At Rise of AI 2017 in Berlin, Fabian Westerheide, founder of the conference, gives the opening keynote from the venue hosting a rebuild of Konrad Zuse's Z1, the first programmable computer built in Europe. He recounts how Rise of AI grew from a 15-person meetup at the c-base hackerspace to over 300 attendees, and defines artificial intelligence as algorithms that use data to make and improve their own decisions within given frameworks. He walks through AI already in use, including IBM's Deep Blue and Watson, AlphaGo's win over Lee Sedol, Cortana, Amazon Alexa, self-driving cars, AI-run stock markets, and younger AI companies such as Berlin's Parlamind and Micropsi Industries and London's Exlo Dynamics. He then looks ahead to brain-computer interfaces and an emerging internet of agents by around 2025, describes new jobs such as AI trainer and AI controller, and closes by arguing that without a strong European AI ecosystem, AI will end up thinking American or Chinese rather than reflecting European values.
Key takeaways
- Rise of AI grew from a 15-person meetup at Berlin's c-base hackerspace in stages to over 300 attendees with a waiting list by 2017.
- AlphaGo beat Lee Sedol at Go just one year after Nick Bostrom had estimated in Superintelligence that it would take humanity another decade.
- Berlin's Parlamind and Micropsi Industries, and London's Exlo Dynamics, were building AI for email response, robot learning, and drone swarm control respectively.
- Westerheide expects new jobs such as AI trainer, AI controller, world creator and community manager as AI takes over more routine work.
- Westerheide argues that without a strong European AI ecosystem, AI will end up reflecting American or Chinese values instead of European ones.
Detailed notes
From Zuse's Z1 to Rise of AI
Westerheide opens by noting the venue also houses a rebuild of Konrad Zuse's Z1, the first programmable computer built in Europe in the 1930s and 40s, which had 200 bytes of working memory and 5.3 hertz of processing power. He recounts starting Rise of AI as a small meetup at Berlin's c-base hackerspace with 15 people discussing whether a singularity would happen, growing to 70, then 180, and now over 300 attendees with a waiting list, which he reads as AI having reached a critical mass of public interest.
Defining artificial intelligence
Because language is imprecise, Westerheide defines AI as a set of algorithms that use and generate data to make their own decisions and improve within a given framework. He compares algorithms to cooking recipes, and frameworks to the rules and motivation that guide employees or citizens: AI systems process input, use memory, make decisions, and learn by improving, all within rules that humans set for them.
AI already among us
He surveys existing AI systems: IBM's Deep Blue beat the world chess champion in 1996 and 1997, and Google's AlphaGo beat Lee Sedol at Go, a game Nick Bostrom had estimated in his 2014 book Superintelligence would take humanity another ten years to solve, achieved one year later. IBM Watson, using 90 servers with 16 terabytes of memory, won Jeopardy and now runs as a platform for thousands of startup applications. He points to Cortana and Amazon Alexa as everyday AI, self-driving cars that share collective learning across a fleet, AI-run stock markets, and Google's shift to an AI-first strategy across Maps, AdWords and Search.
Younger AI companies from Berlin and beyond
Westerheide highlights younger companies as proof of European AI talent: Parlamind, a Berlin startup that drafts and can auto-send email responses for customer support; Exlo Dynamics in London, which built mission-control software letting a single operator manage large drone swarms by setting premises rather than piloting each drone manually; and Micropsi Industries, based in Berlin, which builds AI that lets robots learn and adapt to a process instead of being explicitly programmed, which he cites to show Berlin's own innovation talent.
Brain-computer interfaces and the internet of agents
He describes existing brain-computer interface headsets that read electrical signals to control computers or let paralyzed people type, and cites Elon Musk's Neuralink and a Facebook research device, expected around 2019, aimed at letting people type by thought, developed by a researcher formerly at DARPA. Looking to around 2025, he predicts an internet of agents in which every app becomes an AI agent that fulfills a person's needs on command, first by typing, later by thought, and describes AI extending into virtual, augmented and physical worlds, including satellite-guided agricultural machinery.
Future jobs and the case for European AI
Westerheide argues AI will not replace people outright but let companies produce the same output with fewer people, and describes emerging roles: AI trainers who teach systems how a job should be done, AI controllers who oversee large numbers of automated systems, world creators building virtual and augmented realities, community managers, and consultants who explain change to others. He closes by warning that without a strong European AI ecosystem, AI trained mostly in the US or China will carry American or Chinese values rather than European ones, which he says is the reason Rise of AI exists.
Quotes
Artificial intelligence is for me a set of algorithms which is using generating data to make their own decisions, to improve within given frameworks.
Fabian Westerheide
We don't have a strong European ecosystem for artificial intelligence, that is why we host this conference.
Fabian Westerheide
Artificial intelligence is just a tool for us to be more human.
Fabian Westerheide
Questions answered in this talk
How does Fabian Westerheide define artificial intelligence?
According to Fabian Westerheide at Rise of AI 2017, artificial intelligence is a set of algorithms that use and generate data to make their own decisions and improve within a given framework of rules and motivation set by people.
How did AlphaGo's win compare to expert predictions?
According to Fabian Westerheide at Rise of AI 2017, Nick Bostrom had estimated in his 2014 book Superintelligence that it would take humanity another ten years to solve the game of Go, but Google's AlphaGo beat Lee Sedol just one year later.
What new jobs does Fabian Westerheide expect AI to create?
According to Fabian Westerheide at Rise of AI 2017, expected new roles include AI trainers who teach systems how a job is done, AI controllers who oversee automated systems, world creators building virtual and augmented realities, and community managers.