“Machine intelligence is the last invention that humanity will ever need to make”Nick Bostrom
As humanity has advanced, we have created increasingly ingenious labour-saving inventions from the humble plough to the steam engine to the, seemingly, again humble Roomba. As hinted at by the last example (and for the keen-eyed, the title) AI could well be the invention to end all (human made) inventions.
A Quick Definition and History
Leading scholars in “Artificial Intelligence” use the term to refer to the field of studying: ‘Intelligent Agents: any system that perceives its environment and takes actions that maximize its chance of achieving its goal.’. This definition also makes differentiating between AIs and Algorithms quite easy, as an Algorithm doesn’t perceive and doesn’t have the ability to update its behaviour naturally.
AI has had a long, if not wholly illustrious, history. There have been multiple research projects where promises were not lived up to and as such funding and talent left the field. A little after the 1950s people believed it would be possible to make a fully intelligent machine, unfortunately both the computing power and the amount of data available at the time, was incredibly limited and of course, other problems beset individual projects.
The lack of progress caused funding to dry up, which mostly came from government institutes and agencies. However, with the birth of the Internet in the early 1990s, communication between colleagues became many times easier and as such many modern paradigms and ideas sprouted around this time, and in the late 1990s the roots of modern AI systems began to grow.
The development of AI isn’t too dissimilar to the development of nuclear fusion, both have the potential to completely revolutionise our lives and both have had an arduous road to get where they have reached thus far.
Worldwide Spending on AI
Data, Data Beyond Your Wildest Imaginations
As companies accrue more data it becomes increasingly hard to make useful inferences from that data. Nowhere is this more obvious than in the life sciences sector over the last few years; with COVID came the rise of preprints and preprint servers and with such a glut of data and information, came a number of misunderstandings due to inabilities to process the data adequately.
To put things into perspective, in 2020 Facebook gathered 4 petabytes (1 petabyte = a million gigabytes) of data every day and 44 zettabytes (1 zettabyte = a trillion gigabytes) were stored in the cloud(1). Google processes more than 20 petabytes of data every day.(2)
A Match Made in Heaven
However, now that we have both volume and quality of data , if recorded properly and processed effectively, we can use it to make real predictions, from the stock market to predicting which chemicals will have a desired effect.
AI and data have an interdependent relationship, without copious amounts of data our AIs wouldn’t be anywhere near as good as they are, but without AI we will never be able to make sense of all the data we have now and will be processing in the future. Novartis uses AI to predict untested components researchers should explore to find new cures.(3)
AI assistants are agents capable of performing a task or service with a certain degree of intelligence based upon a command or a question. They come in many forms with various functions from Amazon’s Alexa to Apple’s Siri.
They differ from the automated phone menus that are used to route a call to the correct department as those do not have “a certain degree of intelligence”.
AI assistants are commonly able to connect to the internet, which means that devices that are purely made to house the AI assistant (e.g. Amazon Echo) are IoT (Internet of Things) devices, and as such are able to answer a wide variety of questions as they can simply draw the answer from online, unfortunately this cannot be said about specific tasks as there is no easy way for a speaker to bake a cake.
There’s a reason big businesses are spending a lot of money making AI assistants (Google had spent $3.2 billion on acquiring Nest Labs the company behind the smart thermostat), as AIs are the only structure that can adapt over time, without direct outside intervention.
As a customer uses an assistant it improves its functionality since it adapts to the user; a small scale example of this process is autocomplete/autocorrect, where you don’t need to type the entire word or even correctly type the word on your phone for it to suggest a word or correction. This information can be relayed back to the business which allows them to update and improve the basic, default AI behind the assistant.
Of course, businesses must be ethical with all this data, for example ensuring the data is transmitted securely and keeping the user well informed, an oversight in this has led to many scandals and can lead to user dissatisfaction which eventually leads to a loss in customer retention.
It is impossible to predict what the future holds for us but as computing power continues to increase and as people are living more digitised lives, the capabilities of our AIs will continue to grow and make our lives ever easier.
AI is more than just one singular technology, each one has a unique architecture and the data used to train it also has a massive influence on the outcome (something akin to nature and nurture in humans). Just as the work and decisions made in the late 1900s have transformed our lives today, the decisions made today will affect the outcomes for many generations to come.
If we are ever able to make a general AI as intelligent as a human then not only will it be our last invention, it will be the only invention that makes inventions.