Our misplaced faith in AI is turning the internet into a cesspool of misinformation and spam

The printing press is surely near the top of any list of the greatest inventions of all time. Before the printing press, books were transcribed by hand, and innovations were spread by explorers and adventurers like Marco Polo who traveled the Silk Road to Asia and returned to Europe with news of paper, paper money, compasses, coal, porcelain, and more. Isolated people who lived on distant islands, in dense forests, or on the wrong side of deserts might be deprived of great inventions for centuries.

Conspiracy theories and paranoia are as old as the human race, but the internet has made their circulation fast and frenzied.

The printing press made books less expensive and more accessible and allowed ideas and inventions to be spread widely and improved. Even more powerful communication innovations appeared in the 19th and 20th centuries with the development of the telegraph, telephone, radio, and television. There were also growing concerns about their overuse. Though it is hard to disentangle correlation and causation, education studies repeatedly found lower test scores for Americans who watched more television than others. Likewise, there were concerns about teenagers who spent hours a day on the telephone.

Nonetheless, the perception that communication technologies had a positive impact on our lives and, particularly, innovation, was deeply entrenched. Many economists believe that communication is the lubricant that spurred the diffusion of scientific advances, useful inventions, and technological improvements.

These beliefs were one reason for the optimism surrounding the Internet in the 1990s. Originally intended to allow military and academic networks to share information and computer resources, it soon became available to everyone. Many expected a golden age as the development and diffusion of new innovations accelerated.

There is certainly a wealth of information on the internet — ranging from specialized research papers to detailed instructions on how to adjust the height of a reel lawnmower. (Yes, one of us did that this past weekend.) One of us is also happy that his 10-year old son can find so many high-quality videos on YouTube, helping him learn about science, technology, and history.

The use of large language models by unscrupulous actors means these tools are unleashing a tsunami of spam, phishing, clickbait, and viruses that is going to drown the internet.

However, optimism about the internet ushering a new golden age has been battered and crushed. One unexpected development is the number of hours people spend on the internet. Worldwide, internet users spend almost 7 hours a day on the internet—and very little of that time is spent reading research papers, preparing for a do-it-yourself project, or watching history videos. An average of 2 1/2 hours a day is spent on social media; teenagers, on average, spend more than 6 hours a day. Not surprisingly, some experts speculate that the internet has made us less productive.

These many hours spent online might be rationalized as entertainment or social bonding, but the reality is hardly benign. The promise of Facebook, Twitter, Instagram, and other social media platforms was that they would bring people closer together by allowing friends and family to maintain and nourish social ties by sharing events in their lives—a new dog, a birthday party, and even what was eaten for lunch. When Facebook went public, Mark Zuckerberg issued a statement of intent that repeatedly stressed the communal power of sharing:

At Facebook, we build tools to help people connect with the people they want and share what they want, and by doing this we are extending people’s capacity to build and maintain relationships.

This was surely corporate BS, and certainly fantastically naive. One outcome of social media that should have been foreseen is the allure of posing—exaggerating or diminishing as needed in order to appear happier and more successful than we really are, which provokes others to do the same. I’m having more fun than you. I’m more successful than you. I’m happier than you.

This bragging contest was amped up when Facebook introduced the Like and, later, Share buttons; and when Twitter in turn introduced a Retweet button. The goal of social media shifted from sharing and bragging among family and friends to becoming a social media celebrity—amassing followers by posting content that persuades strangers to watch and click. Hyperbole, stunts, and dishonesty are rewarded.

The internet is an extremely efficient transmitter of disinformation spread by individuals, businesses, and governments. Many people—especially teenagers and children—have become addicted to videos that promote skin bleaching, weight loss, drugs, firearms, racism, misogyny, and self-harm. Russia has an army of cyberwarriors spewing a “firehose of falsehoods” all day every day.

Conspiracy theories and paranoia are as old as the human race, but the internet has made their circulation fast and frenzied. A 2019 YouGov poll of British adults found that 20 percent believe that the Moon landings were probably or definitely faked. Similar polls in the United States and Russia have given similar results.  A 2016 Public Policy Survey found that 4 percent of Americans (12.5 million people) believe that the U.S. government is controlled by alien lizard people. Millions believe that the Earth is flat; school shootings are false-flag operations; and Bill Gates orchestrated the COVID-19 crisis so that he could use vaccines to insert microchips in our bodies. Instead of disseminating useful information and debating important ideas, the internet is too often used to promote whacky theories and titillate users with inane gossip. Instead of nurturing communal feelings, social media is tribalizing in that the world becomes increasingly divided into groups convinced that they are right and others are wrong.

It is going to get worse, much worse. Chatbots like ChatGPT and other large language models (LLMs) are, in a similar techno-utopian vein, supposedly going to revolutionize the way we write, research, and search the internet. But these amalgamations of algorithms aren’t “intelligent” in any real sense of the word, as we have argued previously; as a result, they tend to often spit out misinformation and falsehoods. And their use by unscrupulous actors means these tools are unleashing a tsunami of spam, phishing, clickbait, and viruses that is going to drown the internet. Just this past week, ChatGPT told voters in one Australian town that their mayor had been sent to prison for bribery and corruption, and that a Washington Post story reported that a Georgetown University law professor had sexually harassed a student on a class trip to Alaska. The mayor was not in prison and had, in fact, helped expose a bribery scandal involving the Australian National Reserve Bank. The law professor had never taught at Georgetown, had never taken students on a trip, had never been accused of sexual harassment, and there was no Washington Post story. Our faith and excitement in the promise of chatbots like these to revolutionize access to information is misplaced.


Want more health and science stories in your inbox? Subscribe to Salon’s weekly newsletter The Vulgar Scientist.


Even if the major tech companies limit their LLMs’ capacity to generate disinformation, bad actors won’t — and it is surprisingly easy to construct LLMs. A few weeks ago, a Stanford research team built a ChatGPT-level LLM for $600. Soon after, another group built one for $300.

In the very near future, most things (maybe nearly everything) on the internet will be garbage. A tempting conclusion is that we now live in an era of too much communication. Instead of spreading good ideas and spurring innovation, the internet has spurred distraction, dissent, and disinformation.

How will this play out? We don’t know but some scenarios seem likely. Unless the social media platforms shut down the bots, sane people will flee. What other rational response is there when your Facebook, Twitter, WhatsApp, Instagram, and TikTok accounts receive hundreds, thousands, millions of daily messages? For other parts of the Internet, perhaps a viable solution is a paywall inside which humans vet content for accuracy and usefulness. It will be expensive, but not as expensive as trying to navigate a vast unmonitored internet in which almost everything is BS.

Read more

about AI

Comments

Leave a Reply

Skip to toolbar