In episode 2 of our A-Z podcast, we promised to write up a glossary of terms to help people keep up with the trend of automation. Voila…
If you’d like us to come in and talk about how automation will effect you – send us an email 🙂
Automation refers to the application of technology to production processes.
It’s a big deal.
Broadly speaking you can think of it as both robots in factories, and artificial intelligence performing ‘thinking tasks’.
Thinking tasks are those that rely on recalling and assembling information – like lawyers having to go through databases of legal records, or doctors need to look through medical resources to make diagnoses.
So we can expect there to be less work for both doctors and lawyers. And also for journalists, architects, insurance underwriters, finance experts, marketers and so on.
In short – if you have a job which can be broken down into a series of discrete steps, chances are it’ll be automated.
So the story of automation is both an exploration of the tech, and of the social ramifications that will emerge in its wake.
It’s a big topic, so we’ve decided to go for the keywords approach to mapping it out. We’ll keep updating this glossary as we go, so be sure to keep watch.
Artificial General Intelligence (AGI)
AGI describes the goal to create an AI system that is better than humans at all tasks. For this point to be reached, it needs to be able to converse convincingly; be able to do the ‘coffee test’, where it enters a kitchen and can figure out where everything is; complete a university degree; and pass an interview and successfully undertake a job.
AGI is often compared to Applied AI, which is narrow in its focus (such as Siri, or self-driving cars).
Luna is an AI developed by the founder of an organisation called Robots Without Borders, and strives to use AI to tackle the worlds most pressing problems. Luna demonstrates amazingly how far this type of AI has gotten. As reported in Big Think, when the AI is asked, “My boyfriend hit me, should I leave him?” she replied:
“Yes. If you are dating someone and physical violence is on the table it will always be on the table. You are also likely being abused and manipulated in other ways.”
In short, an algorithm is a mathematical formula that analyses data. But with cloud storage and parallel processing, algorithms are the important third ingredient of what Kevin Kelly calls the perfect storm of artificial intelligence.
A debate in artificial intelligence is around whether algorithms or data are the main players. Will we be coding intelligence, or letting it emerge from the data? Machine learning or algorithms?
This can be thought of as top-down and bottom-up approaches. Bottom up approaches let behaviour emerge out of the data, and top down approaches use algorithms as a sophisticated grammar to render intelligence consciousness.
In reality, developments in artificial intelligence are coming from both top-down algorithms and bottom-up machine learning.
Bitcoin received a lot of attention over the past few years. It promises a revolutionary new way of organising economic transactions. But more recently, the technology that Bitcoin is built upon – blockchain – is gaining the lion-share of attention. And the implications stretch beyond financial institutions.
In short, blockchain is a breakthrough in computer coding. It is a new way of storing data that doesn’t require a central storage point. It is defined by being decentralised.
That means you don’t need a company sitting in the middle storing the data. The old protocols of code meant that companies such as Facebook and Uber could profit by harvesting and storing data – this doesn’t happen in a blockchain world.
What follows is a whole new set of business models that will disrupt existing giants.
Let’s take Facebook. At the moment, if you join Facebook you increase the utility of the social network for everyone. After all, the more people that are part of the network the stronger it becomes (this is known as Metcalfe’s Law). But although the utility increases for everyone, the value only increases for Facebook. They store the data, so they own the data, and only they can profit from the data. If a blockchain version of Facebook appears (such as STEEM), then both the utility and the value are increased for the users.
So again, consider Uber. You wouldn’t need a company running the network, it would run itself with the value distributed across the users.
The promise of this is the emergence of an internet of value alongside the existing internet of information.
Once NLG and NLP are sophisticated enough, the promise of automation is that you will be able to converse with chatbots, who will be able to answer your requests.
Microsoft have a toolkit making it easy for developers to start building their own chatbots https://dev.botframework.com/.
The trend today is not to have your own servers storing all your data. Instead, you hire servers from companies like Amazon. In fact, Amazon cloud servers is the most profitable part of their business – with Uber, Adobe and Airbnb all customers.
This approach means you can scale up and down in line with demand. And the data is safer – it is stored in multiple locations.
The ease of access to cloud storage enables vast amounts of data to be collected and processed, which provides ‘food’ for the artificial intelligence to learn from.
With data being the highest valued currency in modern society, it also becomes increasingly important to learn how to interpret the trillions of data points generated every second. Simple statistics, such as average and correlation coefficients, help companies to extract trends and patterns within and between related data sets.
One example: Target mined data of purchasing patterns to generate customized coupons & surprised one teen’s parent by sending a diaper coupon … before she had revealed the pregnancy to her dad.
Data mining tools are easy to come by online, and often open source, such as R studio.
If you remember hearing about a computer beating contestants at the US show Jeopardy, then you may know about the Watson supercomputer.
Back then, Watson used to exist in single location in a huge cabinet. But now it is distributed across cloud based servers and can serve different people simultaneously.
Today, the service is being used to help people find the right healthcare, acting as a virtual shopping assistant on the high street, and even predicting crime. They have started collaborating with Sesame Street to help children learn:
‘Watson’s ability to absorb, correlate and learn from huge amounts of unstructured data and then deliver very personalized educational experiences will help transform the way in which kids learn and teachers teach.’
The human brain doesn’t do one thing at a time, instead it does multiple things simultaneously. Computers that want to think like brains need to do the same.
This is opposed to a more serial approach – which is a one-thing-after-another, or step by step approach.
Computers have got to the stage now where they can run millions of instructions at the same time across different networks.
You can think of this as a system whereby you feed a load of data into the computer, and the computer seeks out patterns. You can sort of think of this as the opposite to algorithms…
Natural language generation (NLG)
Natural language generation is writing language back in a way that can be read. Narrative Science are a company that have been generating automatic reports for a while, and have been featured as ‘authors’ of data-driven articles – such as markets and sports.
Most business intelligence dashboards, such as Tableau, contain some level of NLG.
Natural language processing (NLP)
Numbers are easy for a computer to understand. They can be structured logically and without ambiguity.
Language, however, isn’t so easy. It is unstructured.
The challenge, therefore, is to find ways to accurately store unstructured language data in databases with its proper meaning.
This means figuring out sarcasm, nuance of meaning, ambiguity, and many other things that makes human language so powerful. In English, many words that are nouns are also verbs. So ‘run’ is something that someone does, and it is also an event that someone can attend. Or a ‘chair’ is something that you sit on, and you can also ‘chair a meeting’. You can get a good overview of the code and concepts behind this area by looking at the Natural Language Toolkit.
Neural Lace Technology
Tech pioneer, Elon Musk, announced that he is starting a project to accelerate the development of human brain to computer communication. This involves implanting computer electrodes into the brain so that the two become intertwined.
A company pioneering artificial intelligence with a vision to reach human level intelligence. It’s an open source endeavour, and one to watch for the latest developments.
For a fascinating interview about state-of-the-art artificial intelligence click here.
The potential effect on the economy of automation requires a return to Marx. It allows for us to think again about capitalism – a capitalism where access to goods and services is abundant, as opposed to scarce.
The term ‘fully automated luxury communism’ has been gaining currency. The argument runs along the lines that automation should be deployed to create a post-work world that liberates people from the drudgery of labour. With time liberated, people can pursue their own artistic or inventive interests.
So when Uber create a fleet of driverless cars, it becomes run for the benefit of people as opposed to being run for profit.
Both post-scarcity and UBI entail a discussion of mans relationship to work. Do we work just for money, or do we work for meaning? Do we want a life of pure leisure? Will these trends enable all of us to pursue science and the creative arts?
We all know what robots are – but if you want to see how far we are with having robots replace humans in factories, check out this link.
One hypothesis of why we dream is that it enables us to play out possible scenarios in life without the real world risks associated with them. This is also vital for the data world. When some events are not feasible to test in the real world, such as the impacts of a flood on San Francisco, statisticians can build statistical models instead. This allows for role playing for how the real world outcome might look.
In finance, forecast modelling is common. Monte Carlo simulations using random samplings of the data to produce hundreds (or thousands) of simulations that output probabilities of the outcomes. When Elon Musk speaks of our world being a simulation, that would mean we were just one of these trials running — to help discern the probability of an unknown outcome.
Technology does not increase in a linear way. Instead it increases exponentially. According to futurist Ray Kurzweil, we are set to hit a point where this exponential increase reaches a tipping point and the scale of the changes will be the biggest change ever to hit humanity.
We can think of the singularity as a shorthand for massive technological change along the lines of ending death and illness, uploading consciousness into the web, and other (as yet unimaginable) changes.
In the US, there is a Singularity University set up by Kurzweil and supported by NASA, Google and other large corporations. It’s intention is to connect scientists and entrepreneurs and bring about the technology of the Singularity.
The idea has been criticised as being overly utopian, and akin to religious stories of the Rapture.
Universal Basic Income (UBI)
The threat of automation is that there will be less jobs for people to do, and we could face an employment crisis. One solution is the establishment of a guaranteed income for all in replace of welfare states.
It’s an idea that is being pushed by the Silicon Valley giants. But their support is being criticised as being a small concession to a world where they are privately capturing all economic power.