Why Artificial Intelligence Keeps Me Up at Night
Back when I started my programming journey, AI was what I had the most superficial interest in just because it seemed cool and primed for innovation. Over time, as I became more and more exposed to different facets of computer science and engineering, I realized that AI was indeed very cool, but not really for me, and it's perhaps this newfound lack of enthusiasm that made me underwhelmed when ChatGPT made waves across the tech world in late 2022. I found it impressive, but I still felt like this wasn't going to change much of anything at all. Many others were in the same boat, showing how ChatGPT tried to prove that the square root of 4 was irrational, among various other gaffes you can do with some clever prompt engineering. There were flaws, and there still are many flaws, but there's now a nagging thought at the back of my mind that we're on the edge of a precipice and too few of us care to look down. My mind is foggy; it already feels like I cannot comprehend the scale of the disruption that is about to occur, and I'm worried.
I'm worried that I'm falling behind the curve for one, but that's not really what keeps me up at night. I'll probably spend some time exploring the technology in more detail, but my learning is within my control. What's not in my control is what the world will do when faced with all the ethical issues that AI brings. I want to go through a few of my concerns, even though the most I can do is scratch the surface.
On Consensus
Consensus is a scary thing because relying on it can undermine the search for truth, and this fear is exacerbated when truth cannot objectively be found. Once a critical mass is achieved for a particular opinion, it becomes enshrined as a fact, and with that comes the concomitant prestige of believing something most people believe. Relying on consensus in AI design can result in several issues unless certain safeguards are in place. Can AI evaluate an idea on its own merits, or does it reflect the whims and fancies of its creators? Can an AI break through the barriers of consensus? Every good idea starts out in the minority, and whatever type of artificial intelligence we're training should be as reflective, representative and objective as possible. The worst outcome I can see is bad actors poisoning the well and being placed in positions of power to influence others under the guise of normalcy and behind the veil of consensus.
When AI takes our jobs eventually, the least we can do is make sure that we ourselves retain our own intellectual autonomy, and can continue to fight for what we believe is right. In fact, what we believe is right should be the basis of what the AI believes is right. I would rather it be the case that society retains its axiological monopoly, so that it's AI that needs to constantly catch up to us instead of the other way around. In an ideal world, we should be the ones setting boundaries on ethics and aesthetics, and AI should be the one always lagging behind.
On Art
It's funny how we expected that the first jobs AI would replace are jobs involving physical labor. What many of us didn't realize is that physical labor (and hardware) is complicated. Nowadays, it seems like the greatest outrage over the use of generative AI is specifically within the visual art space, and I get it. Here you have this bunch of 1's and 0's looking through what you've made with painstaking effort and years of discipline and training, and immediately learning from it to spit out something similar soon after. The end result is definitely (for now) of an inferior quality, but maybe that's fine for the average person. Small art gigs are probably starting to vanish, as people just go to MidJourney or DALL·E for their basic art needs. I tried out creating a cartoon-style avatar using just Bing and a general description of how I look, and the result was good enough to probably use non-ironically, although for now I'd rather not associate my online identity with it simply due to its origin.
This is the industry that, I feel, will be disrupted the first and the hardest, especially for those who still need to make a name for themselves. Going back to the point on consensus, the incentive structure might be such that some artists are incentivized to create content that's deemed by general opinion or the powers that be to be too unsavory for AI, and that doesn't seem sustainable in the long term. The next step, then, would be to manage this disruption properly, perhaps by finding ways to harness AI as an assistant or tool of some sort, or by putting money where our mouth is and supporting actual artists if it's within our financial capacity.
Of course, secondly, there's the issue of ownership. How do we regulate consent when it comes to art being used to train a model? I'm no legal expert, but I'm assuming this is a conundrum we're still struggling with. It only seems fair for a cut of resultant revenue to go to artists, or for artists to prevent their art from being a part of a dataset unless compensation is provided upfront. However, this seems impossible for anything that has been shared online. "Right click+Save" is the bane of NFT and digital art enthusiasts alike.
I'd like to end this section on a more positive note. I believe that humans are biased towards other humans, and our tastes in art will change to reflect that. Furthermore, even if AI innovates on its own and responds to feedback to adjust itself, the interplay between the human experience and the infinitude of possibilities ensures that AI can never have a monopoly on artistic innovation for as long as we exist.
On Trust
Being able to recreate pictures, videos and audio might just completely upend what can count as evidence in a court of law. Maybe the lawyers among my readers can elucidate the current literature on the topic, but it seems to me that this can sabotage the pursuit of truth and magnify differences in power and influence. I would love to have a chat if you have an opinion on this.
On Further Progress
We've been here before, and we're still struggling to wrap our heads around what happened. Social media has indubitably been the most disruptive technology of the century, and many of us didn't realize it even though there was a initial huge backlash typical of the greatest technological innovations. This is yet another case of Amara's Law.
“We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” – Roy Amara
Social media has been the greatest source of ethical concerns within the tech space, and, in my opinion, we have failed on many fronts when it comes to managing them, and many of these echo the ones I've mentioned above. The formation of echo chambers across the political spectrum has resulted in less open discourse and more polarization and fake news. The creation of global digital marketplaces has probably resulted in a lot of failed ventures (not limited to the art space) due to the inability to compete. Memes are being stolen without credit being given. (That is a joke.) On the issue of ownership and consent, we have increasingly lost ownership over our own data and in many cases have transitioned from consumer to product.
Even now, I'm not convinced that society has really figured it all out when it comes to social media. Looking at how we have been dealing with all the related issues does not inspire much confidence. If we're not careful, AI will be the next groundbreaking technology that society needs to catch up to, and that's very scary indeed. Let AI catch up to us instead, singularity be damned.
I sound like some AI doomer, but I have to say that it's insanely cool how fast this technology is progressing, and I'm quite optimistic when it comes to tech and how it will improve our society. That doesn't mean I believe that we should just throw caution to the wind. We should look at everything that will come afterwards in the context of everything that has come beforehand, and move ahead in a principled and careful way. For now, the best I can hope for is that some very smart people are well ahead of me in thinking about these issues, and that you are now intrigued enough to think about them too.