Artificial Intelligence (AI) seems to be everywhere nowadays.
It’s in the news, with updates on breakthrough developments. It’s in social media, with cautionary comments about the next big thing from academics and digital heavyweights. It’s in everyday conversations, with discussion about potentially dire consequences for humanity and the promise of a re-imagined future.
So how much of today’s discussion and predictions are ‘show me the money’ real and how much is just hype?
In this introductory post for non-technical readers, I’ll provide a brief summary on where AI’s been, where it is today and predictions for its future.
Quick Trip Down AI Memory Lane
So we all know that AI is not new. In fact, it’s been around for more than half a century, dating back to the work of Turing, McCulloch, McCathy, Minsky, Rochester and Shannon from the 1950’s.
After its initial growth spurt, progress slowed through the 1970’s and 1980’s in what scientists have referred to as the ‘AI winter’, when projects struggled to deliver to expectations and funding became scarce.
Toward the end of the 3rd Industrial Revolution in the 1990’s, (you’ll recall this is the one related with computing, automation and the internet) a range of factors converged providing a catalyst for AI’s resurgence.
Exponential increases in computing power and storage (and declining prices), new programming approaches and algorithmic advances, data proliferation at an unprecedented rate, and increasingly sophisticated data analytics all contributed to step-change advances in capability.
Artificial Intelligence Introduction – ‘AI Inside’ ?
AI is essentially about machines performing cognitive functions that are (or were) normally associated with human intelligence. Sensing, assessing, reasoning, inferring, predicting, acting, learning and even creating are examples that are increasingly being delivered by AI technologies.
These capabilities are realised not purely through technology but rely on a range of disciplines working in concert including mathematics, neurosciences, linguistics, informatics, psychology, philosophy and others. Combinations of sensory, behavioural and cognitive layers interact to deliver the overall AI user experience.
Some of the more notable current AI technologies include:
- Machine Learning (ML) – enables artificial systems to learn by having access to vast amounts of data to improve from ‘experience’ without needing to be explicitly programmed. ML uses statistical techniques to identify patterns in data from which it makes inferences and predictions. These are tested against inputs to improve accuracy and performance. Deep Learning is a specialised subset of ML that uses multi-layered neural networks to solve problems from data that is unstructured or unlabelled.
Current ML approaches include:
- Natural Language Processing (NLP) – enables artificial systems to understand language through the analysis, manipulation and generation of human languages. The technology includes classification, context extraction, machine translation, text generation and question answering.
- Speech – enables artificial systems to speak and hear by recognising and transcribing human speech. The technology includes speech to text and text to speech.
- Computer Vision – enables artificial systems to see by automating the human visual system to understand digital images and videos. The technology includes object, facial and character recognition, scene reconstruction, event detection, video tracking, 3D pose estimation, motion estimation and image restoration.
- Expert Systems – are computer systems that emulate the decision-making ability of human experts. Expert systems are designed to solve complex problems by reasoning through bodies of knowledge, represented mainly by if-then rules rather than through conventional procedural code. The technology includes inference engines and knowledge bases.
- Planning & Optimisation – incorporates artificial intelligence in automated planning and scheduling, typically for execution by intelligent agents, autonomous robots and unmanned vehicles. Planning environments are typically dynamic and unknown and must be discovered and optimized in a multidimensional space. The technology includes realisation reduction, classical, probabilistic and temporal analysis.
- Robotics and Intelligent Automation – includes the next generation of AI enabled robots that are able to do a broad range of tasks in unpredictable environments. Technologies include robot process automation (RPA), an emerging form of business process automation enabled by software robots or AI workers.
All AI is not Equal – Path to Super Intelligence
Like many other sciences, not all AI’s are created equal.
There are progressive levels, with their own distinct characteristics and capabilities together with substantially differing implications for humankind.
Nearly all AI driven systems today are defined as Narrow AI or ANI (also known as Weak or Applied AI). IBM’s Deep Blue in the 90’s and Watson in the 2000’s beat the best human competition around but only in the specific games for which they were developed. Even Alexa, Google Assistant and Siri can smartly answer a broad range of questions but can’t make you that cup of coffee just yet. Narrow AI is showing its value across a range of use cases where systems are able to outperform humans with repetitive functions, such as service call engagement, medical diagnosis, driving, etc.
The next level up from ANI is Artificial General Intelligence (AGI, also known as Strong AI). Ben Goertzel, respected AI author defines AGI as “a system intelligence that has a general scope and is good at generalization across various goals and contexts”. In other words, AGI aims to create machines capable of successfully performing any intellectual task that a human can, although its speed and ability to process data would be far greater. AGI has been and continues to be something of an AI holy grail with many organisations actively researching this space.
The third level in the progression is Artificial Superintelligence (ASI). ASI is a hypothetical intelligence that surpasses that of the smartest human. Vernor Vinge, a mathematician and fiction writer, coined the term “the singularity” in 1993 describing the inflection point where machines become smarter than humans. This is the level that’s often depicted in doomsday science fiction movies where we’re all lucky that Arnie’s on our side.
Head of Facebook’s AI research facility, Yann LeCun commented late last year that AI still has a long, long way to go before it approaches anything near the intelligence of a baby, or even an animal. Estimations of exactly when (and if indeed) we’ll get to a future where AGI or ASI are a reality vary, with some prediction speculating AGI may be around as early as the 2030’s and ASI some time afterwards in the 2060’s. There are still many challenges and issues before getting near this potential future (not least of which are some of the ethical considerations) but these predictions certainly do give us food for thought.
Recent Developments – Pushing Back the AI Frontiers
Just search on recent AI breakthroughs and you’ll get a sense of the rate of progress being made – I’ve noted a few highlights below.
Google CEO, Sundar Pichai, confirmed the companies AI focus in May last year announcing, “An important shift from a mobile first world to an AI first world.”
From its establishment in 2010, the Google Brain team (that started out as a Google X project) has made significant progress with breakthroughs in deep learning ‘cat detection’ in 2012 and more recently with developments in AI devised encryption systems, image enhancement, translation and robotics.
Google’s Deepmind division has also made significant progress over the last 4 years (after joining the Google fold in 2014). Recent developments include teaching its AI to dream, fight, play games, learn like humans (addressing AI’s fundamental ‘catastrophic forgetting’ problem) and even enabling it to build new AI’s that outperform human-built AI systems.
Google X delivered a number of notable ‘moonshot’ AI enabled inventions over the period including Google Glass (augmented reality head-mounted display) and Waymo’s driverless vehicles, that was transitioned to a separate company in 2016.
From its early lineage in Google Voice Search in 2011, Google has continued evolving its NLP, speech and ML capabilities and in 2016 unveiled the Google Home Assistant. Google also recently released its Pixel mobile phone, ‘built for AI’ with Google Lens visual search capability and baked in AI assistants.
Microsoft’s CEO, Satya Nadella, also called out AI as a core focus last year, highlighting the intention to democratize and harness AI’s potential and infuse intelligence into all of its products and services.
Microsoft has successfully developed voice search and virtual assistant technology with their Cortana virtual assistant. Cortana was first demonstrated in 2013 and has subsequently been integrated across a number of products. The company has also made great strides in the virtual/mixed reality space with the HoloLens mixed reality headset that’s the first fully self-contained, holographic computer, enabling interaction with high-definition holograms.
Microsoft has also continued to make strides with its AI research and development, building an AI capable of answering questions after reading passages of text and more recently, developing an AI that matches human performance in translating news from Chinese to English.
Facebook also has a strong AI focus having hired leading AI innovator, Yann LeCunn, to direct its AI Research lab back in 2013. Although a number of its AI initiatives are perhaps less visible and ‘under the hood’, it continues to actively research and use AI extensively. Services are AI enabled with neural networks to identify and filter specific content, ranks posts, perform facial recognition, language translations, determine the most relevant ads, and much more.
Amazon has “rebuilt itself around AI” over the last decade. Early advances in applying machine learning to product recommendations proved its potential, promoting adoption in other areas.
Amazon’s smart speaker, the Echo with its Alexa voice platform was conceived in 2011 and developed in-house from the ground up. In its development, Amazon was able to leverage its global Amazon Web Services (AWS) infrastructure capability to provide the cloud backbone for the service. The Echo, released in 2014, was the first home assistant on the market and gave Amazon a jump on the competition and provided a source of valuable input to boost its machine learning and AI efforts.
Having developed its machine learning platform, Amazon offered it as an AWS service to customers in 2015. The service was further extended to include, a text to speech service called Polly, a natural language service called Lex, an image and video analysis service called Rekognition, and a fully-managed machine learning platform called Sagemaker. Other Amazon AI powered highlights include Amazon air drone delivery service and Amazon Go, a deep learning powered cashierless grocery chain.
Many other companies are also currently making substantial investments in similar AI technologies. Apple, IBM, Nvidia, Uber together with Chinese industry giants Tencent (Chinese ecommerce multinational), Baidu (82% of dominant Chinese search volume) and Alibaba are just a few to note.
Perhaps one of the more ‘out there’ recent developments is Neuralink, a company launched by Elon Musk last year. The company is working on creating a brain-computer “neural lace” interface that would see human brains in symbiosis with AI.
AI Market’s Growth Potential
You would guess that the AI market is expected to be boom and you’d be right.
In fact, AI is already in a number of the devices and services that we’re using everyday, although they don’t specifically carry an ‘AI inside’ label. In fact, a survey last year of 6,000 customers in six countries by Pagasystems found that 84% of consumers are regularly using at least one AI-powered service or device.
IDC is predicting a 50% compound annual growth rate for global spending on AI, reaching $57.6 billion by the year 2021.
Large technology players have been doubling down on AI not wanting to be caught behind the next wave of digital disruption. AI start-up investment has also ramped-up significantly over the last decade. Governments have identified the disruptive potential of AI and are wanting to position for the future with China, for example, planning to build an AI industry worth around $150 billion by 2030.
Associated with the growth in demand is the need for skills. A recent report from Chinese technology company, Tencent, indicates there are about 300,000 “AI practitioners and researchers” worldwide, but millions of roles available for people with these qualifications.
The recent (2017) Stanford University AI Index study provides a supporting snapshot of the growth in demand for AI skills as well as the upward trend in the number of enrollments in popular AI technologies at selected US universities.
AI Adoption and Opportunity
Indications are that we’re still at the early stages of AI enterprise adoption with a recent study by Statistica indicating that only 5% of those surveyed had adopted AI extensively in their organisations, in comparison to 22% that had neither adopted or had plans to adopt AI. These figures are backed up by another study by the EIU that indicated that 75% of executives surveyed will be “actively implementing” AI in their companies within the next three years.
Adoption level of AI in business organisations worldwide, as of 2017 (Statistica).
In terms of opportunity, a recent Deloitte survey indicated that 83% of the most aggressive adopters of AI and cognitive technologies said their companies have already achieved either moderate (53%) or substantial (30%) benefits.
Wrapping Up
“We have only begun to tap into all the ways it will make people’s lives more productive and creative.” – Bill Gates on AI’s potential in May last year.
Since the advent of computing we’ve been interacting with reactive, programmed machines through screens and keyboards (and there was that step change to using a mouse in the 80’s).
We have seen tremendous advances in AI over the last decade that offers a glimpse of a future that is very different from our past. Tomorrow’s AI enabled world has ‘ambient’ computing, ubiquitous and connected. Computers are voice and vision enabled – they can see and talk to you. Machines are intelligent, predictive, proactive, able to learn and teach.
“AI is one of the most important things that humanity is working on. It’s more profound than, I don’t know, electricity or fire.” – Sundar Pichai, Google CEO earlier this year.
If that sounds like hot air then just look around – talking bots, virtual home assistants, self-driving cars, self-learning computers, translating mobile phones, etc. are all around today!
The AI enabled paradigm-shift is here and with increasing focus and levels on investment, it’s reasonable to expect that the pace of change will only accelerate. It remains to be seen how we’ll all adapt, harness AI’s potential for the greater good and hopefully avoid its potential downsides.
It’s a great challenge, as Bill Gates points out, “…an amazing time to be alive.”