.png){kind=small}
As the AI rage continues to take the world by storm, debates and opinions on whether AI will become sentient, overtake humans and eventually enslave them, are gathering momentum. However, this is not the first time such a prognosis has been made. The advent of factories that followed the industrial revolution, the invention of computer and its evolution into multiple form factors, the development of Internet in its simplest form to the intricate Web 3.0 model that drives much of digital commerce today and the transformation of automobiles from manually operated to driverless systems, have all seen machines taking over repetitive, manual tasks in one way or the other and/or significantly improving the way humans live/work. However, this time around the pace of development, the scope of application and the level of alignment (or misalignment) force us to think that sci-fi movies such as Terminator or Her, could well be a reality, in the not-too-distant future. Why I say so?
One, the ongoing pace at which AI is developing, or supercharging is insane. The biological evolution of human intelligence took nearly 4 billion years, whereas the evolution of AI is happening at a frenetic pace. Algorithms that merely automated mundane tasks 15 years ago can now hack human brains, swing election results and/or modify human behaviors, and all this being just the tip of the iceberg. The adoption too is equally steep. According to McKinsey research, enterprise AI adoption has grown 2.5X in the past five years the average number of AI capabilities embedded in organizations has nearly doubled from 1.9 in 2018 to 3.8 in 2022. The rapid strides made by modern AI (RPA, NLP, NLU etc.) over the past 50-60 years are dwarfed when compared to the volume, scale and speed at which the current Large Language Models (LLMs) that power Generative AI operate at.
Two, most of the AI systems known to us till date such as smart speakers, digital assistants, and autonomous vehicles have focused on performing a single task using a pre-defined data set and operating within a pre-defined range. As such, they merely reproduced human ideas and lacked creativity, reasoning and imagination. Generative AI, on the other hand can actually exhibit human intelligence, learn by itself, create new ideas and make complex judgements under uncertainty. OpenAI’s ChatGPT can write codes, poems, ad copies, essays and it has even passed a horde of simulated exams, the most notable being the Uniform Bar Exam. The multi-functional application of generative AI and the ability to create downstream use cases, makes it all the more powerful.
Three, historically much of the AI advancements have taken into consideration the societal reset, regulatory action and the economic re-adjustment to counter the knock-down effects. This is not to say, there haven’t been negative consequences following their rollouts, but they have been far more controlled and paced out, giving some elbow room for governments and corporations to make necessary course corrections. Today’s AI development lacks the alignment checks and guardrails, with creators, at best offering lip service on adhering to safety, transparency, governance and privacy standards.
Image by Gerd Altmann from Pixabay
The potential productivity gains from generative AI, particularly in the fields of medical research, drug discovery, customer service, legal/investment advice, and marketing, appear highly promising. However, alongside the promises, there are also significant threats to consider. The fact that experts like Geoffrey Hinton and Yoshua Bengio, who are often regarded as the Godfathers of AI, have quit their jobs and turned into AI doomers, underscores a tale that deserves serious attention.
Among the various areas where generative AI is likely to make a substantial impact, the following appear to be particularly prominent and imminent:
The dangers to democracy, art and creativity: Generative models will be used to create fake or misleading information, such as deepfakes, fake news, and fake reviews. Their ability to automate misinformation campaigns and perform micro-targeting at scale would give autocratic governments greater power to suppress dissent and implement surveillance systems. At a time, when democratic systems are under pressure from political extremism and far-right fundamentalism, this could have detrimental impact on political engagement and electoral outcomes. By creating cloned human voices, hyper-realistic images, videos and audios in seconds, generative AI can become a powerful tool to mislead voters, impersonate candidates and undermine elections at a scale and speed not seen before.
LLMs will adversely affect originality and novelty in art, literature and creative works. In the pursuit of productivity and hyper growth, content generation will rely primarily on ubiquitous, previously AI-generated content as source material, recycling errors at a massive scale. Going by recent news reports, plagiarism, copyright and privacy infringement, and data breaches will assume centerstage in courtroom battles.
Economic dislocation: The job losses resulting from deployment of generative AI will be widespread, but will likely be concentrated in knowledge services. According to a recent study by Accenture, 40% of all working hours could be impacted by LLMs such as ChatGPT-4. This is because language tasks account for 62% of the total time employees work. With learning curve lagging behind development curve coupled with lack of support for reskilling and transition and absence of government-backed social security, this could lead to economic dislocation with second order effects on mental health and social engagement.
Widening inequality: As the AI arms race intensifies, not just corporations but governments will increasingly be vying to control the AI systems. Generative AI will be the wild card for governments to run state surveillance, execute nation-state sponsored cyberattacks, and establish territorial sovereignty. Rich nations that are currently the hotbed of AI development will harvest gains from AI to control poor nations, influence geopolitical relations and fund technological innovations. The result would be widening economic inequality and social injustice, rendering poor countries as data colonies.
Deepening climate crisis: A lesser evident outcome of the exponential growth in generative AI will be felt on global warming and climate change. According to industry estimates, ChatGPT emits 8.4 tons of carbon dioxide per year, more than twice the amount that is emitted by an individual and for training GPT-3, it consumed 1,287 megawatt hours of electricity and generated 552 tons of carbon dioxide - the equivalent of 123 gasoline-powered passenger vehicles driven for one year. The enormous resource consumption and carbon emissions by these LLMs will likely worsen the climate crisis and aggravate environmental degradation as more companies jump into the fray to develop and deploy LLMs. And while the OpenAI chief might be betting big on nuclear fission to counter the ecological impact from LLMs, it might just sound like a case of robbing Peter to pay Paul.
Battle for intimacy: The advent of social networking platforms helped intensify the clamor for attention and ‘digital’ presence. With generative AI, the battle will shift to intimacy. A generation that has been born on social media, and currently courting with Metaverse and brain-computer interface (BCI) will seek greater levels of involvement with AI ‘partner’ who will fulfil all their desires, ignore their flaws and easily be able to shape-shift their moods/interests, much in the DIY style that appeals to their instant gratification behaviour. With the institution of marriage failing and relationships cracking, AI may become the ideal partner with no strings attached, the only one being you throw open your data vault.
It will be truly fascinating (and perhaps scary) to watch how these developments unfold over the next decade, and what experiences they would create – more importantly if they would be creative or destructive. The alarm bells heralding a dystopian future are growing louder day by day with several experts and intellectuals making a clarion call for a pause on further advances in generative AI. At the moment, the AI juggernaut seems unstoppable for the creators are more answerable to shareholders than to humanity. Fast-forward 50 years, and AI could well be the last thing humans ever create. A Frankenstein Monster – creator and controller of human intelligence!