Ex-Machina: bridging the gap from sci-fi to sci-fact
Last night I decided to re-watch 2014’s Ex-Machina. Alongside Marvin Minsky’s ‘Society of Mind’, a book about the rule based and model based organisation of cognitive functions, and Spike Jonze’s Her, Ex-Machina is one of the main pieces of creative art that excited my awe in artificial intelligence and its effects on the human condition.
In this post, I’ll be looking at a number of key points that were depicted as sci-fi, but which might not be too far from becoming sci-fact.
The IoT and the state of privacy
As our devices have become more connected, it has become easier for companies such as Google, Facebook, Squarespace, Netflix and Amazon to gather more information to provide personalised services. This personalised service platform has become possible through the accumulation of big data on the general level, and then on the individual level. Whether its the predictive search on Google, or the recommendation/suggested content on YouTube and Netflix, connectivity has become a matter of convenience through the ability of machine learning algorithims to accurately predict what we would like from our previous behaviour.
It was after watching Google I/O conference, that it became clear to me that the next frontier of applied convenience (which I believe is the manner in which these companies use big data), will be through the use of object recognition. With the advent of the Homepod, Alexa skills and Google Home products, our smart devices will become able to track our movements and be trained to follow us within a crowd. The opening of Ex-Machina shows just this:
The phone as well as laptop that Caleb Smith was using, was monitoing his facial movements, body movements, as well as the movements of his coworkers behind him as can be seen above. With our object recongition systems reaching near human accuracy in our neural nets, and AR applications such as Google Lens on the horizon, our smart devices will soon have the capacity to watch us in a literal sense.
The age of ambient computing will come in step with the age of cognitive ai systems — systems that can hear, track, see and speak to us, in a way that is tailored to each of our traits and behaviour patterns. In my last post I wrote about Google’s Multimodel Ai algorithim, that is able to transcribe languages, parse a language, classify an image as well as caption an image all at once, with the added importance of its ability to get better the more tasks it was given to perform. I pointed out that this lateral task approach opposed to a single task approach, would be the cornerstone of modeling the way towards artificial general intelligence.
Google’s efforts are not alone in this space. MIT has taken up the challenge of making their algorithim’s multi-modal as well, with researches working on ‘creating a way for them to link, or align, knowledge from one sense to another’. With AI systems that know what an image is, can then recognise that the sounds it hears are coming from that image (in the case of an animal, ambulance or person), and then infer a response from the audio, visual and linguistic cues, the way towards general AI is well on the way (but not around the corner).
A Mind made of many minds
When Caleb meets the protagonist AI Ava in Ex-Machina, he is immediately met with an entity that is more than just a machine. Her stochastic brain allows her to infer, inflect as well as make jokes, and her responses to his questions as well as questions she herself poses, shows a cogntive capacity that demonstrates human-like operations at work.
Her cogntive functions are a product of her synthetic brain. Her brain’s software is modelled from the data collected by the ficitonal company Blue Book — the film’s version of Google. The company’s name pays homage to the work of the philosopher Ludwig Wittgenstein, one of my faviourite thinkers as well as one of the most instrumental philosophers in the 20th century by demonstrating the importance of language in how thinking occurs. For Wittgenstein, to imagine a language was to imagine a game or way of life. A game in the sense that there are rules both implcit and explicit for players to follow in order for the game to be played. And a way of life, in the sense that there is a culture or range of cultural traits (such as idioms) exhibited in a language. Ex-machina plays this out by showing that Ava is trained on the language of what it means to be human, and therefore learns more than just speech and ethical behaviour, but also lying, subterfuge and manipulation.
Her ability to accomplish these human traits are founded on her understanding of what she sees and hears. The object recognition from the phone and laptop we see at the beginning of the film, is the same software that allows Ava to track, understand and respond to Caleb’s gestures both verbal and non-verbal. More than that, her ability to tell when he is aroused, lying or keeping something from her, is informed from her collection of facial gestures collected from the entire web. Her creator, Nathan, informs Caleb that this data was collected legally yet unethically, with the telecommunication companies knowing he was monitoring civilians and allowing him to do so.
Ava’s brain is therefore not centralised, but a single mind made of a network of minds. This network is both computational (the servers than run the algorithims which her single brain-casing was trained on) as well as organic (the data provided from the civilians who used Blue Book).
Whether this will be the same for our own world of general AI will remain to be seen. Google’s I/O conference demonstrated how much better their object recongition algorithims are getting at recognising us through Google Photos. Likewise, Deepmind’s algorithims and Kinetics dataset are being trained on videos that we have uploaded onto YouTube, showing off a range of actions and tasks from playing tennis, opening objects, hugging to scrambling eggs. The accumulation of visual and audiotry data on the internet will provide lenses for the AI’s of tomorrow to gain understanding of our actions in a range of contexts, made possible through our upload rates which are rising by the day through our social media apps.
The future of AI will be imagined through our sci-fi movies as well as TV shows, while its reality will be slowly modelled and created by the progress of both human researchers and their slowly evolving algorithms.
Thanks for reading :)
