Back to forum
Reading·Passage 3·Identifying writer's views or claims
medium
In many countries in the West, hysteria the transformative effects of the coming about the future of artificial intelligence (AI) ‘AI revolution’, they fail to realise the is everywhere. There seems to be no shortage complexity around deploying advanced of sensationalist news about how AI could machine learning systems in the real world. cure diseases, accelerate human innovation and improve human creativity. Just looking One of the most promising varieties of AI at the media headlines, you might think that technologies are neural networks. This form we are already living in a future where AI of machine learning is loosely modelled on has infiltrated every aspect of society. the neuronal structure of the human brain, but on a much smaller scale. But what many While it is undeniable that AI has opened up politicians do not understand is that simply a wealth of promising opportunities, it has adding a neural network to a problem will also led to the emergence of a mindset that not automatically mean that you’ll find a can be best described as ‘AI solutionism’. solution. Similarly, adding a neural network This is the philosophy that, given enough to a system of government does not mean it data, machine learning algorithms can will be instantaneously more inclusive or fair. solve all of humanity’s problems. But, in fact, instead of supporting AI progress, AI systems need a lot of data to function, this mindset actually jeopardises the value but the public sector typically does not of machine intelligence by disregarding have the appropriate data infrastructure to important AI safety principles and setting support advanced machine learning. Most unrealistic expectations about what AI can of the data remains stored in offline archives. really do for humanity. The few digitised sources of data that exist tend to be buried in bureaucracy. More often In only a few years, AI solutionism has made than not, data is spread across different its way from the technology evangelists’ government departments that each require mouths in Silicon Valley in California to special permissions to be accessed. Above the minds of government officials and all, the public sector typically lacks the policymakers around the world. The human talent with the right technological pendulum has swung from the dystopian capabilities to fully reap the benefits of notion that AI will destroy humanity to the machine intelligence. utopian belief that our algorithmic saviour is here. For these reasons, the sensationalism over AI has attracted many critics. Stuart We are now seeing governments pledge Russell, a professor of computer science support to national AI initiatives and at the University of California, Berkeley, compete in a technological race to has long advocated a more sensible and dominate the burgeoning machine-learning realistic approach that focuses on simple sector. While many politicians proclaim everyday applications of AI instead of the hypothetical takeover by super-intelligent The medical profession has also recognised robots. Similarly, Rodney Brooks, professor the drawbacks to AI. The IBM Watson for of robotics at Massachusetts Institute Oncology programme was a piece of AI that of Technology, writes that ‘almost all was meant to help doctors treat cancer. Even innovations in robotics and AI take far, far, though it was developed to deliver the best longer to be really widely deployed than recommendations, human experts found people in the field and outside the field it hard to trust the machine. As a result, imagine’. the AI programme was abandoned in most hospitals where it was trialled.

One of the many difficulties in deploying machine learning systems is that AI is Similar difficulties arose in the legal domain extremely susceptible to adversarial attacks. when algorithms were used in courts in the This means that a malicious AI can target US to sentence criminals. An algorithm another AI to make it behave in a certain calculated risk assessment scores and way, such as forcing it to make wrong advised judges on the sentencing. The system predictions. Many researchers have warned was found to amplify structural racial against the rolling out of AI without discrimination and was later abandoned. appropriate security standards and defence mechanisms. Still, AI security remains There are some crucial lessons here for an often overlooked topic when machine everyone aiming to boost investments in learning systems are installed. national AI programmes. These examples demonstrate that there is no AI solution If we are to reap the benefits and minimise for everything. Using AI simply for the the potential harms of AI, we must start sake of AI may not always be productive or thinking about how machine learning can useful, and not every issue is best addressed be meaningfully applied to specific areas by applying machine intelligence to it. of government, business and society. This All solutions come with a cost and not means we need to have a discussion about AI everything that can be automated should be. ethics and the distrust that many people have towards machine learning.

Most importantly, we need to be aware of the limitations of AI and where people still need to take the lead. Instead of painting an unrealistic picture of the power of AI, it is important to take a step back and separate the actual technological capabilities of AI from fantasy.
Stuart Russell’s proposals regarding the use of Al are impractical.
A
YES
B
NO
C
NOT GIVEN