“This book is outdated.” So begins Toby Walsh's Faking It: Artificial Intelligence in a Human World. Such a declaration is rare in a book that aims to address pressing issues concerning humanity. In this case, artificial intelligence (AI) is disrupting daily life.
The book consists of 10 chapters covering a variety of issues, from faking intelligence to faking creativity to sharing possible solutions to combating counterfeiting, conveying complex ideas in approachable language . In contrast to the opening text, this text is truly timely and succeeds in highlighting the capabilities and limitations of machines and predictive intelligence. The book also addresses modern paranoia, such as job security, technology oversight, and ethical scrutiny, and is both a “guide and a warning.”
But before delving into these concerns, Walsh, a professor of artificial intelligence at UNSW and a fellow of the Australian Academy of Science, points out how faking intelligence is not a modern phenomenon. He uses the example of the “mechanical Turk.” It was “an impressive device that toured Europe and the Americas since its debut in 1770.” […] Until it was unfortunately destroyed in a museum fire in Philadelphia in 1854. ”
However, the first scientific meditation on imitating human intelligence is credited to Alan Turing. The British mathematician's paper entitled “Computing Machinery and Intelligence” is often cited as being pivotal to the advancement of AI. Turing, who was accused of being a homosexual, died of cyanide poisoning a few weeks before his 42nd birthday.
While several milestones have been achieved in AI research over the past 70 years, there have also been a series of false incidents. Walsh points to Expensify's “SmartScan” technology as an example. The software company claimed to use AI to automatically process receipts, but it turned out that the data transcription tasks were performed by “lowly paid people.” The book is full of such anecdotes, but what makes it such a rich read is that Walsh is an informed, sensitive, and politically aware writer. Two of his arguments stand out in particular because they also illustrate his gendered perspective on technology-related issues.
First, let's look at a machine learning (ML) algorithm developed by Stanford University in 2018. The algorithm helped separate “gay photos from straight photos.” Walsh rightly points out that AI gaydar is probably one of the worst uses of pseudoscience, given that “more than a dozen countries around the world have the death penalty for homosexuality.” .
Second, Walsh wants readers to question the names organizations give to personal assistants. Whether it's Apple's Siri or Amazon's Alexa, they're “almost always styled as a woman waiting for your orders. What does that say about our society?”
While Mr. Walsh highlights his concerns, he also focuses on what can be achieved by integrating AI into our lives. He said self-driving cars would “provide mobility for people who can't drive” and that in the future there would be “personalized AI tutors” for students. The development of the former required AI researchers to grapple with object recognition for decades. They helped address this problem through deep learning. “But deep learning is not the only reason AI tools are hot today,” Walsh wrote. He said this is another “key element” of his “Transformers architecture for neural networks, which in 2017 he introduced by his team of AI researchers at Google.” Masu.
You may have used ChatGPT. The “T” stands for “Transformer” in the free-to-use Large Language Model (LLM)-based AI system developed by OpenAI, which is used for computer vision, speech recognition, [and] Natural language processing. “This advancement shows that in the future, we will interact with machines rather than “point, click, and touch.” Machines will be intuitive to operate. Humans no longer need to “repeat the underlying context.” [their] questions and commands. ”
Read more: Toby Walsh – “We are afraid that what we create will displace us.”
The movie that reminded me of all of this was Her (2013), directed by Spike Jonze and starring Joaquin Phoenix. He plays Theodore Twombly, who is romantically attracted to virtual assistant Samantha (vocals: Scarlett Johansson). This film helped people imagine and discuss an exciting future. Perhaps this is why his ChatGPT announcement in 2022 received such an overwhelming response. Walsh said that in his first five days of service, he attracted more than 1 million users, and soon after, 100 million unique visitors per month. He compares his ChatGPT hit response to his Spotify and Instagram. The former took him 5 months, but the latter he registered 1 million users in 2 and a half months. It's not without reason that Elon Musk tweeted that “ChatGPT is scary good.” It had to be so. “Less than 1 percent of Wikipedia is complete” of the text “poured into GPT-2” and “the total text consumed by GPT-3 is equivalent to the amount a human would read in full text.” “About 100 times as much,” writes Walsh, “I write a book about every day of my life.”
But would anyone consent to the use of the data supplied to train such a tool? If the tool can create different narrative arcs based on prompts, how does it work? Shouldn't we be shocked by what we've achieved? Questions like these are being debated around the world, and OpenAI's data collection practices have come under intense scrutiny. Additionally, the organization's disregard for copyright has led to lawsuits from authors like Mona Awad and Paul Tremblay. In addition to creator rights, the indiscriminate use of AI-based tools is also a cause for concern. Rashmika Mandana's deepfake video and the Guardian's January 2024 report on the alleged rape of a minor in the Metaverse are just a few examples.
“Video and audio fakes are now being generated using deep learning,” Walsh wrote. From identity theft to election fraud to fake profiles on dating apps, there are a variety of issues facing organizations and AI researchers working with generative AI and LLM-based tools. Walsh said deepfakes are “perfect for catfish who want to create fake profiles to attract unsuspecting people.” It's estimated that about one in ten online dating profiles is fake, and “romance scams” cost about $50 million a year. ”
Citing the European Digital Services Act, which is “scheduled to come into force”, he suggested that government or regulatory intervention, rather than outright censorship, could be the solution. Walsh said China has come up with “stronger rules against deepfakes due to concerns about their potential impact on social 'stability.'” Furthermore, he warns: The ultimate fake AI is about to enter our homes and offices. ”
Walsh also claims that the hype surrounding AI is also the result of “bad journalistic” practices. To support his argument, in a chapter entitled AI Hype, he shares his five headlines published by major news outlets and argues that such a culture of excess can have consequences. I'm pointing it out. However, these aren't necessarily uninteresting. Take, for example, the announcement about a yet-to-be-developed Teslabot designed to avoid a “robot takeover.” Its main feature will be its slowness and weakness. [anyone] You can easily overtake and overwhelm them. ” Walsh said the release “featured a person in a full-body white bodysuit pretending to be a Teslabot.” He suspects someone has been fooled and concludes that “most of the claims being made about AI are exaggerated.”
Two of the biggest and most widely accepted are mass layoffs and making human-centered work irrelevant. Walsh said the COVID-19 pandemic has resulted in more hiring than layoffs. Further he writes: [that of elevator operators and locomotive firemen] In the 1950 U.S. Census, 270 jobs were completely eliminated by automation. ” This highlights the fact that it would be great for people to get used to AI and be better armed with it. Instead of being hostile towards it. “Rather than robots taking jobs from humans, humans using AI will take over the jobs of humans without AI,” Walsh said. This means we need to strive to streamline human and machine intelligence to co-create a better world.
One of the best examples of this is the book itself. Walsh wrote a few paragraphs using his ChatGPT, but it's almost unintelligible. He did this partly to entertain himself and partly to investigate whether human readers could be “fooled” into thinking it was written by the author. It was for a reason. This inherent playfulness not only serves to strengthen Walsh's argument, but is also a hallmark of his storytelling. His emphasis on engagement with his readers is evident in the fact that he always shows and doesn't tell. The image was created by DALL-E — an OpenAI tool that “can create realistic images and art from natural language descriptions” — based on his prompt in the book: “Cats in the style of Picasso's Cubist paintings.” , is a good example. As Walsh argues, there are limits to computational creativity, but what humans do have is the advantage of “collective evolutionary” learning. Despite the enormous challenges facing AI researchers and their beneficiaries, Walsh remains hopeful, believing that “the beauty of science is that it can self-correct.” Mistakes are identified and corrected. ”
Saurabh Sharma is a writer and freelance journalist based in Delhi. Instagram/X: @writerly_life.
