I’m usually not very interested in trending tech topics, mostly because I see an abundance of claims that state that you cannot miss knowing about this, or that you should discard everything you’re using and implement this new thing instead, which is the best thing ever, which has no replacement, which is the future of all things (until a new one comes up).
I tend to stay away from the crust of the wave, remaining at the shore. With my feet on the ground, I can observe both the new thing and how it affects its universe, what happens when the wave crashes and where people are left off. This has been my usual behavior with technology, and I proved myself that it’s a good one. I perceive I’m in a good spot, I like it here.
AI has been a boom recently, but it’s lacking the new factor. AI has been a thing since before the eighties (or 2000’s if we want to compare the current actual capabilities). I think all the attention it gained is due to how hardware, software and even data are being offered as a service now, concept newer than AI (early 2000).
AI, being it no more than hardware and software, experienced the effects of SaaS/PaaS. Particularly with AI, the computing power needed is very high and costly, which is also true for the software behind it, that comes in the form of very specifically applied algorithms to large amounts of data. Combining the two, we can say that AI is computation and data as a service.
Nothing prevented businesses to do the same calculations, predictions or content generation AI is now doing. Well, money did, or the lack of people with enough expertise to provide good and logical results. Which is another conclusion I reach: AI has its place where the quality of the content does not matter enough, otherwise, businesses would be paying each buck for getting quality results. But now, given it is relatively cheap to perform these tasks, businesses allow a higher error rate. Once again, we see quality decline at the expense of delivering quick and cheap results. I’ve seen folks on the internet comparing this to outsourcing, where you can’t really control the quality or accuracy of the output, and iterating over it gets exponentially more expensive.
About AI taking jobs…yes, it might (not AI per se, but people choosing outputs of unknown biases and quality over controlled and transparent ones), which is another reason to remind ourselves that learning processes is not a stable career path. History has proven all processes to be automatable, and if we want to have a long developed career over a certain topic or field, we should learn and study process generation.
I also think it’s too early to tell if businesses are making a good decision leveraging a portion of their processes to be performed by “AI”. My take is that the more you involve black boxes to a business model, the more confusing it gets in the long term, the more difficult it is to correct or modify and the more dependent the model becomes of the black box.
I think lots of money are being invested here because it makes business models cheaper, but not better. And on my experience, businesses become more successful with quantity over quality, whenever the definition of success is a larger net worth instead of providing sound products and an enjoyable, humanized working environment.
It started with hardware as a service, then software as a service, now calculations or content generation as a service. Who maintains those services? It’ll be a smaller amount of people, meaning a larger amount of bias, meaning an oligopoly where it wouldn’t be crazy to have sudden price increases to businesses that have already adopted AI on their processes.
What will happen when AI starts to be trained on itself, where data becomes more corrupt at each iteration? (like the chinese whisper / broken phone game).
What will happen when the dead internet theory stops being a theory?
Am I saying that we should ignore and disregard this topic? Not at all. It’s a good thing to learn, a good tool to have (specially if you have in your workflow automatable tasks -I still think it would be better for you to learn how to automate them-). It’s also being a hot topic for businesses, where you don’t want to be an outsider when it comes to job positions.
I predict an overall growth on the AI topic for the next ~5-10 years that will make businesses grow exponentially, where it’ll then plateau into a corrosive tool to have inside a business model that businesses will pay higher prices to understand the implications it’s having on their processes.
For now, I’m staying away from digging into the ethical implications this has (deepfakes and so on), but that might be the most important implication.
A big “we’ll see”. On my side, I’ll focus myself on learning the mathematics and logic behind it, while being around the topic and learning what it is and how businesses are using it.