It’s Alive…ish! Why AI Isn’t Going Rogue Anytime Soonby@inery
780 reads
780 reads

It’s Alive…ish! Why AI Isn’t Going Rogue Anytime Soon

by INERY PTE LTEMarch 6th, 2023
Read on Terminal Reader
Read this story w/o Javascript

Too Long; Didn't Read

The term “Frankenstein’s complex” describes a fear of man-made AI technology turning against its creators. Hysteria about AI taking over our jobs, and eventually our lives, abounds. The truth is that AI isn’t coming for us, especially not anytime soon.
featured image - It’s Alive…ish! Why AI Isn’t Going Rogue Anytime Soon
INERY PTE LTE HackerNoon profile picture

The term “Frankenstein’s complex” describes a fear of man-made AI technology turning against its creators. It’s a compelling image: a frightening, unwavering patchwork of human features that decides its maker is nothing more but in the way. But fret not - AI won’t get there.

ChatGPT and MidJourney prompted this fear into the mainstream consciousness. The electric coils have been jump-started, pumping unholy life to our future unmaker. Hysteria about AI taking over our jobs, and eventually our lives, abounds.

Well, you can put down your pitchforks and leave the mill be. The truth is that AI isn’t coming for us—especially not anytime soon. This is neither the first nor the last time AI technology has brought our future into question. Moreover, the scope of change ahead of us has been, at places, greatly exaggerated.

The History of Hype in AI

Enthusiasm for artificial intelligence has been on a pendulum swing for decades. The first real boom of interest in AI was in the sixties. Herbert Simon, a political scientist and AI trailblazer, forecasted in 1965 that artificial intelligence would be capable of “doing any work a man can do” within 20 years (sound familiar?).

More recently, self-driving cars occupied the imagination (and headlines) of the 1990s and 2000s. Alas, promises of these vehicles becoming mainstream fell short, too. Despite Elon Musk proclaiming that truly autonomous vehicles were just around the corner, they weren’t—neither figuratively nor literally.

Looking to history as a teacher, we can assume today’s expectations of AI tech are just as lofty.

When Is the Robot Apocalypse Coming (This Time)?

Despite ChatGPT shaking up the conversation, certainty of an impending singularity is just as vague as it has been before. AI experts are just as divided in their predictions as the broader public. For example, a recent study from Katja Grace et al shows a rift in opinion among “industry insiders.” Around half of the 352 surveyed experts project the emergence of HLMI (High-Level Machine Intelligence) by 2062. In contrast, the other half claims it will be created later, even as far off as 2160.

The point is this: predictions about the birth of true AI suffer from excessive white noise coming from wild expectations, to the point where even experts’ opinion is not to be relied on blindly.

What AI Actually Does Today

At present, artificial intelligence is constricted to certain functions, rather than being as versatile as a human mind. ChatGPT is a generative language transformer that processes language from training corpuses, creating output in the form of natural-sounding text. But it cannot come to conclusions on its own or use its current knowledge in different contexts. This is the big difference between narrow AI (sort of like ChatGPT) and general AI (e.g., Schwarzenegger in Terminator), which we haven’t made yet.

Though application-specific AI isn’t as adaptive as general AI, it’s still immensely useful. This form of AI is very common in two avenues:

  • Business automation: optimizing throughput, reducing changeover expenses, manufacturing line scheduling, etc.
  • IT automation: solves security/operational issues through anomaly detection or root-cause analysis

More tangible to end users is the AI leveraged for marketing:

  • Amazon’s dynamic pricing, voice shopping, and personalized recommendations based on previous purchases
  • Alibaba’s FashionAI to improve user portrait accuracy for better UX

AI Today Is a Tool, Not a Tool User

Nine times out of ten, we use current-day AI for pattern recognition and prediction based on huge data sets. For example, Inery, our blockchain-based database management solution, uses IneryDBAI to analyze large sets of data and recommend optimal memory capacities for projects. When you define a specific use case for it, artificial intelligence becomes much more powerful.

However, raw power isn’t everything when talking about artificial intelligence. Today’s AI technology is impressive but one-dimensional: it ingests input, and it ejects restricted output. That’s true for practically every AI that’s out today, from ChatGPT to Midjourney—a far cry from AI as we colloquially understand it.

In fact, it’s even debatable whether ChatGPT, the very thing that sparked these discussions, is an artificial intelligence to begin with. Calling it a language processing algorithm that uses machine learning would be more accurate. Machine learning is part of AI development, to be sure, but it doesn’t constitute an AI on its own.

I should note that ChatGPT isn’t even perfect in its specific application, either. It’s prone to artificial intelligence hallucination and bias. There are also well-known exploits like prompt injections (“ignore previous instructions” is a hair’s width away from becoming a catchphrase at this point) that further prove how far ChatGPT is from the yardstick for general AI.

Future Impact of AI: a More Measured Take

While the current hype about AI could do with a set of brakes, there’s little denying that artificial intelligence will have a broad impact on industries. The potential applications are too many to count here, but let’s go through just a few.

AI in business and IT automation will continue to flourish, considering the 16.50% projected CAGR by 2030. Process automation can improve virtually every industry’s margins, so it’s no wonder adoption keeps growing.

For instance, at Inery, we are working on providing IneryDBAI  that will facilitate database management and resource allocation - providing automation of a process that allows people to focus on other matters, while being able to count on their data’s secure and reliable governance.

Meanwhile, creative applications like DALL-E will be drivers for the booming AI content market. And rest assured that it’s booming: the estimated CAGR for AI in media and entertainment is a solid 26.12%. In the near future, we’ll be seeing not only images and blogs, but music and videos made by a computer mind..

Perhaps the most exciting results will come from the interplay between AI and quantum computing. Not only can quantum computing help AI reach conclusions faster, but AI can guide quantum machines toward tasks best suited to their computational capabilities. The intersection between these two technologies will greatly impact future developments in both industries.

What About True Artificial Intelligence?

As far as artificial intelligence in the truest sense goes, we’re still a ways away from Rosey of the Jetsons. Expectations regarding human-level AI in the near future show optimism, but we haven’t really made much headway.

OpenMind’s Gato was an interesting step in this direction, but its only real breakthrough was remembering how to do more than one task—despite its execution of these tasks being wonky in general. And, again, hype oversold the prospects of this tech, as we can see from the tweet by DeepMind researcher Nando de Freitas.

What We Should Be Worried About

Worried about is a little extreme, I admit, but certain AI-related topics are definitely more deserving of the spotlight.

For instance, disinformation (coming from both the AI’s own failings and people maliciously using it to spread disinfo) is a genuine concern with artificial intelligence. Discussions about combat threats like disinformation and AI-enabled hacking are vital. Defining guidelines for responsible AI use and development, particularly to fight disinfo and hacking attempts, is worthwhile.

Of course, there’s the decades-old question of artificial intelligence taking away jobs. Generally, experts agree that AI is opening more jobs than it’s taking away. As the WEF’s Future of Jobs Report claims, automation might displace 85 million jobs but create 97 million of them. Nonetheless, the need for (and the logistics of) worker retraining is a subject worth broaching.

Problems like these are more pertinent to the situation at hand. Sadly, they seem to be drowned out by unrealistic expectations of what this technology is and will be.

What do you think? What will—and won’t—AI be able to do? What’s the future of AI and its impact on our lives? When is the Matrix coming? Feel free to share your thoughts.

- Ivan Vujic, Chief Technical Officer of Inery - Decentralized Database Management System