Artificial Super Intelligence and the Humble Human

 

“Science fiction warned us about machines that think; it never warned us how ordinary that moment would feel.”

For centuries, intelligence was the one domain humans assumed would remain uniquely ours. Strength could be mechanized, memory could be stored, speed could be amplified, but thinking felt different. Today, that assumption is quietly dissolving. Artificial Intelligence is advancing not in steady steps, but in accelerating leaps, challenging our definitions of reasoning, creativity, and even understanding itself. And yet, when we look closely, the distance between computation and humanity remains both profound and revealing.

Artificial Intelligence is progressing in leaps and bounds, practically on a day-to-day basis. In many areas, it already outperforms humans by a wide margin. To evaluate how close AI comes to genuine human reasoning, we rely on benchmarks such as ARC, which attempts to measure AI performance in comparison to humans.

ARC (Abstraction and Reasoning Corpus) is a benchmark created by François Chollet to measure general fluid intelligence in AI—the ability to learn new skills and reason abstractly without prior domain-specific training. It has evolved into multiple levels: ARC-AGI-1, ARC-AGI-2, and ARC-AGI-3 (preview). Current leading AI models perform far below human level on the hardest levels: humans solve nearly 100% of tasks, while even the best AI models struggle to reach 50% on ARC-AGI-2 and under 20% on ARC-AGI v2.

The current performance landscape looks like this:

• Humans vs AI: Humans consistently solve 100% of ARC tasks, highlighting the massive gap in reasoning ability.
• AI Progress: While ARC-AGI-1 was nearly “solved” by large models, ARC-AGI-2 exposed their limitations—brute force and scale are insufficient.
• Breakthroughs: GPT-5.2’s ~53% on ARC-AGI-2 is a major leap, but still far from human-level reasoning.
• Future: ARC-AGI-3 will test interactive reasoning, pushing AI beyond static puzzles into dynamic environments.

In summary, ARC is the North Star benchmark for AGI, measuring whether AI can truly think, generalize, and adapt like humans. Despite significant progress, leading models remain far from human-level performance, especially on ARC-AGI-2 and beyond.

While researching this topic, I was reminded of Data, the humanoid android from Star Trek: The Next Generation. Curious, I asked Microsoft Copilot how Data might score by comparison. This was its response:

ARC-AGI-1 (basic abstract reasoning puzzles)
Humans: ~100% | AI: ~75–87% | Data: Effortless.
Data routinely demonstrates the ability to infer rules, patterns, and transformations far beyond simple grid puzzles. He would solve these instantly, likely faster than humans.

ARC-AGI-2 (harder, resistant to brute force, requiring deep abstraction)
Humans: ~100% | AI: ~53% (best models) | Data: Near-perfect.
Data shows fluid intelligence: he learns new games (like poker or strategy simulations) without prior training, adapts to alien logic systems, and reasons abstractly across domains. He would likely match or exceed human-level performance here.

ARC-AGI v2 (leaderboard variant, where top AI scores are <20%)
Humans: ~100% | AI: ~15% (best) | Data: Essentially flawless.
Data’s ability to generalize across unfamiliar contexts—whether deciphering alien languages, solving engineering crises, or improvising in diplomacy—suggests he would achieve near-100% here, like humans.

ARC-AGI-3 (interactive reasoning, dynamic environments)
Humans: ~100% | AI: Not yet tested | Data: Beyond benchmark.
Data thrives in interactive, real-world reasoning: commanding starships, negotiating with adversaries, and adapting in real time. He embodies the kind of general, embodied intelligence ARC-AGI-3 aims to test.

At this point, I listened again to a discussion between Data, Geordi (Chief Engineer), and Dr. Pulaski (Chief Medical Officer), which reminds us how far away even the fictional android is from being human.

The context is that Data and Geordi enter the Holodeck to play Sherlock Holmes adventures. Since Data knows all the stories, he keeps jumping the gun and spoiling the fun. Geordie got frustrated and abandoned the game. They both came to the café and discussing the frustration about playing with Data and Dr Polaski was listening to this conversation.

“What we were doing You are wasting your breath ,lieutenant. saying that to data is asking a computer not to compute” Said Dr Polaski.

“Am I so different from your doctor.” Asked Data

“Are you able to cease speaking on command. In medicine I am often faced with puzzles that I do not know the answer.” Said Doc

“She's right there. you always know the answer. to feel the thrill of victory there has to be the possibility of failure and where's the Victory in winning a battle you can't possibly lose.” Georgie observed

“Are you suggesting there is some value in losing” Data asked

“ Yes yes that's the great teacher . we humans learn more often from a failure or a mistakes than we do from an easy success . not you. you learn by rote. to you all is memorization recitation .” Said Doc

‘I don't know about all that. Deductive reasoning is one of data strengths” Georgie commented

“Yes and Holmes is too. But Holmes understood the human soul; the dark flecks that drive and turn the innocent into the evil, that understanding is beyond data.” Said Pulaski

“Now you're just being unfair doctor” Quipped Georgie

“I don’t think so lieutenant. Your artificial friend doesn't have a prayer of solving a Holmes mystery that he hasn't read’

Being a Star Trek aficionado, I then asked Copilot to compare Data with another Next Generation–era AI: the Emergency Medical Hologram from Star Trek: Voyager.

Data would consistently outperform the Doctor in abstract, cross-domain reasoning.
The Doctor would rival or surpass Data in medical problem-solving and human interaction. His emotional growth gives him an edge in empathy-driven reasoning, which ARC-AGI-3 (interactive tasks) is designed to capture.

In short, Data is the embodiment of general intelligence; the Doctor is the embodiment of specialized intelligence evolving toward generality. Together, they illustrate two pathways AI could take. one built for universality, the other for depth and human connection.

AI may bring super intelligence soon, not a normal human being. And will this super intelligence empower the human or annihilate the human is the question.

The idea behind this post is not to be judgmental, but to invite you to join me on a quest on what more is to being human- better human. Atma with a link to Paramatma ?

“Intelligence may be measured in problems solved, but humanity is revealed in the problems we struggle with.”