Robots and driverless cars
The desire for robots to be able to act autonomously and understand and navigate the world around them means there is a natural overlap between robotics and AI. While AI is only one of the technologies used in robotics, AI is helping robots move into new areas such as self-driving cars, delivery robots and helping robots learn new skills. At the start of 2020, General Motors and Honda revealed the Cruise Origin, an electric-powered driverless car and Waymo, the self-driving group inside Google parent Alphabet, recently opened its robotaxi service to the general public in Phoenix, Arizona, offering a service covering a 50-square mile area in the city.
We are on the verge of having neural networks that can create photo-realistic images or replicate someone’s voice in a pitch-perfect fashion. With that comes the potential for hugely disruptive social change, such as no longer being able to trust video or audio footage as genuine. Concerns are also starting to be raised about how such technologies will be used to misappropriate people’s images, with tools already being created to splice famous faces into adult films convincingly.
Speech and language recognition
Machine-learning systems have helped computers recognise what people are saying with an accuracy of almost 95%. Microsoft’s Artificial Intelligence and Research group also reported it had developed a system that transcribes spoken English as accurately as human transcribers.
With researchers pursuing a goal of 99% accuracy, expect speaking to computers to become increasingly common alongside more traditional forms of human-machine interaction.
Meanwhile, OpenAI’s language prediction model GPT-3 recently caused a stir with its ability to create articles that could pass as being written by a human.
Facial recognition and surveillance
In recent years, the accuracy of facial recognition systems has leapt forward, to the point where Chinese tech giant Baidu says it can match faces with 99% accuracy, providing the face is clear enough on the video. While police forces in western countries have generally only trialled using facial-recognition systems at large events, in China, the authorities are mounting a nationwide program to connect CCTV across the country to facial recognition and to use AI systems to track suspects and suspicious behavior, and has also expanded the use of facial-recognition glasses by police.
Although privacy regulations vary globally, it’s likely this more intrusive use of AI technology — including AI that can recognize emotions — will gradually become more widespread. However, a growing backlash and questions about the fairness of facial recognition systems have led to Amazon, IBM and Microsoft pausing or halting the sale of these systems to law enforcement.
AI could eventually have a dramatic impact on healthcare, helping radiologists to pick out tumors in x-rays, aiding researchers in spotting genetic sequences related to diseases and identifying molecules that could lead to more effective drugs. The recent breakthrough by Google’s AlphaFold 2 machine-learning system is expected to reduce the time taken during a key step when developing new drugs from months to hours.
There have been trials of AI-related technology in hospitals across the world. These include IBM’s Watson clinical decision support tool, which oncologists train at Memorial Sloan Kettering Cancer Center, and the use of Google DeepMind systems by the UK’s National Health Service, where it will help spot eye abnormalities and streamline the process of screening patients for head and neck cancers.
Reinforcing discrimination and bias
A growing concern is the way that machine-learning systems can codify the human biases and societal inequities reflected in their training data. These fears have been borne out by multiple examples of how a lack of variety in the data used to train such systems has negative real-world consequences.
In 2018, an MIT and Microsoft research paper found that facial recognition systems sold by major tech companies suffered from error rates that were significantly higher when identifying people with darker skin, an issue attributed to training datasets being composed mainly of white men.
Another study a year later highlighted that Amazon’s Rekognition facial recognition system had issues identifying the gender of individuals with darker skin, a charge that was challenged by Amazon executives, prompting one of the researchers to address the points raised in the Amazon rebuttal.
Since the studies were published, many of the major tech companies have, at least temporarily, ceased selling facial recognition systems to police departments.
Another example of insufficiently varied training data skewing outcomes made headlines in 2018 when Amazon scrapped a machine-learning recruitment tool that identified male applicants as preferable. Today research is ongoing into ways to offset biases in self-learning systems.
AI and global warming
As the size of machine-learning models and the datasets used to train them grows, so does the carbon footprint of the vast compute clusters that shape and run these models. The environmental impact of powering and cooling these compute farms was the subject of a paper by the World Economic Forum in 2018. One 2019 estimate was that the power required by machine-learning systems is doubling every 3.4 months.
The issue of the vast amount of energy needed to train powerful machine-learning models was brought into focus recently by the release of the language prediction model GPT-3, a sprawling neural network with some 175 billion parameters.
While the resources needed to train such models can be immense, and largely only available to major corporations, once trained the energy needed to run these models is significantly less. However, as demand for services based on these models grows, power consumption and the resulting environmental impact again becomes an issue.
One argument is that the environmental impact of training and running larger models needs to be weighed against the potential machine learning has to have a significant positive impact, for example, the more rapid advances in healthcare that look likely following the breakthrough made by Google DeepMind’s AlphaFold 2.