Seldon at the All Party Parliamentary Group on Artificial Intelligence
On 20th March 2017, Seldon was invited to speak in Parliament at the first APPG on AI on the question “what is AI to me?”. Here’s what I said.
This time last year British artificial intelligence company DeepMind challenged the Go world champion.
And in the end, the computer won.
The reason people went crazy about this, is not only because the computer won. We’ve seen computers beat people in Chess and Jeopardy with brute force – using processing power to try every possible move.
But it’s in the fact that there are more possible board combinations in Go than there are atoms in the universe – winning requires strategy and creativity.
AlphaGo played moves that the professionals had never dreamed of.
And harnessing AI will drive the 4th industrial revolution.
Science fiction writers – from Mary Shelly to Stanley Kubrick – have dreamed about creating intelligence for decades. In 1965, Irving J Good, Bletchley Park cryptographer, hypothesised that “the first ultraintelligent machine is the last invention that man need ever make”
And long before the recent obsession with AI in industry, the academic world has been openly researching and publishing papers on techniques such as deep neural networks that are responsible for some of the most groundbreaking recent developments.
AI can now fulfil its potential in the real world due to open-source AI software frameworks, massive increases in the availability of data and compute power. We’re entering an era of quantum computers that are 100 million times the speed of ordinary computers.
The All-Party Parliamentary Group on Artificial Intelligence (APPG AI) was created with the purpose: to unpack the term, to gather evidence to better understand it, to assess its impact, and, ultimately, to empower decision-makers to make policies in the sphere
Our Purpose
Machine learning is everywhere in our lives.
It recommends products online, removes spam from your inbox, and decides which of your friends’ status updates you should read.
Computers have already acquired superhuman abilities across hundreds of new and very focused domains.
But machine learning can also help us tackle some of the world’s biggest problems like
- Drug discovery and image diagnostics in healthcare.
- Predicting crop yields in agriculture.
- Increasing power efficiency – from data centres to the national grid and soon smart grids.
I’m the founder of a start-up called Seldon. Our purpose is to enable people to predict and shape the future with machine learning.
Seldon’s open-source machine learning deployment platform helps data science teams solve problems faster and more effectively.
Currently only a couple of percent of machine learning models move from R&D to production, but over 50% of businesses now want to build their own models in house.
We now have a global community of thousands of developers coming to Seldon for their machine learning building blocks. Including some of the world’s largest companies like Hewlett Packard and Barclays. We were part of the Techstars Barclays Accelerator fintech program – a shining example of large enterprise driving innovation though collaboration with tech start-ups.
What is AI?
So, what is AI?
Artificial Intelligence is an umbrella term that includes everything from natural language processing to automation.
There has been significant amount of progress in recent years, but researchers are still a long way off developing true artificial general intelligence (AGI) with agents that demonstrate human-level intelligence or super intelligence that goes way beyond.
Most of the successful applications of AI today are in fact based on machine learning, which allows computers to solve very specific – or narrow – problems at a superhuman level.
Here’s how it works:
In the software development approach human programmers solve problems by combining rule-based programming with data to produce some output.
Machine learning flips this traditional programming methodology on its head. The raw ingredients for machine learning the expected output and some data set – for most use cases labelled data is necessary for what we call supervised learning.
Sourcing the a clean labelled dataset is a challenge that many organisations underestimate.
The data is fed through one or more algorithms to train a model. A model is effectively a computer program designed to do one job well. Generally classifying, grouping, predicting or recommending things.
Deep learning is a part of machine learning inspired by neuroscience. Just as paths strengthen between neurons in the brain as people learn and observe the world; though the training process, connections between the nodes of the neural networks strengthen – and there are levels of abstractions across each layer of nodes as the network figures out its own rules for pattern recognition.
Neural networks have proven themselves have a higher accuracy than statistical approaches, but the results are harder to explain due to the levels of abstraction and complex relationships between the nodes.
In reinforcement learning, agents interact with virtual or real-world environments, learning from a reward function just as people learn in the real world.
Post-jobs world
Previous rounds of automation replaced muscle power with machines, now machines are replacing our cognitive abilities.
According to an Oxford study, almost half of jobs are likely to be replaced by cognitive machines within the next 15 years.
More than 850,000 public sector jobs could be lost by 2030 through automation.
But there’s a shortage of the very people who have the skills to build machine learning models. There will be a $30bn excess demand for data scientists by 2018 in the US alone.
Are we moving to a post-jobs world with large parts of society living off of some universal basic income or will AI create new jobs?
I believe that new jobs will be created, but these must be matched by policies that prevent a polarised post-jobs economy. It’s also important that people understand what they can do to adapt such as learning new skills.
We’re all living longer and machine learning can deliver the economic efficiencies required to help offset the impact of the aging population on the economy.
What is AI to me?
AI is about using machines to solve new problems. It’s about automating and augmenting our decision making. It’s about making the world a better place. Embracing AI is one of the biggest opportunities for the UK economy at this time.
But there is a short window of opportunity to build upon the great work that has come before with DeepMind and others. It is critical that the UK fulfils its potential to become the centre of excellence for AI that attracts the world’s smartest researchers and entrepreneurs.
As a founder I call upon the lords, ladies and gentlemen here today, the people of influence in government and industry to consider the once in a generation opportunity that AI presents. AI is a positive revolution that will impact every aspect of our lives.
Thanks to open-source, many of the core technologies are readily available for anyone to use and build upon.
We’re in a time of exponential change that calls extraordinary collaboration between government, enterprise and start-ups to:
- Find ways to protect and streamline the visa system to allow global talent in AI to work in the UK.
- Make AI & machine learning a core part of the UK’s curriculum on both a formal and informal basis.
- Provide additional support for research-based AI companies to play the long game.
My name is Alex Housley, we’re Seldon, and we’re building the foundation for you to create the 4th industrial revolution. Come join us!
Follow the APPG AI
The APPG AI Secretariat is Big Innovation Centre — special thanks to Birgitte Andersen for organising, and Co-Chairs Stephen Metcalfe MP and Lord Clement-Jones. Supporters include Barclays, Deloitte, KPMG, PwC, Olswang, BP, EDF Energy and University of Oxford Computer Science Department.
There have been two further APPG AI meetings on more focussed topics including ethics and the future of the office — many more topics will be covered through 2018. APPG AI is temporarily on hold with all communication suspended until after the General Election (this post is not official communication). To learn more please follow @APPG_AI, and I look forward to participating in future meetings and posting updates on this blog.
