At Rise of AI 2016 in Berlin, an event for 100 selected guests, host Fabian Westerheide opens the evening by asking what the singularity actually means. He recounts a novel in which an email company's team builds a smarter reply feature using natural language processing and neural networks, and notes that Google shipped a comparable real feature, Smart Reply, in 2015, arguing that fictional AI technologies tend to become real within one to three years. Drawing on the blog Wait But Why, he offers three definitions of the singularity: computers matching or exceeding human performance, humans losing control of their own creation, or people simply believing a machine is superior to them. He argues humans build AI because they are lazy and want more output for the same effort. He then presents 21 jobs the audience named in advance as most at risk from automation, grouped from low-paid routine roles to professional, analytical and office jobs, while arguing that work built on emotion, free thinking and creativity is safer because science does not yet understand how those capacities arise.
Key takeaways
- Westerheide argues AI technologies typically move from science fiction to real products within one to three years, calling them 'science facts.'
- He defines the singularity three ways: matching human performance, losing control of AI, or believing machines are superior to humans.
- He argues humans build AI because they are lazy and want more output for the same effort.
- Audience input produced 21 frequently mentioned jobs at risk, grouped into low-paid roles, professional roles, analytical roles and office jobs.
- He argues emotional, creative and independent-thinking jobs are safer because science does not yet understand how creativity or feelings work.
Detailed notes
Defining the singularity
Westerheide opens with a story from a novel about an email company whose team builds a feature to make replies smarter using natural language processing and neural networks; in the book, the system starts to improve on its own and becomes the first AI. He notes that in 2015 Google shipped a comparable real feature, Smart Reply, which suggests replies to emails before the user has read them, and says such technologies tend to reach the real world one to three years after appearing in fiction, calling them 'science facts' rather than science fiction. He then poses three possible definitions of the singularity, drawing on the blog Wait But Why: the moment computer performance meets or exceeds human performance, illustrated by dependence on pacemakers or smartphones; the moment humans lose control of their own creation, as in Hollywood depictions; and the moment people believe a machine is superior to them. He argues machines already exceed humans physically and in several cognitive tasks such as chess and driving.
Why build AI at all
Asked why humans create something that could be a threat, Westerheide answers that humans are lazy: throughout history, tools and inventions have served two purposes, working less by having something else do the work, or using the same effort to produce more output.
Which jobs are at risk
Westerheide asked the audience in advance which jobs they thought AI could replace and received hundreds of answers, which he condensed into 21 frequently mentioned roles across four groups: lower-paid, more routine jobs such as cab drivers, baristas and security guards; professional roles including teachers, lawyers and surgeons, illustrated by software that can recognize brain cancer in under a second, a skill that takes a surgeon 20 years to learn; analytical roles such as investment analysts, financial and stock traders, and data scientists; and office jobs centered on entering and processing data. He says jobs built on emotion, free thinking and creativity are safer for now, because science does not yet understand how creativity or feelings arise well enough to replicate them. He closes by asking whether half of today's jobs will eventually be done by machines, and what people will do with the resulting free time.
Quotes
We humans are lazy.
Fabian Westerheide
Will it come to a phase where fifty percent of the jobs which we have today will be done by machines?
Fabian Westerheide
Questions answered in this talk
What are the three ways Fabian Westerheide defines the singularity?
At Rise of AI 2016, Fabian Westerheide describes three possible definitions of the singularity: the moment computer performance meets or exceeds human performance, the moment humans lose control of their own creation, and the moment people believe a machine is superior to humans.
Which jobs does Fabian Westerheide say are most at risk from AI?
According to Fabian Westerheide at Rise of AI 2016, jobs most at risk include cab drivers, baristas and security guards, professional roles such as teachers, lawyers and surgeons, and analytical roles such as investment analysts, financial and stock traders, plus office jobs built around data entry.
Which jobs does Fabian Westerheide say will not be replaced by AI?
At Rise of AI 2016, Fabian Westerheide says jobs involving emotion, free and independent thinking, and creativity are unlikely to be replaced, because science does not yet understand how creativity or feelings arise well enough to program them.