The documentary about AI that teaches you nothing about AI
Over the weekend people could stream the documentary ‘Do you trust this computer?’ by Chris Paine (Who Killed the Electric Car?).
This post first appeared as part of Curious Robot’s newsletter. It collects stories and news about the ethics and impact of technology. Interested? You can sign up here.
The documentary shows a side of artificial intelligence (AI) focused on killer robots, autonomous weapons and artificial intelligence taking over the world. This view has been so dominant, that it starts to hurt.
Artificial intelligence is a wide field of new types of computers that are trained rather than programmed. They help with problems that are too complex to write a computer program for like autonomous driving, image recognition or language processing. Many forms of it already exist.
Superintelligence is a form of AI that will far surpass the human mind. It is considered our last invention, because it will be able to invent things better than people. It does not exist (yet). Some AI researchers argue it will exist in 50 years, others say it will take much longer or even believe it will never exist.
What’s the message?
The main message of the documentary comes at the end: ‘The pursuit of artificial intelligence is a multi-billion dollar industry with almost no regulations’. Yet Paine does not ask the featured CEO’s and experts what role they are playing in it, nor does he push them to explain how they are ensuring this apocalyptic world doesn’t become real.
For example, Paine shows Rana El Kaliouby, cofounder and CEO of Affectiva. Her company is developing AI to recognise and respond to human emotion. In the documentary she’s seen interacting with toddlers playing with a toy controlled by facial expressions. However, the documentary doesn’t mention the company creates the technology to make ads more effective.
Killer robots, drones and human extinction
The conversation about artificial intelligence is completely dominated by killer robots, autonomous weapons and computers set out to destroy humanity. ‘But it is a valid concern and many global and regional organisations have this high on the agenda’, said Virginia Dignum, professor on social artificial intelligence at the Delft University of Technology (TU Delft). ‘But it is not the only issue.’
She refers to the many algorithms that are biased against certain groups of people, feed existing inequalities and are not kept accountable in any way. Propublica reported in 2016 on software used across the country to predict future criminals that is biased against blacks. (If interested, a must-read is: Weapons of Math Destruction by Cathy O’Neil). There are barely any regulation in place to avoid this.
These algorithms often become dangerous, because its creators don’t really understand why the machine offers its solution. They are black boxes that people can’t question, nor hold accountable.
‘That is why we need to express the values we put into machines explicitly and ensure the computer is inspectable, verifiable and accountable’, said Dignum.
‘It is about algorithmic transparency (knowing why a decision is made), the data you use (without bias and complete), the stakeholders, who is paying for the technology, and so on.’
There is also no agreement yet on when superintelligent machines will exist. Existing software does one task really well at the most. ‘And it is unlikely that combining all the different tasks will result in one big intelligent machine that surpasses humans’, argued Dignum.
Lack of diversity
The documentary also lacks any diversity. It shows barely any women or people from non-western regions and cultures. ‘We see white, male men of a certain age. Women should be a part of this conversation. Just like people from other disciplines, regions and cultures’, said Dignum.
According to Dignum, a lack of diversity leads to a dominant focus on winning: it is us or the machines. ‘This goes completely against the ideas of support and coexistence many AI researchers aim for.’
Paine also leaves the viewer hanging and doesn’t offer any solution or way forward. Even though the featured AI researchers definitely have profound insights about responsible AI, they are not given the space to talk about those in the film. ‘By only addressing our fears, it won’t increase public awareness and it keeps people from demanding the boundaries these systems should have’, said Dignum. ‘I seriously start to wonder if this fixation by tech corporations on a dystopian future are a way for them to move public attention away from their practices and avoid regulation and corporate responsibility.’
Share us like candy!
Do you like what we do?
Do you know people who might be interested as well?
Don’t keep us all to yourself, but share us to everybody you know!
