ARTIFICIAL INTELLIGENCE
AI Will Solve Everything
Or not
Everyone and their mother are talking about AI these days. Including me.
I’ve recently made my case regarding how our overreliance on technology has caused us to believe that it can magically fix everything. But it really can’t. Technology is limited by our resources, and we have been far too wasteful with our one-time inheritance for many decades now. The escalating climate crisis is putting pressure on us, and the only viable way forward is to reduce emissions, as fast as we can.
However, late-stage capitalism in its desperation to maintain endless growth on a planet with finite resources, will peddle anything it can as a possible way to maintain the status quo.
This is delusional.
Excluding technologies that are guaranteed to not be a solution, but at best a mitigating factor to our climate woes (such as carbon capture and storage), or even a pipe dream that is always two decades away (such as nuclear fusion), our last, best hope is what we call Artificial Intelligence.
Let’s consider for a moment the fact that LLMs (Large Language Models) aren’t “true” artificial intelligences, in the sense that those models are, in effect, extremely advanced… parrots. LLMs have no ability to understand what they are saying, even if in most cases their replies seem concise and logical. And this might not change by simply scaling up the processing power and the data volumes that are fed to these systems, a process that becomes exponentially more costly in energy and water use as these LLMs increase in capacity.
In other words, an LLM can surprise us with its apparent eloquence, although it can botch (of all things for a “machine”), math and will even confidently offer wrong information as true. It can write poems or pretend to be anyone we tell it to, including a virtual boyfriend or girlfriend, but there’s one very important thing that it cannot do: propose entirely new ideas.
This doesn’t apply to other kinds of AIs, such as deep learning models based on neural networks that can play games and do it in completely new ways. It’s not that simple, of course. Even the best such deep learning system is what is called an Artificial Narrow Intelligence or ANI. In other words, it can be trained to play chess or Go or even a videogame, such as Starcraft very well, better even than the best human players, but it can’t play all of those games together and can do little more than that. And an ANI is still vulnerable in unexpected ways. Apparently, a relatively simple algorithm can exploit AlphaGo’s blind spots and beat it, but this very same algorithm can easily lose to amateur-level human players.
In other words, an ANI can be taught to foil human players or another ANI, but not both. It’s a specialized tool.
Neural networks are seemingly still our best bet to attain the Holy Grail of AI research that is called Artificial General Intelligence (AGI). In other words, true AI that can think for itself or, to be more precise, can execute any mental task a human can and do it better, faster or both.
This would be the first, deciding step towards a theoretical ASI or Artificial Superintelligence. In theory an AGI could either by design or on its own improve itself rapidly enough that it could evolve into an ASI and bring about the supposed technological singularity.
This would be an abrupt leap in technology with unforeseen consequences for humanity, including a possible extinction event.
When all else fails, a supposed godlike AI might be our only chance to get us out of our current predicament: the vicious circle of technological progress that is at the same time killing our habitat, destabilizing our climate, sustaining our current population and, perhaps, offering us a way to escape the confines of our planet (and the certain demise that comes with remaining here forever).
Or it might destroy us that much sooner.
Imagine this, though: let us suppose that the newborn ASI proves to be benign to humankind and biological life in general. Imagine that, despite its completely alien way of thinking, and the vast intelligence gap that would separate us from such an entity, this ASI comes to understand the value of life in a universe that isn’t exactly friendly to it as a whole. What if it tells us that the only way to escape our encroaching doom is to limit resource use to the absolute minimum to sustain ourselves and advance only the technological areas that are crucial to our long-term survival as a species?
What if, even with its help, fusion power and faster-than-light travel were still decades away and would require significant sacrifices from our side, namely the end of our consumerism and wastefulness? Sure, it might have a better plan on how to bring about these changes with less disruption of our economy and our society. But would that affect our fear of change itself? Are we capable of abandoning the greed and exploitation that have been part of human progress since the first human laid claim to a piece of land?
Would we accept this solution or ignore it and go on waltzing towards the precipice?
Wouldn’t it be the height of cosmic irony to develop an impossibly intelligent machine that could only tell us what we already know?
