let's talk about ai

this is a topic ive been itching to write about for months now, and finally have the motivation to get to in light of recent events

ill start by saying this: unfortunately a ton of ai discourse is polluted by people who have no knowledge of the field whatsoever, and are fueled by their instinctive ideological response to it (especially amongst the anti-ai demographic). that’s not to say that you need a phd in a topic before you’re allowed to have an opinion on it; in fact, I think everyone SHOULD be opinionated, especially in socio-political matters. however this makes it so that in debates or discussions on the topic, they tend to come out looking worse despite having the more regulated and morally just stance

your reason for being against ai CANNOT be because it sucks and lacks in quality. it needs to have a moral foundation to it. this was most prominent when the ai gen art discourse was going strong. i saw heaps and heaps of people clowning on the art style and the quality of work made by ai models, and i agree, it usually is ugly and jarring to look at. but what if it wasnt? what if it could make amazing art which was indistinguishable from real works? what then? funnily enough, we’ve seen many examples of this in the recent past. there have been people who’ve made gorgeous art that was entirely generated by ai, which was commended by people, who eventually got exposed. without fail, every time, people switch up after the news breaks and retreat into post-hoc rationalizations about how the work was “soulless” all along

if we want to build a coherent, robust, and morally grounded opposition to the current trajectory of artificial intelligence, we have to stop retreating into aesthetic snobbery

we must abandon the comfortable delusion that human beings possess an untouchable monopoly on beauty, logic, or scientific insight. the navier-stokes breakthrough showcases that

man do i have a lot that i can say about it

i have spent plenty of time deriding the self-appointed prophets of the ai boom. i still harbor little patience for their brand of corporate messianism. yet recent months have made one thing starkly clear: wherever your sympathies lie across this divide, the benefit of the doubt is no longer an option. the frontier laboratories steering this technology have squandered any claim to our collective trust

that’s not to say that the technology behind it should be undermined though. i think we’re at the point where the long standing cope of llms being stochastic parrots is no longer applicable. despite the financial and industrial constraints that surround it, criticizing the epiphenomenal quirks of a growing architecture reads as pure cognitive dissonance to me; the arguments proposed against it are bound to be outdated either way. and even in the event that llms do prove to be a dead end, i highly doubt that the frontier labs would take that as a cue to give up and work on other endeavours. there’s too much poured into the promise of agi for there to ever be a clean back out

i think that there also seems to be a great deal of misconception on what agi even looks like and spells for us. for agi to be “the end of humanity” as is feared by people, the way it would manifest doesn’t need to be a hyper-dystopian society where robots gain sentience and start waging an all out war on humans. the world is fragile, and is comprised of a ton of moving parts: water, food, weaponry, oil, healthcare, technology; any of these getting majorly revamped can drastically affect how we function as a society. the future really is limitless, and i dont trust the current pioneers of the models to have the amount of authority over it as they do, especially since they’ve proved time and time again to make rash and financially motivated decisions despite standing for and promising the very opposite (one culprit stands out in particular here :p)

out of all the gloom and doom that this entire landscape is encapsulated by, one of the stark benefits that i see blooming out of it is its impact on research and education. if there was ever a time to push for a reform, it’s now. i honestly wish i’d started my cs degree 5 years later so that hopefully the incredibly mind numbing method of cramming a whole textbook’s worth of syllabus at surface level to reproduce it in exact terms during exams would be dead during my tenure. this could honestly be a whole new post in and of itself; it’s ridiculous to me that we still follow this tired and dysfunctional architecture

as for research, i don’t think people are thinking nearly big enough about what ai could actually mean for science. most of the conversation is still about jobs, chatbots, writing, coding, images, productivity etc. but some of the biggest changes may come from ai helping us solve problems humans have been idle on for decades or even centuries and simply haven’t been able to crack

think about how much of modern life ultimately traces back to a relatively small number of humans figuring out something fundamental about reality. newton understanding motion and gravity. einstein changing our understanding of space and time. people figuring out electricity, evolution, germ theory, dna, computation. a human notices something and dedicates their life to thinking about and working on it - connects ideas nobody had connected before, and suddenly entire branches of civilization become possible

now imagine having systems that can read basically the entire scientific record, analyze enormous datasets, run simulations, compare millions of possibilities, notice patterns humans missed and do it across EVERY field at once. there are mathematical problems people have spent their entire lives trying to solve (the recent one being the catalyst for me to write this post). there are parts of the ocean we barely understand and species we haven’t discovered. huge portions of the rainforest we haven’t properly catalogued. biological systems doing absolutely insane things that we still don’t fully understand,,, for example, there’s a jellyfish that can essentially revert back to an earlier stage of its life cycle instead of continuing toward death (isn’t that so surreal to think about :o). animals that regenerate limbs and organs. creatures that survive radiation, freezing, crushing pressure or almost no oxygen. nature has already solved an unbelievable number of problems that humans haven’t. we just don’t necessarily understand how to translate those solutions into something useful for us

what happens when ai can map millions of species, proteins, genes and biological mechanisms and start finding connections between them that no individual scientist could ever hold in their head? maybe something an animal evolved 50 million years ago eventually teaches us how to regenerate damaged human tissue. maybe some obscure organism in the rainforest contains a mechanism that becomes a cancer treatment. maybe studying aging across thousands of species reveals something we’ve completely missed about why our own bodies age. we’ve never had the faculties to properly speculate about these possibilities until very recently

physics is an even bigger question mark: dark matter, dark energy, quantum gravity, what actually happens inside black holes, whether there are forms of matter or physical phenomena we haven’t discovered yet, whether our current understanding of space and time is even close to complete. maybe time travel as we imagine it is impossible. maybe it isn’t. the point is that we genuinely do not know what a more complete understanding of physics eventually makes possible. imagine explaining gps or nuclear power or an iphone or landing something on mars to someone living before newton, there’s no way to detail them in the context of what was known until then

we are extremely used to living inside the discoveries humanity has already made, so we forget there could still be enormous layers of reality we simply haven’t figured out yet. ai doesn’t magically know the answers - we still need experiments, evidence, scientists and the physical world to tell us when an idea is actually true. but for the first time we may be building a tool capable of searching through the space of possible ideas at a scale no human mind ever could,,, and if even ONE of those discoveries ends up being on the level of electricity, antibiotics, relativity, or computing, the world on the other side of it could be almost impossible for us to imagine from where we are now

my biggest fear is the corporatization and bastardization of this technology. the arms race to gain a monopoly over intelligence, the desire to sell it at will and commodify it, and the mutual competition that the frontier labs operate on lead me to be skeptical about whether a morally sound and universally beneficial outcome lies ahead