Article: Could AI Take Over the World?

Photo Source: Unsplash

Recent advances in generative AI, such as OpenAI's ChatGPT, have enthralled the general public with their remarkable natural question-answering, translation, and document and image generation capabilities. The ease and comfortable tone with which these chatbots converse on almost any topic gives the impression that these machines have a consciousness—they feel alive. The propensity of people to anthropomorphize non-human animals and even objects such as machines and vehicles helps us to ascribe human qualities to chatbots today. The human-like conversation with the rapidly improving AIs naturally leads to the age-old speculation around when and if a machine superintelligence or general AI might emerge that exceeds all human intelligence across every discipline.

Such a machine would presumably use its super intelligence to improve itself by writing new code that would only exponentially accelerate its intelligence. Such an intelligent machine could take over the production of robots that would expand its control of the physical world, much like the workers in an ant or termite colony work in the service of the queen. Such an embodiment of AI in robots would require a magnificent amount of energy and resources. Today, the International Energy Agency (IEA) has determined that large generative AI machines in data centers worldwide consume 1.0-1.5% of the world's electricity, and the IEA expects that to double in the next two years. (iea.org) The complexity of a super-intelligent robot that manages to build and sustain itself while rapidly expanding its intelligence may seem impossible at this point. Still, it does not stop experts from speculating what will happen to people once human intelligence gets eclipsed by our machines.

Not surprisingly, experts across the AI machine learning world still disagree on when and if a general AI will emerge and how it would affect humanity and the planet. The Institute of Electrical and Electronic Engineers (IEEE) recently collected the opinions of twenty-two AI experts from academia, industry, and non-profits to answer two questions. (spectrum.ieee.org) The first question asks, "Is the success of GPT-4 and today's other large language models a sign that an AGI is likely?" the second question asks, "Is an AGI likely to cause civilizational disaster if we do nothing?" The authors of the article "The AI Apocalypse: A Scorecard" assembled the scorecards from media clips and assigned answers to the two questions based on the experts' writings and other media. Interestingly, for the first question of whether generative AI portends a general AI, 60% of the experts said "No," and 36% said "Yes." Regarding the second question regarding general AI leading to a disaster for human civilization, only 20% predicted yes, half the remaining experts answered "No," and the rest answered "Maybe."

Most experts did not see general AI as imminent or as the fall of human civilization. For example, Yann Lecun, the Chief AI Scientist at Meta, formerly called Facebook, cannot see a future where large language models such as GPT will ever lead to general AI because the models cannot acquire common sense or human-type understanding. Along the same lines, Professor Gary Marcus of NYU predicts that it will take enormous effort to get a machine to reason about our world. Other AI experts may not see general AI arising. Still, they attribute real danger to today's large language models, perpetuating human bias baked into the data that trains the models and the potential for dangerous misinformation getting propagated by the emerging LLMs. Of those that predict a general AI emerging, four see a disastrous outcome that will not allow a second chance for humanity. The most extreme being the total extinction of humanity. 

The general AI that destroys humanity would have to reverse the trend in history that centrally controlled and designed systems such as economies tend to fail. The failure often comes from ignorance of natural market forces that drive the distribution of goods and services. The nature and complexity of building, maintaining, and feeding a general AI and the necessary robotic support managed by a central super intelligence appear to be doomed when looking at analogous systems. Imperfect data and imperfect systems do not suggest constructing a vastly superior intelligence capable of world domination and control.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.