As we move into a future being increasingly transformed by a disruptive artificial intelligence (AI), what other changes can we begin to expect in the coming year?
Admittedly, while AI programmes have become more remarkably adept at figuring out what a person requires, they’re still constrained by statistically generated or scripted language.
So how close are we to imbuing machines with true understanding?
That simply depends on how rapidly it continues to evolve.
Extending the repertoire of natural language AI assistants
Most people are now familiar with AI assistants interpreting and processing natural, human language in an almost conversational manner.
But as you’ve probably experienced, the truth is they presently recognise only a narrow range of directives, and any unexpected deviations can lead to confusion.
However, the digital assistants’ repertoire is already expanding to literally fill in any gaps.
A technique enabling AI to utilise unlabelled text has already been developed, as a means of avoiding the expense and time of categorising and tagging all the data manually.
Moreover, a study of millions of sentences has led to a system enabling AI to predict missing words, which has performed as well as humans when put through a multiple-choice test.
Making the Smart Home a reality
Smart Home appliances have been around for quite a while now, yet its wholesale adaption is yet to materialise.
Presently, anyone wishing to transform their home into a smart one must choose a particular system over any others, as they each remain incompatible. Once you’ve picked your voice assistant, it’s only going to work for you if it’s linked up to accessories from the same supplier.
It also presents a whole new cyber security challenge, as criminals find they can hack into the cameras that are supposed to be deterring them.
The first challenge is being addressed as companies come together to create an open-source standard for internet-connected home products.
This in turn leads to a more overall solution to the second challenge, as security itself becomes more standardised and controllable.
The rise of Cryptocurrencies
Based on a global public network of thousands of computers, decentralized cryptocurrencies like Bitcoin provide freedom from corporate and government censorship.
Throughout last year, digital currency chipped away at the all-powerful dollar’s influence over the global financial system.
Seen as a parallel financial system that the US can’t control, countries like Russia, Iran, Venezuela and North Korea are increasingly taking interest in the potentialities of either setting up or adopting digital currencies.
China in particular has already set in motion plans to promote its own currency, the renminbi, as an alternative by speeding up development and encouraging international adoption of a digital version.
The cooperate world has also entered the fray with the Libra, whose network will be controlled by a small number of vetted private entities.
Lifesaving monitors on your wrist
Although they’re yet to approach the sophistication of hospital equipment standards, ECG-enabled smart watches come with a distinct advantage: they’re monitoring a wearer’s condition at all times.
A smart watch-compatible band capable of detecting atrial fibrillation, a frequent cause of blood clots and strokes, is already available. Software utilising deep learning can also screen for hyperkalemia, or elevated potassium levels in the blood.
An FDA-cleared ECG feature, embedded in the watch itself, was launched just last year, leading to an announcement by respected health-device maker Withings of a forthcoming watch that will contain an ECG reader.
Moreover, preliminary results of both an app and a two-sensor system that can detect a certain type of heart attack have been recently presented to the American Heart Association.
2020 Vision: Enabling computers to see things the way you do
Whereas humans identify, say, a type of bird by its plumage, beak or size, a neural network that simply detects pixel patterns can hardly distinguish the bird from its background.
As long as a neural network remains vulnerable to such mistakes, it can hardly be applied safely to more high-stakes situations such as healthcare, where a doctor might need help classifying a type of tumour.
Inspired by the way humans see things, a new image recognition algorithm has been developed that scores highly on tasks such as bird species and a car model identification.
Accurate computer vision is a vital component of self-driving vehicles.
It’s also increasing essential in retail-style environments, where it can be used at register-less checkouts or in making visualised recommendations.
Physicists have successfully developed chips utilising the strange behaviour of quantum physics by generating and manipulating single particles of light within programmable nano-scale circuits.
The primary difference between regular and quantum physics is entanglement, which describes how, when a pair or group of particles is generated, interact, or share spatial proximity, the quantum state of each particle cannot be described independently.
Yet in experiments involving two of the newly developed chips, a collapsing of the entanglement link allowed the individual quantum state of a particle to be transmitted to another particle on the receiver chip.
In other words, the photons on either chip share a single quantum state.
Through encoding quantum information in light generated inside the circuits, these new chips can process information with high efficiency and extremely low noise.
Manipulated video is recognised as a problem by everyone from technology companies to the US government, who are all hoping to develop deepfake-busting technology.
It has even been suggested that the companies creating the tools used to manufacture deepfakes take on responsibility for vetting users and penalising anyone violating their standards.
There may be other solutions, however.
To make it easier to tell if the media has been tampered, ‘verified-at-capture’ or ‘controlled-capture’ technologies use a variety of techniques to sign, geotag, and timestamp an image or video when it’s first created.
The sad end of emotion-recognition tech?
Currently used to assess job applicants and even people suspected of crimes, emotion recognition tech is a $20 billion market.
It’s hoped it might also be applied in deducing gamers’ emotional states once its adapted for use in VR headsets.
And yet there’s a distinct lack of evidence that machines can work out how we’re feeling. Whereas there is evidence that emotion recognition tech amplifies race and gender disparities.
In fact, it’s far from easy to use facial expressions alone to accurately tell how someone is feeling, according to a recent study by the Association for Psychological Science.
And in its annual report, research institute AI Now has gone even further, with calls for facial recognition technology to be banned from use in decisions that affect people’s lives, at least until the risks have been studied properly.
Data ethics and innovation: the new digital partnership
Obviously, all future innovation relies upon maintaining the free flow of data and avoiding digital protectionism.
Equally obviously, technology companies need to focus on becoming responsible enterprises with proven track records of trust, data stewardship, and a commitment to enabling a new, beneficial partnership between technology, public policy, and society.
Would you like to know how you can do this?
The future of AI’s impact on society is discussed in ‘IBM Vision 2024’, outlining an essential roadmap to building a trustworthy, open and inclusive digital Europe.
It sets out ideas for a new digital partnership based around similarly minded technology companies with both data ethics and innovation at their core.
To be a part of it, you can follow all the latest data ethics’ developments here, as well as keeping up with IBM Cloud, IBM Analytics, and Quantum computing news as you optimise your and your clients’ Journey to Cloud.