Machines have learned how to master complex tasks. In 2016, Deep mind’s AlphaGo defeated World Go Champion Lee Sedol through machine learning (ML) methods; it studied the game and learned as it played. Gameplay has long been used as a standard for judging the progress of AI, and the self-learning abilities it demonstrated here were almost humanlike.
Two years later, in 2018, the world’s first commercial autonomous vehicle hire service was launched by Waymo. After years in the making, self-driving cars were finally out on the road for public use.
Elsewhere, AI has become adept at identifying human faces, translating spoken word into text, and spotting cancerous tumours. Indeed, right now AI is playing a huge role in mapping and containing the spread of COVID-19.
These serve as just a few standout examples of what AI has accomplished in recent years. But are we in danger of reaching an AI ‘winter’?
Will we reach an AI ‘winter’?
With so much progress being made in the field, it might seem inevitable that we will soon reach an AI ‘winter’ – or, in other words, a period when fewer and fewer advances are being made.
After all, the hype surrounding AI has peaked and troughed over the years as people first overestimated the abilities of this technology, and then re-evaluated its capabilities in a more measured and realistic way. The peaks here are known as AI summers, while AI winters refer to the troughs – periods where people become disillusioned with AI and progress starts to plateau.
This is not an unlikely scenario in the near future; indeed, despite billions of dollars being funnelled into the development of AI tools and some amazing milestones being achieved, the holy grail of artificial general intelligence (AGI) – where machines possess true human intelligence – remains elusive.
Can greater investment into the AI sector change this? If not, what else is holding us back?
Current techniques can only carry us so far. AI to date has veered in the direction of deep learning and ML, which although impressive, simply describe the pattern processing abilities of machines. Still, AI lacks consciousness and the ability to process abstract knowledge – key characteristics of human intelligence.
Let’s consider the feat of deep learning; a milestone which is both a blessing and a curse. On the one hand, it is the science behind many AI applications today, and has enabled powerful tech solutions that weren’t possible before. However, it has also caused scientists to narrow in too much on exploiting this ability, rather than focusing on producing new breakthroughs.
Research today has now largely shifted to applications: how can deep learning be used in novel ways? As a result, less people are diverting attention to making discoveries that will give us the new “deep learning”. And if the pace of research fails to keep up with applications of technology that we already have at our disposal, then we risk hitting an AI winter.
The field of AI has come a very long way in the last decade. But we have only just scratched the surface of what is possible, and there is simply not enough clean slate research happening right now. We need ingenuity and determination to avoid progress from stalling, which will enable us to make scientific and technological advances that will create truly intelligent machines.
It is not investment that will hold us back; nor the capabilities of technology. It is the mindsets of the organisations at the cutting edge of AI.