Advanced targeting (right message, right time, right place)
Sales
Once you’ve got those customers, how do you make sure they complete a purchase
Combining customer demographic and past transaction data with social media monitoring can help generate individualized product recommendations.
You can display personalised content and offers to each and every individual prospect by analysing their location, device, past interactions, likes and dislike, age, education, job role, etc.
“Next product to buy” recommendations that target individual customers—Amazon
L’oreal - lets consumers submit cheek-swab samples — it then analyses a user's skin health based on the ecosystem of microbes of bacteria, fungi and based on the results, a user can make more personalized product choices.
Customer Service
Finally, how can you make sure you can delight these customers, making them stay for as long as possible or return for another purchase
Targeted emails
Service queries
Enhanced product features
Chatbot - Chatbots can assist customers 24×7 and they can retain customer data. In other words, customers don’t have to repeat themselves with every interaction.
Here are a few examples of the 3 platforms that enable these 3 steps to come to life
Botify - enables you to integrate a bot into your site
Digital Genius - This AI tool can suggest or automate answers, and classify tickets and messages to quickly route them to the right team, freeing up your support agents’ time
Crayon - Enables you you to do better market research and track your competitors activity
Gong - Allows you to track and monitor sales calls picking out winning trends
Phrasee - Will help you write more enticing subject lines
Law Geex - A tool that will automate the create of legal contracts, freeing up time for your lawyers
But finally I do want to say before leaving that you don’t have to paint your whole company with AI overnight. It’s not for everyone and it’s not for everything.
For example, according to Mintel research Although 25% of people like receiving personalized entertainment recommendations for things like TV shows or music playlists, 39% say they do not.
So thank for your time and I’ll be happy to answer any questions you may have
You can also use this link to download a special whitepaper we wrote on AI