ChatGPT’s Trial By Fire
Is ChatGPT sexist? Does the AI chatbot pose a danger to society? How accurate are ChatGPT's answers to complex questions?
Let's deal with the sexism charge first. Sharmin Ali is founder of a deep-tech startup called Instoried. She complained to journalists Arpika Bhosale and Aastha Atray Banan in Sunday Mid-day: "At the core of any open AI is a man - not a woman - who inputs 'variables'. They tell the AI what words in a sentence are positive or negative, such as happy, sad, angry or joyful. Now add AI's auto-learning into the mix, and the machine is assigning its own values to new words and framing its responses to users accordingly. These then include the inherent bias of the man, who assigned the first variable, a value which is now multiplied on an unimaginable scale."
Turn now to the second concern: does ChatGPT pose a danger to human society? Elon Musk, founder of Tesla and SpaceX and reluctant owner of Twitter, thinks it does. Along with a bunch of AI experts, Musk has called for a six-month moratorium on developing more advanced AI systems like GPT-4. The Future of Life Institute issued a letter last month signed by Musk and over 1,000 other experts. The letter warned: "Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable."
Is that paranoid? Many AI specialists think it is. ChatGPT has built in safeguards to block pornography, homophobia, racism and, yes, sexism. ChatGPT comprises more than 100 trillion parameters and is constantly improving, upgrading and updating data.
The latest avatar of ChatGPT was unveiled recently. Bloomberg called it Chat GPT's "iPhone Moment" which poses a "threat to Google's sprawling tech empire."
Exactly how? Bloomberg explains: "Just as we were getting used to the idea of ChatGPT writing emails and conducting research, Open AI has upgraded it with capabilities that make it even more of a threat to big tech companies such as Google, ChatGPT's new plugins, released last week with relatively little fanfare, are a major leap forward for the technology. In a nutshell, it is going from merely generating text to taking action on the web, turning it into the type of powerful virtual assistant that Alphabet Inc.'s Google and Apple Inc. have been trying to build for years."
That brings us to the third question: how accurate is the output from ChatGPT when asked a relatively straightforward question but one with an historical, economic or scientific context?
To test this, I first asked ChatGPT: "What is the probability that life as we know it or as we don't (for example, plasmoid) exists in the universe? Is distance the only reason we've not made contact with extraterrestrial life?"
In seconds, ChatGPT responded: "As an AI language model, I cannot provide a precise probability of the existence of lie in the universe since we do not have enough evidence to make an accurate estimation. However, given the vast size and age of the universe, it is highly probable that life exists beyond Earth, whether it is similar or vastly different from life as we know it.
"Distance is certainly a significant factor that makes it challenging to detect extraterrestrial life, especially if it is beyond our current technological capabilities. The vast distances between stars and galaxies mean that even with our most advanced telescopes and instruments, we can only observe a tiny fraction of the universe. It is also possible that other civilisations have emerged and gone extinct long before we have developed the capability to detect them."
A little underwhelming but accurate enough. So I asked another, more contemporary, question: "India is now the world's 5th largest economy. Its GDP of $3.36 trillion will overtake Germany and Japan by 2026 and 2028 respectively to become the world's 3rd largest economy. What geopolitical changes will this engender?"
Once again, ChatGPT gave a super-quick response: "With a larger economy, India would have greater leverage in global affairs and would be able to exert more influence in international organisations like the United Nations. India could potentially become a more prominent player in global issues such as climate change, trade, and security. India's rise could potentially put it in direct competition with China, which is currently the world's second-largest economy. India and China have had a long-standing border dispute and have often competed for influence in the region. A stronger Indian economy could lead to a more assertive Indian foreign policy and potentially heighten tensions with China.
"Overall, India's rise to become the world's third-largest economy would be a significant development with far-reaching geopolitical implications. While it is impossible to predict exactly how this will play out, it is clear that the rise of India will have a major impact on the region and the world as a whole."
Like any innovation, AI chatbots are a work in progress. It's wise not to be taken in by the hype or be unsettled by the paranoia. ChatGPT is overall a great innovation but it needs a reality check. I wrote six months ago in these pages that the "metaverse" was simply an upgrade on 3-D which has been around since the 1990s. Sure enough, Meta founder Mark Zuckerberg, who changed Facebook's name to Meta to push the metaverse, has quietly stopped talking about it.
Will ChatGPT meet the same fate? Absolutely not. This is a revolutionary innovation with loads of potential. It will improve with time. It will also have to sort out copyright issues with artists and content creators.
But is ChatGPT a threat to human society, as some fear? That's even less probable than encountering extraterrestrial life in the universe.
