From online-first to mobile-first, we are currently entering the phase of AI-first. Artificial Intelligence is the consequence of an online, digital, connected, mobile, data-driven, computerized world. Let me explain why.
The 5 Main Drives for the Rise of Artificial Intelligence
There are five main drivers supporting the current Rise of Artificial Intelligence.
The more data the better for Artificial Intelligence
Nowadays almost everything is based on data. We humans measure, store and analyse everything imaginable. Every machine is based on data, with everything ultimately reducing down to a zero or a one.
The amount of data we produce as humans and machines is increasing every year. The more data we produce, the more of it we can feed into Artificial Intelligence systems. Why? Because we need it. The more data generated, the more complex our environment gets. Artificial Intelligence helps us to reduce some of that complexity.
Cloud solutions have started a process in which data is no longer stored on just one physical server, but is accessible everywhere at all times. This means the data can be harnessed more easily and used by other software, including AIs. Some call it the API economy, and even now Machine-to-Machine (M2M) communication already creates more data than human-to-human communication.
Artificial Intelligence is embedded into hardware
Chips are the brains of computers and they are getting a lot smarter. Google builds its own AI chips, as do Microsoft and Nvidia. The special chips they create are suitable for running neural networks and other machine learning tools which can support smarter AIs.
Soon these chips will meet the Internet of Things wave. AI-strengthened chips will be integrated into mobile phones, tablets, cars, drones, vending machines, robots and TVs. We can expect an intelligence upgrade for electronic items.
Software becomes smarter and hybridized
Every day there are really smart people out there improving on tools to create Artificial Intelligences. We are also starting to see more and more hybrid systems which combine expert systems, machine learning, bots and deep learning.
Deep learning has been around for 15 years, but has only come to the public’s attention in the last 3 years. There are therefore still a lot of findings and research results to be analysed and applied to real-world problems.
Universities researching machine learning and cognitive systems are now getting more attention, funding, clients and applicants. The academic world is awakening from its hibernation.
It is going to be easy to build your own Artificial Intelligence
You want to build your own AI one day? Deepmind has opened their own open source lab. OpenAI offers Universe for training of AIs. Google also shares Tensorflow. You can also use Theano and Torch.
There are already plenty of tools and soon there will be even more. Day by day, creating, building and training Artificial Intelligence systems will become easier and easier until we reach a democratization of intelligence creation.
Capital follows the entrepreneurs into AI
Over the last two years we have seen a surge in AI companies. Not everything which is labelled AI contains an AI inside. Maybe 90% of the companies are applying machine learning, but they aren’t building a self-improving, cognitive system.
Nevertheless, the trend is very positive. And where the ideas are, the money follows. Americans VCs have poured billions of dollars into AI startups.
Google, Twitter, Intel, Apple, Microsoft, Salesforce, Facebook, eBay, Oracle — they have all recently purchased young AI companies.
Consequence of the Rise of Artificial Intelligences
New technologies always bring about change.
Four steps towards Artificial Intelligence
There are four steps for a company to reach Artificial Intelligence.
Firstly, data needs to be collected, stored and analysed. This is outdated thinking and tells you something about the past.
Secondly, machines start to make predictions based on the data. These predictions help humans to make faster, easier and better decisions.
Thirdly, machines make predictions, execute them, measure the results and then change the inputs and constraints to optimize the output goal.
Fourthly, machines achieve automation. They are becoming motivated cognitive agents and are able to use their various learned behaviours to create new transferable knowledge.
Most (old-economy) companies are at stage one or below. They know they have data and start collecting it.
Many digital companies (ecommerce, mobile, gaming) work with their data and are often on stage two.
Google, Amazon, Facebook & Co are leading the applied AI fields. Some parts of their companies are on stage three.
Stage four will come. Some brilliant teams are working on it.
You have to train an Artificial Intelligence
Artificial Intelligence systems have to be trained. It’s no longer a case of programming them and thinking they’re finished. An AI is never finished. There is always more it can learn, get better at, improve and ultimately deepen its knowledge.
Modern AI is like a child. You have to teach the AI everything: how to understand text, how to watch videos, how to listen to audio or how to generate language.
Some fields are easier to learn, like computer vision. Other tasks are much harder, such as understanding text.
Artificial Intelligence is eating your business
The new wave of AIs, often implemented by digital companies and start-ups, will eat many smaller businesses for breakfast. AI is a horizontal technology. It is impacting many industries (logistics, automotive, pharmaceutical, insurances, media, manufacturing, retail), systems (networks, cities, states), companies and humans.
In general, AI achieves two things. It makes processes more efficient and it makes stupid machines smarter. A self-driving car is smarter than a human-driven car. Using the Google search algorithm is more efficient than going to the library.
A future with less human work
However, this leads to several challenges.
One challenge is that applied AIs will drive companies out of business, force political change and force individual humans to adapt.
The other challenge is that millions of tasks will be done by machines. New jobs will be created (e.g. machine trainers), and there will be less demand for human labour. Why? Because that is why we invent machines in the first place (watch my TEDx talk to learn more or read 22 jobs disappearing in future); so that they work for us.
Artificial Intelligence is neither evil nor bad. It is a tool we use to generate more wealth, more happiness and more health for humans on earth. But change is never comfortable. AI forces the human species to leave their current comfort zone.
AIs brute force experiences
So what else can we expect? AI systems are already developing their own languages and encryption methods. Other AIs are able to generate images and videos. There are plenty of exciting things coming to the market over the coming years.
We can assume that some technical problems will be overcome: unclean data, unstructured data, limited access to data and biased data.
Furthermore, machines struggle with high degrees of abstraction during their learning phases. Humans today still learn more efficiently than machines. To achieve similar levels of intelligence, we have to train machines with significantly more data. Google’s AlphaGo was only able to beat the human world champion at Go because the AI system practised by playing millions of games. Since AIs can learn faster and in parallel, they brute-force experiences.
Another academic challenge is to build cognitive systems which can generate transferrable knowledge. This means combining different modules of an (artificial) brain.
AI is good for humankind
Sure, there is more: AI-assisted teaching, the jobless future, income equality, personalized medicine, self-moving objects, personal assistants for everyone.
The wide application of Artificial Intelligence will lead to prosperity, lower energy costs, more mobility, free education, longer and healthier lives and increased luxury.
Currently, AI is a bit overhyped because Hollywood, Techcrunch and Journalists find it easier to write about AI than other topics. Nevertheless, the underlying trends are strong. AI won’t go away. We should look forward to and embrace it.
If you would like to have a full day of great insights into Artificial Intelligence, then join our conference Rise of AI May 2017 in Berlin. If you just like to talk, ping me via twitter. If you look for capital, then check out Asgard — human VC for AI.
