Artificial intelligence: Not living up to the hype

• Follow

At a time when artificial intelligence (AI) is increasingly being looked upon to improve the lives of people, there are perhaps reasons to step back and see the areas where AI that haven’t really delivered. The hate speech on Facebook, for instance, or self-driving cars, or AI-based cancer detection, are some of these areas, according to a report by Bloomberg.

Social media

In social media, where venomous content and hate speech are routinely disseminated, AI-based solutions have clearly not helped make platforms, such as Facebook, safer spaces. In 2020, a study by New York University Stern School of Business recommended that Mark Zuckerberg-owned Facebook, as AI wasn’t up to the task of double the number of content moderators to 30,000 to keep its platform free of content that could trigger mental-health disorders. The author of ‘Weapons of Math Destruction, a book about the societal impact of algorithms, Cathy O’Neil has stated that Facebook’s AI “doesn’t work”.

Given how easy it is to dodge Facebook’s algorithm against fake news and misinformation by using special characters, such as ‘vaccine’ instead of ‘vaccine’, Zuckerberg himself has admitted that it’s difficult for AI to moderate posts because of the nuances of speech.

Self-driving cars

Similar is the problem with Elon Musk’s promise of delivering self-driving cars. The promise was first made by him in 2019 and the technology was supposed to be rolled out within a year. However, Tesla’s customers are still waiting for the promised fully autonomous driving vehicles. And the wait is unlikely to be short given how Musk himself tweeted that generalised self-driving technology was “a hard problem.”

The highly touted AI-based cancer detection tools have also left much to be desired. A study published earlier this year in the scientific journal ‘Nature’ analysed several machine-learning models designed to detect signs of Covid-19 in X-rays and CT scans and found that not even one could be used in a clinical setting. Another study published in the ‘British Medical Journal’ found that 94% of AI-based systems were less accurate in detecting signs of breast cancer than radiologists.

The market for AI, however, has been booming for the past many years, thanks to solid tech marketing. According to a report by MIT’s scientists, the very datasets that are used for building AI systems, such as computer vision and language processing, are replete with errors. To make adjustments, some businesses are shifting from pitching ‘AI-as-a-service’ to spending more on data-preparation software, according to Brendan Burke, a senior analyst at PitchBook.