Artificial intelligence: A new dawn for ARC or history repeating itself?

In the last edition of the newsletter we talked about the ARC framework (read more here)—a three-phase cycle of Augmentation, Replacement, and Creation, that can be used to illuminate the trajectory of major technological advancements.

These days, conversations inevitably gravitate towards artificial intelligence—an ever-evolving dance of myths and facts that has captivated our minds like no other, presenting promises that appear both extravagant and feasible.

It's only natural that my readers are curious about AI and the ARC framework and have a few follow-up questions.

Does AI fit the ARC framework? If yes, which of the three stages—Augmentation, Replacement, and Creation—are we currently navigating through? Will AI resemble past technological advancements, or does it present a unique departure from the norm?

Let's explore.

Is AI Following the ARC Framework?

The short answer is yes!

The ARC framework describes technologies that are transformative. "Indiscriminate change and unimaginable growth" is how I described their impact on humanity. Does AI fit that description? This might sound like a rhetorical question, but let’s examine.

The current ARC of generative AI really started on November 30th, 2022, when OpenAI introduced ChatGPT. Right out the gate it achieved an astounding feat—reaching 100 million users in just two months, faster than any other consumer app before, including TikTok, which took nine months to hit the same number.

Since then, GPT-3 had 175B parameters, but now we have models with over a trillion! We've progressed from text-only to multimodal tools like Runway Gen-2, enhancing AI systems with text, images, and audio.

We know large language models (LLMs) at times produce inaccurate information—as seen with Google's Gemini and Microsoft's Copilot providing misleading details about Super Bowl LVIII in February 2024—but they are undeniably making significant strides. The reliability of these models has also improved, though it’s still a mixed bag. For example, GPT-4 achieved a top 10% ranking in a simulated bar exam, marking a remarkable advancement from GPT-3.5, which ranked in the bottom 10%.

As AI becomes increasingly integrated into our daily lives, I'm reminded of Arthur C. Clarke's renowned words, "Any sufficiently advanced technology is indistinguishable from magic." The progress in AI indeed seems magical.

So, the answer to our question is a yes. AI is indeed a transformative technology. And if our hypothesis is accurate, it will adhere to the three phases of the ARC framework.

Are We in the Augmentation Phase of AI?

The response to this question is considerably more intricate than one might initially perceive.

Generative AI is revolutionizing industries by automating tasks, enhancing creativity, and aiding decision-making. For instance, platforms like Copy.ai help create tailored ad copies, email campaigns, and social media posts. In customer service, tech like Rasa, Dialogflow, and IBM Watson Assistant powers chatbots that handle routine questions, freeing up human agents for complex issues. Nference streamlines medical report generation in healthcare, while BenevolentAI boosts drug research. Financial sectors use DataRobot for risk assessment and QuantConnect for algorithmic trading. In manufacturing, Visio Ingenii improves quality control and Llamasoft enhances supply chain logistics. Creatively, Runway ML produces art, music, and poetry, with tools like Figma supporting design projects. With new applications emerging daily, there's a clear shift towards using generative AI tools to enhance productivity.

It is undeniably in the augmentation phase of the three stage ARC cycle.

Or is it? "AI is expected to handle about 95% of the tasks currently carried out by agencies and creatives," exclaimed the headlines, quoting Sam Altman. OpenAI has consistently articulated its goal of developing AGI, or artificial general intelligence. In a recent study by researchers Pamela Mishkin and Tyna Eloundou at OpenAI, an analysis of 19,265 tasks across 923 occupations revealed that LLMs could perform tasks typical of 80% of American workers.

Furthermore, research on the impact of technology on the workplace depicts a landscape where some roles remain resilient, but others face imminent replacement. Sectors like supply chain optimization have witnessed workforce reductions. Fields such as legal research, financial analysis, and predictive maintenance on fixed assets are witnessing rapid transformations as generative AI tools begin to supplant human roles.

Although instances of such transitions are fewer compared to the positive use cases of augmentation, they are still noteworthy. Are we already transitioning into a Replacement phase?

But wait, the plot thickens. Consider the case of Insilico Medicine. This pioneering biotech company harnesses generative AI to create a new era of drug discovery. Their platform identifies target molecules, generates innovative drug candidates, and forecasts clinical trial outcomes. For example, they developed a treatment for idiopathic pulmonary fibrosis using AI, and the FDA approved their AI-crafted small-molecule treatment for solid tumors. Traditional methods for such achievements could have incurred costs exceeding $400 million and taken up to six years. Yet, leveraging generative AI, Insilico accomplished these feats at one tenth of the cost and one third of the time—reaching the initial phase of clinical trials just two and a half years after project initiation.

Such breakthroughs herald a new healthcare era, where AI-driven drug development will transform patient care and create a burgeoning new industry at the nexus of technology and medicine.

This starts to feel like the Creation phase of the ARC framework.

So which phase of the ARC framework are we in? All three!

Generative AI is moving so fast that we are starting to see all three phases of the ARC framework overlap and run almost simultaneously.

Which bring us to the next question.

Will AI Behave Differently than Previous Transformative Technologies?

The short answer, again, is yes!

In just a few years, AI is progressing through the three phases of the ARC framework simultaneously, though at varying speeds. This stands in stark contrast to past groundbreaking technologies, which often required decades or even centuries to complete a predominantly sequential run through the ARC phases.

And this difference in journey is due to a key trait—generative AI learns rather than just following programming.

Before conjuring visions of a dystopian future, consider this: GenAI isn't programmed, it's taught. This dynamic of learning and teaching sets AI apart from any previous technology.

Should we approach AI differently?

Absolutely! Despite AI lacking human traits, the key is to interact with it in a profoundly human manner—teaching it akin to nurturing a child, with attentiveness and accountability.

Keeping pace with AI can be demanding, but so is raising a child. Just as we would never write off a child’s future because of an outburst or tantrum, the same goes for AI. Bias, hallucinations, and a lack of explainability can become teachable moments. As teachers, we set boundaries for kids, but we never want to quash their spirits. Similarly, there is a place for AI regulation, but not so much that it represses our ability to creatively leverage its applications. We can remain agile, respond to issues as they arise, and adjust our approach accordingly. 

Generative AI is proving to be both familiar and unique. It is following the ARC framework, but it is driving the three phases of Augmentation, Replacement, and Creation almost simultaneously. We have created it just like previous technologies but it's learning on its own. Like previous technologies, we are the ones to guide its future, but not as programmers but as teachers.

This naturally brings us to the most significant question: How?

As a leader within my team, organization, or community, how can I influence the future of AI?

Where is the compass to help us go forward in this journey?

Stay tuned as I address these topics in the next article of this series.