Artificial Intelligence 101
An explanation of what artificial intelligence really is.
This article is only for people who have a vague understanding of artificial intelligence and its potential impact on our society. This article is not for those with preexisting domain knowledge, but may be a handy re-cap anyway.
Artificial intelligence doesn’t have a single definition. Instead, there are three distinct classifications of AI: Narrow, General and Superintelligence.
Narrow
All currently available artificial intelligence is classed as ‘narrow’ AI. The ‘narrowness’ refers to the span of its knowledge. Narrow AI typically has a single purpose and is trained to only handle specific tasks.
As the availability of artificial intelligence becomes widespread, the majority of them will remain ‘narrow’. As they are the classification of AI that will automate work, Narrow AI will have a huge impact on our future, influencing transport (self-driving cars) and all types of work — from manual labour to white-collar jobs.
Well known examples of narrow AI are board game players such as IBM’s Chess player, Deep Blue, which famously beat Gary Kasparov in May 1997 and DeepMind’s Go player, AlphaGo, which beat Lee Sedol in March 2016.
Microsoft have also built AI that achieves a perfect Ms. Pac-Man score, so it’s great to see that every major company is starting to achieve equally important things with machine learning 😄
Other examples of narrow AI that are publicly available and ready for use in the workplace are Salesforce’s Einstein, which helps you get more out of your Salesforce data and Shopify’s Kit, which helps you do Facebook and Twitter advertising so that you can sell more on your Shopify store.
Feather fits into this category too, we build ‘narrow’ AI that is designed to help marketers and businesses with their marketing so that they can reach wider audiences and find new customers.
If you’re interesting in finding out more about artificial intelligence’s potential impact on the future of our society, I’ll link you to a few interesting talks at the end of the article.
General
General artificial intelligence refers to AI that has acquired ‘human-level intelligence’, this means the AI should be able to perform a wide variety of tasks, learn how to acquire new skills and can hold a conversation at a sufficient enough level to be able to pass a Turing test.
There are no known examples of general AI yet, however, attempts are certainly being made, notably at DeepMind.
A novel approach to developing general artificial intelligence is ‘whole brain simulation’, trying to artificially emulate how a human brain works so that artificial intelligence could potentially make a decision in the same manner that human intelligence does.
There is scepticism surrounding ‘whole brain simulation’ as a means of accomplishing general artificial intelligence, as imitating biological systems has historically not been the best approach to accomplish the same outcome.
Birds and planes can both fly, in very different ways ✈️.
It’s generally agreed by industry professionals that we’ll likely see some form of basic general artificial intelligence by 2025 at the latest.
Superintelligence
Superintelligent artificial intelligence is the classification that people worry about. It is possible that this level of intelligence may never be successfully developed, however, as it’s definitely within the realms of possibility. Ethical and moral debates surrounding superintelligence need to be had.
If you’re interested in investigating some of the considerations that are being made by industry experts, I highly recommend Nick Bostrom’s book, ‘Superintelligence’.
The existence of sentient superintelligent artificial intelligence is typically reserved for science fiction (see Skynet & HAL 9000), however with modern advances in machine learning over the last few decades, mixed with the huge amount of data we have, it’s now safer than ever to assume the possibility of superintelligence existing by the end of the 21st century.
Footnotes
If you’re interested in reading more about how programmers build artificial intelligence, I’ll be discussing that in much more detail in more articles in the near future.
If you don’t use Medium, I’ll be sharing the articles on Feather’s Twitter & Facebook pages as soon as they’re published, so follow us there instead.
If you‘re too impatient, I suggest you go read the Wikipedia articles for Artificial Intelligence, Artificial Neural Networks & the GOFAI approach.
I’d also highly recommend you watch the two following videos, they’re both explanations of the impact that artificial intelligence will have on our society and how automating jobs in the 21st century could be significantly different to the automation of jobs in the 20th.
Kurzgesagt are a YouTube channel that produce incredibly informative videos time and time again, this video about automation is no exception:
As a long time Cortex listener, it doesn’t surprise me that CGP Grey produced this informative video about how robots outperform humans, the implications of powerful AI and why this differs from previous forms of automation throughout the 20th century.
I also highly recommend you read through some of OpenAI’s and DeepMind’s publications.
If you’re interested in reading more about machine learning, I’ll be writing more articles about machine learning and artificial intelligence as well as how I’m utilising data analysis and machine learning to build Feather 🤖
