AI is too dangerous to be left with the techies
Tech giants are pushing the limits of AI development though they are yet to fully understand its inner workings
By Hari Kumar
The year is 1440, and the Catholic Church has called for a six months moratorium on the use of the printing press and the movable type. Imagine what could happen if commoners get access to books. They could read the Bible for themselves and society would be destroyed.
This was a tweet sent out by Yann LeCun, a French computer scientist and the chief AI scientist at Meta. This sarcastic poser came after the Life Science Institute issued an open letter signed by thousands of tech experts, including Elon Musk, Apple co-founder Steve Wozniak, and others, demanding a temporary halt to AI experiments. LeCun opposed any such ban, saying it would be like “taking a step back in terms of the progress of humanity”.
LeCun is not alone in opposing a ban on further research into regenerative AI, which has been proceeding at an explosive pace in the last few months. Microsoft founder Bill Gates and Google CEO Sundar Pichai have also been vocal in their opposition to calls for a halt in the current AI progression.
Still, the majority of experts do not dispute the claim that new AI platforms will have a profound impact on everyone. In a survey conducted by Stanford University for its annual AI index report, considered by some as the industry's state of the union report, 73 percent of respondents agreed that “AI could soon lead to revolutionary societal change”, while 36 percent believed it would cause a “nuclear-level catastrophe”.
The race to create more and more powerful AI programs is inevitably drawing comparisons to the atomic race during the Cold War era, but there are two crucial differences between the two. While the United States and the Soviet Union were racing to create more destructive nuclear weapons, the competition was between government-run facilities. In the case of artificial intelligence, governments have been in ringside seats, while the actual slugfest is being carried out between private companies.
The other major difference between the nuclear race and AI development is even more concerning.
AI systems are becoming so complex that researchers are unable to determine how these programs arrive at certain conclusions. Despite this, more and more powerful AI systems are being developed and incorporated into software that spans different fields, such as defence, medicine, and finance.
The hi-tech firms in the race spend billions on making their AI programs more powerful, but very little attention or funding is devoted to understanding what goes on inside the black boxes of their systems.
As the AI revolution is predicted to generate trillions of dollars in business, there is very little incentive for leading companies to hold back. Even those who demand caution have been caught with their hands in the cookie jar. A few weeks after calling for a halt on AI research, reports say Musk himself is getting ready to build a platform similar to ChatGPT.
The proliferation of AI power has already begun, and a recent report in The Washington Post stated that one tool, described as “Stable Diffusion for perverts”, has been downloaded 77,000 times already. People are also sharing ways to use AI technology to edit real images, including removing the clothing of fully dressed women in photographs. Meanwhile, a group in the US is offering fake bomb threat calls using AI-generated voices for as little as 50 US dollars.
Administrators of chat boards like Reddit report seeing an increased amount of AI-generated text appearing on their popular forums, while deepfake images and videos of world leaders, celebrities, and religious figures are becoming increasingly convincing.
While the debate about AI continues fiercely in the tech world, there is little movement to involve experts from other areas in the ethical aspects of these developments. As Sandra Wachter, an expert in technology regulation at the University of Oxford, recently wrote, this is too important an issue to be left solely in the hands of tech experts, as AI technology permeates almost every field.
“You really need to talk to lawyers, to people who do ethics, to people who understand economics and politics,” says Sandra Wachter, an expert in technology regulation at the University of Oxford. “The most important thing is that those questions are not decided among tech people alone.”
Tech wizards are often skilled at opening up new vistas in the cyber world but may not fully comprehend the impacts of their creations. Take, for example, social media platforms like Facebook, which were originally meant to connect people and bring them together. However, they have ended up being used as tools to organise riots, tear societies apart, and influence the future trajectory of even the most powerful nations.
As revenue and growth become the bottom line for many of these platforms, they have tweaked their workings to turn users into virtual slaves of the algorithms they have developed as a means to boost their company fortunes.
Given this, leaving the development of artificial intelligence in the hands of a few private firms is a worrying proposition. It becomes even more concerning when commentators like Ezra Klein of The New York Times reveal that their interactions with those in the ‘Cerebral Valley’ of San Francisco have converted them from being supporters of AI development to individuals who believe there is an urgent need for restraint.
In a recent column, Klein explained his change in attitude:
“Since moving to the Bay Area in 2018, I have tried to spend time regularly with the people working on AI. I don't know that I can convey just how weird that culture is. And I don't mean that dismissively. I mean it descriptively. It is a community that is living with an altered sense of time and consequence.”
He later said that many involved in creating powerful AI machines do agree that there is a 10 percent chance that their creations could go out of control and disempower or even extinguish humanity. Yet, there is still no responsibility pinned on the creators of these programs who continue to seek more powerful versions.
Some experts now suggest that we should look at how the aviation industry implemented stringent regulations and allowed the private sector to build planes, instead of the nuclear race during the Cold War.
They argue that artificial intelligence is a similar technology with the potential to bring unlimited benefits to humanity, but it can also result in disasters if not handled properly. They propose that no manufacturer should be allowed to release their AI products to the market unless they comply with stringent standards.
The question now is who will step up to bell our fat cats.