Advancements in narrow AI are taking place nearly every week the year of this writing. That trend as with anything in the industry is going to continue increasing exponentially as we refine existing technology and create new ones. So then what better a time then this year to list some of the narrow AI’s that are going to be making waves in the coming four years. From augmenting industries to entirely upending them here are my predictions for some of what is to come (and a few that have already happened!)
Number 1: Narrow AI Will Expand Economies Not Eliminate Them
As anyone who lived through 2022 with an internet connection can tell you, there has been quite a bit of commotion about AI generated images this year. In what I would say is the first AI phenomenon the general public witnessed, the open source image generator DALLE-Mini took the online meme communities by storm. Based off the code of DALLE and trained on open sourced data, it was truly the first time the general public could simply go to a website, type in a prompt, and get 9 images of various AI generated memes and nightmare fuel. The results as evidenced were questionable at best when they weren’t garbled static but I would almost go so far to say that was the very reason it succeeded. The internet loves a good joke and what better irony than the AI overlords destined to replace those of flesh and bone being unable to so much as draw a sky without some weird error or computerized issue in the image. It was all fun and games until the much anticipated DALLE2 AI was released. Even though it was released as an invite only beta once the door was opened a crack the internet forced it open. Suddenly images that were eerily good started cropping up everywhere. A lot had captions giving credit to the AI, but even more did not. Memes were now being made that weren’t making fun of how bad the AI was, but instead were just being made as they always were but with the help of the AI. Even some popular creators began using DALLE2 to generate the thumbnails for videos they’d create. What many people (and artists specifically) soon came to realize however was that in the joke the internet had created, they were the punchline.
Then not even six months distant from the beginning of going mainstream, StabilityAI released a brand new algorithm to the public: Stable Diffusion AI. Based off of technology of the same name this AI improved every single facet of DALLE2. It offered vast improvements on the already questionably good models before it, but one feature alone made it stand out: it could run on home computers. Breaking from the questionably authoritarian trend of trying to keep technology from the public OpenAI and Google had set, the entire model was released to the public free of modification or censorship. At the time of writing this is where we currently are. Anyone with enough technology savvy intuition can spin up and run their own AI capable of creating images every bit as good as a human artist.
With this tidal wave of new art being created in seconds with mere text prompts the online communities of artists have been in uproar. Under the cries that “AI can’t create it just steals art from humans” and “It isn’t even that good” there is a common line of thinking not being said aloud: AI is going to take my job as an artist and I don’t like that. Artists from a purely historical standpoint are underpaid, underappreciated especially in their lifetime, and oftentimes even if popular when alive will see their works resold for vastly more than they charged to create it. Only in the most recent twenty or so years has the prospects of an artist become better, and now AI has come to take that away. Or so these communities think. This line of thinking is understandable but slightly misguided.
When automation comes to an industry that does not mean the manual simply ceases to exist. Quite the opposite actually. When technology enters an industry it expands the entire economic market share of it. A great example of this would be the baking industry. For ages people would go to the local baker to pick up bread products. Then automation hit the industry and resulted in factories that can create millions of loaves a day, pack them, and send them out. Just because that factory is made however does not necessarily mean the bakery in that same town will be out of business. People that want bread from the local baker they know will continue to buy from them, even at a premium to the price of the factory bread next door! However, who previously couldn’t buy from the baker either because he simply couldn’t produce enough bread or those who wanted could simply not afford it, will now have the luxury of being able to purchase the factory created bread. The factory did not simply eat into the existing market share to carve out a spot, but rather created a new market entirely.
Of course this is a fairly primitive example. It does however show what happens when new technology enters an industry. The manual may become a niche of a much larger new market, but it never goes away in its entirety. Humans are social creatures and that alone means some people would prefer buying more expensive products created by another as opposed to one inexpensively made by machines. In the case of narrow AI such as photographic art, music, and soon video and cinema even though it will make production of such things extremely cheap (if not entirely free) it will not replace conventional artists as many in those fields are concerned about. In fact this may truly be the best time in history to be an artist. Why? Simple! Once literally anybody can make a movie, the people that dedicate their lives to manual creative expression will find themselves in that niche market. Talent will be appreciated and the market for human made media may even see a premium pricing come to it such as what happened in many other markets with industrialization. For examples consider the textiles, woodcrafts, soaps and detergents, meat production, agricultural, and yes even baking industry! Industrialization came to every single one of these markets, and every single one of them still has a place for manually created products! People pay premiums for free range eggs and meats from local farmers. They tout the textiles they purchased from the locals of a town or from the website of a small business as something unique and special. The very fact they are imperfect and created by an actual human being becomes a valued feature! This is why artists need not fear about their futures. In the coming decade with art and other industries that are not unskilled labor people will virtually always have the opportunity to keep their career paths. They may well become the niche of a vastly larger market and economy, but as I’ve shown fairly plainly above that may serve to their benefit not detriment.
Number 2: Narrow AI Will Become Ubiquitous
Now that we have the comic argument of AI taking all jobs from humans addressed and out of the way we can focus on what truly will be impacted. As I mentioned before the concept of narrow AI is that it excels in a singular task. It is a far cry from AGI but still will allow for things at the moment unable to be fathomed by the vast majority of humans on this planet. What is to come in the next decade is so fantastical in fact most anybody not directly in the field of AI research would brand it as sheer lunacy. To avoid such a mantle I’ll start off slow and work my way out in the realm of probabilities. We already covered that text to image narrow AI has gone from running on supercomputers to your home GPU in under a year. The next exponential leap forward is already in the works. The Stable Diffusion AI specifically went from an already infinitesimal 8GB down to 4GB, with one of its main developers stating they had designs to eventually whittle that number down to an absolutely mind boggling 100MB. Note that just because the model size decreases does not mean that the quality of the results does. In fact in the example above there was no quality decrease whatsoever between the 8GB and 4GB. As models continue to be created and improved upon we indeed will hit such a threshold of sub 1GB, and whence we do a fascinating series of events will trigger.
AI will not only be available to put in everyday applications, but practical! Video editing software will come in built with a narrow AI model that can on command change a background, change the hair color of the person in the shot, alter the weather and lighting, among millions of other use cases (Edit: I couldn’t even publish this before it was created). Camera applications will have text prompts allowing you to create any kind of facial rig imaginable simply by typing it out and watching it work in real time. Products that generate music on command, or can create an entire animated TV show based on your general questions about genre or you providing an entire script, balance the accounts and logistics of an entire company, extract revenge on your coworker, alter your voice in real time, and the list goes on.
It goes without saying this is as much a blessing as it is a curse. While the fears of misinformation and disinformation are severely overinflated, the potential misuse will be an issue everyone will eventually have to contend with. For every fun face swap there will be a high profile political one. For every app letting you create your own tv show there will be a platform editing their information to conform to social norms, even deleting historic or documentary type information when it conflicts. Which brings us quite nicely to my next prediction.
Number 3: Governmental Structures Will Polarize.
The people astute enough to notice the period at the end of that statement will have realized this is not simply a prediction. Contrary to many people who dare write their thoughts about the subject of AI, I am not one to so immediately dismiss everything about the present and proclaim socialism and Universal Basic Income (UBI) will be the only way forward. It is but one path of many. The common thread we do agree on is this: governance is going to undergo a tectonic shift as a direct result of Narrow AI and the soon to follow Proto-AGIs. While there is plenty that is already transpiring that may well result in governments being toppled unrelated to AI, those are externalities to this statement that might influence an outcome of this shift but not cause it. Permit me to explain. Nationalism for better or worse is once again on the rise throughout the entire world. If the timing of these nationalist movements coincides with the revolution in commercially available Narrow AI (as it indeed appears to be) we could see those of steadfast opinions dominate offices and be able to sustain the promises they make. Systems once thought impossible to remove humans from could become nationalized for fascism of socialism overnight and with the help of AI actually be sustainable.
A quick example would be a postal service. It has a very static layout, a small set of controlled variables, and an unchanging objective being to deliver A to B. A Narrow AI could be tasked with running the financial operations of that system to avoid fraud, another to run the factory floor and coordinate with the other factories, another to sort and label the packages, and still another to load and unload them into vehicles going to those destinations. Even the vehicles themselves could be run by a Narrow AI given its route by still another AI. In such a system there is no single AI that controls the entire system. It isn’t “smart” or alive in any sense of those words. It is simply an algorithm that can account for variables that a normal machine could not. The most fascinating aspect of this is that each Narrow AI can preform the task it was trained on and nothing more yet despite that could fully remove human fallibility and corruption from an entire system.
I’m sure you’re beginning to see the prospects here. One of the greatest shortcomings of governance is corruption, so what happens when the lofty promises of a system such as fascism or socialism can actually be delivered upon? Simple: they are validated. People especially in the next decade are demanding changes to government and radical solutions to problems are going to be attempted on both sides of the political spectrum. Whichever side of the political spectrum in a nation is first to embrace AI and use it to create radical reform will be the one set to rule that nation for decades to come. I must note not all of this subject is to be seen as pessimistic. There will be a lot of good to come up the upcoming struggles for nations’ souls. A lot of corruption will be eliminated as a direct result of AI. National conflicts will breed competition, innovation, and above all progress in technology. What may be most interesting about this subject in particular would be the evident truth that the political authoritarianism governance systems of the 20th century will be given a second round to fight for superiority. It will be an interesting time to observe even if I can’t really say I favor the probable outcome.
Number 4: Progress Itself Will Increase Exponentially
These previous three points all bring me to perhaps the most evident claim I’ve made in this writing. The progress of AI is one of exponents. As Narrow AI paves the way for Proto-AGI every field it touches will be impacted. Material sciences, physics, chemistry, genetic editing, fusion, artificial pregnancies, the arts, language engineering, agriculture and so many many other fields will be revolutionized in ways thought unimaginable. There will be a great many issues as a result of this to be sure. I after all only just recently wrote how technology and progress are not synonyms. The stage has been set to have ethical and moral calamity befall humanity sometime in the next century if not much sooner. Some nations will return to their religions and traditions, others to nationalism and national identity, and still others to race, cults, or any other number of things as ways of finding meaning and commonality to justify this expansion of technology forcing a shift in world view. The current system of globalism will crumble as a result of this, if it manages not to fall apart before then. There are thousands of possible outcomes to this. Most of them may not be as desirable as the proponents of it would suggest. Ultimately Narrow AI will give humanity a brief grace period to learn and accept AI in everyday life before the much larger advancements occur. I hope humanity or at least some of humanity makes the right choices in this time frame. AI has the power to allow for societies virtually free of government to coexist, to socialist dreams of uniform control over every industry being held by the citizens, to the horror filled dystopias portrayed countless times before of hyper authoritarianism and tyranny over the population of a nation the likes of which has never before been seen in human history.