Generative AI: A Paradigm Shift in the Making?

At this rate of progress, the world may look a little different in just a few years

The media narrative on “Artificial Intelligence” (AI) has become somewhat cynical. In the absence of any notable breakthroughs that happen to go viral on social media, you tend to find the coverage to be skeptical due to the lack of widespread adoption, and criticism of the quality and reliability in real-world use.

Obviously, there are some massive caveats here.

AI is a broad term with a booming scientific and technical community that has been churning out progress for some time. When looking at more specific use cases for AI, like machine learning and its application in more niche fields ranging from scientific research, medicine, surveillance, and so on, we have already seen a profound impact and its potential to do even more in the future.

The disappointment, I suppose, lies in the lack of mainstream applicability and relevance in everyday life.

We seem to be waiting for the world we’ve been promised from “insert basically every sci-fi movie here,” while nervously joking about an impending doom as a result of our new AI overlords that would then soon follow.

Now, I don't plan to cover the ethics or technicalities of AI and its societal impacts in this article; I believe it's better to leave that to experts and individuals who are more fully immersed in the field.

Rather, I want to go over what we’ve seen recently, particularly from OpenAI’s DALL-E 2 and ChatGPT demonstrations, the impact they had, and what this could mean in the near future.

General AI still eludes us, but is generative AI enough?

It’s worthwhile clarifying that as impressive as DALL-E 2 and, more recently, ChatGPT appear to be, they do not represent a general form of artificial intelligence.

“Artificial General Intelligence,” or AGI, would require software to be able to understand, learn, and apply any intellectual task that we, as people, can.

While ChatGPT appears to have something akin to a personality, displaying awareness and even apologizing and correcting itself whenever we point out an error or mistake, we should note that the underlying software model is based on natural language processing that essentially mimics the English language by generating responses that mirror how we use it.

Through being trained on unfathomably large datasets and using reinforced learning techniques, the chatbot is basically able to predict what to say next with a high enough degree of accuracy that it appears to be using the language like an actual human being.

But of course it isn’t perfect. 

We’ve already seen numerous “Aha, see! I found a way for ChatGPT to look stupid” moments, particularly around its logic and consistency, and this makes sense once you understand that it’s generating responses based on a statistical likelihood rather than actually understanding your request and forming an intellectual response.

While this may start to sound like a bit of a letdown, in reality, I still believe that this is a technological and productivity marvel.

Imagine having a resource that is effectively trained on the entirety of human knowledge up to a certain point and being able to instruct it to regurgitate something specific in a way that can be helpful and useful at a moment's notice.

In theory, this could be a game changer. We've already seen so many creative uses of ChatGPT that range from mind-blowingly useful to just plain silly; all thanks to the ingenuity of people.

If it’s so good already, do we really need AGI?

I’m not really sure that this is the right question to be asking. Rather, the reality is that AI research will continue in pursuit of it, just like in any other scientific field, in the interest of advancement and knowledge.

But as more breakthroughs are being made each year, with progress now becoming exponential, I believe we may be closer than we think to something that may just be considered AGI, or at the very least nearly emulate it.

We may end up seeing different AI models that are better at certain things than others combined into a single experience that, for all intents and purposes, could genuinely trick us into believing that it is sentient.

Applicability, accessibility and adoption

And that’s the crux of the matter, in my opinion. General intelligence or not, if a “combined AI package” can allow for it to be genuinely and reliably useful, then it may not really matter at the end of the day to the average person.

(Please do note that I say this from a mainstream perspective rather than an ethical or scientific point of view, which are far more nuanced.)

While not a fair comparison, the proliferation of smartphones over the last two decades is an interesting parallel to draw from.

Many of the technologies behind smartphones had already existed for some time prior to the introduction of the original iPhone. Similar attempts had even been made from other manufacturers to usher in the “next big thing.”

Apple, however, can be credited with piecing all of the puzzle pieces together (even the imperfect ones) at a time when consumers were able to “get it,” and supply chains were able to move it into the hands of enough people to reach critical mass in a timely manner.

They understood the importance of demonstrating meaningful applicability in a person’s life and then making sure that the product was accessible to the masses.

Looking back, there were so many gaps with the iPhone, but everyone saw the potential, and human ingenuity kicked in to plug the gaps. Over time, adoption reached a point where we cannot imagine a world without smartphones today.

Could the same happen for AI?

I think so. As humanity continues to advance, it has become harder for us to bridge our accumulated knowledge and the practical application of it. We need some help.

You only need to look at the disarray we see across social media when it comes to polarizing opinions arising from increasing amounts of misinformation and a general unwillingness to fact-check.

Smartphones transitioned us from a user experience requiring keyboards and mice to one based on touch. Increasingly, voice control has become more common, and there are still many who dream of a world where we’ll be able to control our devices with our thoughts.

While the latter is still some time away, the use of voice is looking like the next frontier of how we interact with technology and information.

However, unlike typing on a keyboard, which developed alongside computing devices, language and verbal communication between humans have evolved over hundreds of thousands of years.

For a voice-led user experience to take over, it has to emulate the way we naturally communicate in everyday life between people.

If you think about it, a lot of the groundwork has already been laid. 

From “forced” confrontations with generic chatbots that provide user experiences that resemble chatting with your friends to quirky voice assistants from tech giants that amaze you with one feat only to completely “face-palm” on the next one, the puzzle pieces, so to speak, are beginning to appear, with each piece representing a different application of AI. 

And while there has been progress over the years, let’s be honest: each piece on its own, especially the current generation of chatbots and voice assistants, is still a far cry from what we expect them to be.

So could generative AI, like DALL-E 2 and ChatGPT, be one of the final pieces for a more complete AI experience that we’ve been waiting for?

Well, if it can reliably help us get things done faster and better without adding additional steps, then yes, this could very well be what we’ve been waiting for.

But technology and cool use cases won’t suffice. There’ll need to be a business ecosystem to show applicability, provide accessibility, and drive adoption amongst consumers.

Final thoughts

Of course, there will always be considerations that can and will affect progress. 

Discussions need to be had around the ethics and regulation of AI to prevent abuse and ensure that consumers and rights-holders are protected. Software and hardware improvements are also needed, as the cost of generative AI still has a ways to go before it can be commercially viable at scale.

But once consumers truly begin to see the value of something, market forces and collective human ingenuity will accelerate progress and plug any gaps in bringing it to life.

So how close are we to getting there?

Perhaps closer than we think—take a look at the Chartr infographic below that shows ChatGPT’s uptake and reception on the Internet in December 2022.

The demonstration went viral, with ChatGPT reaching a million users at a record pace, beating out iconic services like Instagram and Spotify by miles.

People are beginning to “get it.” The carrot on the stick is now more obvious than ever.

But we also shouldn’t get ahead of ourselves.

While the world let out a collective gasp in awe of the iPhone back when it was first released, it still took years for true proliferation of the smartphone as industries and institutions played catch up to consumers who had caught onto the “next big thing.”

See you in the next one!