What is AI?
As technology expands into virtually every corner of human experience, it’s clear that computers with sophisticated artificial intelligence (AI) capabilities are more than simply a sum of their chips.
Much like humans, these computers learn from experience. In their case though, the experience is gathered from data. Computers are brilliant at computational intelligence – remembering, calculating probability, discerning patterns in datasets too subtle for people to notice. This capacity is harnessed in the field of predictive analytics using machine learning.
What is machine learning?
AI computers are ‘shown’ what to do rather than told to carry out a detailed set of coded instructions.
If you want to get to grips with some of the basics, you can learn more in our AI: how does machine learning work? article.
The main point to take away from a machine learning program is that they are incredible (often better than humans) at doing just one thing – whether that’s identifying a pattern in financial data, or winning quiz shows. This is called ‘narrow AI’, and all AI software currently in existence is narrow AI.
When we think of ‘artificial intelligence’ our minds automatically skip to the sort of computer you see in sci-fi that’s human-like across a huge variety of tasks – called ‘general AI’. So while narrow AI programs are certainly sophisticated, it doesn’t seem right to call them intelligent.
Next generation AI
The creation of a more general purpose AI, being more versatile and closer to the way human intelligence operates, goes hand in hand with what some call the fourth industrial revolution: the coming together of AI, blockchain, internet of things and 5G. However, bringing these technologies together requires more advances in cloud storage, quantum computing and powerful computational algorithms.
There are some who argue that truly general AI will always remain a fiction. Yet AI innovation continues feverishly. Sub-disciplines of semantics, contextual programming and heuristics are developing, and human interaction is being enhanced through cyber psychology.
Law and policy
Fundamental to AI playing an ever-increasing part in our lives is trust. Principally, trust in outcome and trust in security. We need to see that intelligent machines can operate reliably, safely and consistently.
- Trust in outcome. AI does not ‘do’ the bigger picture, so correlations may not properly reflect the context of a complex situation. AI is also prone to making the kind of errors a human would never make. Systematic evaluation of quality and suitability of training data will be crucial, as will be the involvement of human judgement (at least on appeal) where consequential decisions have been made.
- Trust in security. Privacy laws and the models already developed based on transparency and accountability will be key tenets for the deployment of AI. Privacy by design will facilitate techniques that use personal data without accessing the identity of individuals.
Understanding the economic model
We’re still very much in the early stages. AI needs data, and lots of it, so the phrase ‘data is the new oil' will only become truer. Vast amounts of structured and unstructured personal data are collected through our use of technology; and even more through machine to machine data with the internet of ‘industrial’ things.
Don’t confuse automation with AI. They work together; automation creating huge databases, and AI understanding and manipulating them. Licensing intellectual property not just of AI technology, but also of datasets will become key here, and the laws around competition, exclusive arrangements and the rules on dominant positions will come into play in ways not yet envisaged.
We will be looking at the impact of AI on areas such as privacy, regulation, licencing and competition in other articles soon. If you would like to be added to our mailing list, please sign up for email alerts.