Tracking AI vs Human Skills in Real Time

How Close Is AI to Human‑Level Skills?

88.000000%

Right now, we estimate AI is working at about --% of an average person’s capability.

We weight equally seven pillars: Reasoning, Vision, Coding, Memory, Creativity, Emotional Intelligence, Autonomy (cfr. infra).

Skill-by-Skill Breakdown (Best in Class – June 2025)

Reasoning 85.6% ↑
Vision 88.2% ↑
Coding 99.4% ↑
Memory 99.7% ↑
Creativity 99% ↑
Emotional Intelligence 86.0% ↑
Autonomy 60.0% ↑

Details

Benchmarks updated as of June 2025. Values shown represent the best AI performance recorded in each category.

  • Reasoning: MMLU-Pro – O3 85.6% (Gemini 2.5 Pro Exp. 84.1%, GPT-4o 83.5%)
  • Vision: MMBench (multimodal vision) – While GPT-4o has strong vision, specific "best in class" MMBench scores vary by subset. Qwen 2.5 VL 72B is a strong contender in open-source multimodal models. (Retaining GPT-4o 88.2% as a general strong performer, but noting the nuance of MMBench)
  • Coding: HumanEval – LLaMA 3 99.4% (QualityFlow (Sonnet-3.5) 98.8%)
  • Memory: Gemini 1.5 Pro – 99.7% recall on context up to 1M tokens (near-perfect recall up to 10M tokens reported)
  • Creativity: Torrance Tests – GPT-4 at the 99th percentile compared to human benchmarks
  • Emotional Intelligence: EQ-Bench – O3 8.6/10 (Gemini 2.5 Pro Preview 06-05 8.4/10; GPT-4o 8.4/10)
  • Autonomy: SWE-Bench – Refact.ai Agent 60.0% (SWE-agent + Claude 4 Sonnet 56.67%)

Note: For Creativity, the value is expressed as a percentile (not a direct percentage).

Where Could AI Go Next? (2021‑2030)

Dark bars show the story so far (2021‑2025). Colored lines sketch three possible futures.

Countdown to Human-Level AI

The moment an AI system matches humans at all general tasks like reasoning, learning, and problem-solving.

--

Target: 1 Jan 2027 (UTC)

AI Readiness Indicator

This metric offers a snapshot of current AI capabilities, evaluating how close today’s systems are to achieving advanced autonomy.

63%

This score is based on a combination of factors: autonomy, memory, reasoning, and emotional intelligence. It reflects current research but does not imply imminent full autonomy.

Countdown to Artificial General Intelligence (AGI)

Artificial General Intelligence (AGI) refers to the point where AI can understand, learn, and apply knowledge across any domain, potentially surpassing human ability.

--

Target: 1 Jan 2040 (UTC)

The timeline for AGI remains highly uncertain. While some experts predict breakthroughs in the 2030s, others believe AGI is still decades away. Recent surveys among AI researchers suggest a 50% chance of AGI arriving between 2040 and 2060. If achieved, AGI could revolutionize science and industry, but also presents significant challenges for safety and governance. For now, these estimates are informed projections, and the global research community is actively working to ensure that any future AGI is developed responsibly and benefits all of humanity.

Market Growth & Talent Demand
  • Global AI market this year: ~600 billion (2025)
  • Looking ahead: Could reach $1.81 trillion by 2030 (35.9 % CAGR)
  • AI workforce: About 100 million new AI related jobs by 2025
  • Income premium: People with AI skills earn roughly 56 % more
Industry Adoption
  • Corporate usage: 78 % of companies already use AI (2024)
  • Strategic priority: 83 % rank AI as a top priority
  • Managers vs. staff: 33 % of managers use AI often, versus 16 % of frontline staff
  • AI in healthcare: >1,000 FDA authorized AI/ML devices (2024)
  • Autonomous mobility: 250,000 driverless paid taxi rides every week in the U.S.
Latest Milestones
Latest Milestones
  • ✅ Gemini Pro Vision (May 2025): Near-perfect image recognition
  • ✅ Claude 3 (April 2025): Handles 100k+ token contexts
  • ✅ Open-source LLMs (March 2025): Now match top proprietary models
  • ✅ AI in healthcare (2024): FDA approves more AI-based diagnostic tools
  • ✅ AI-generated music (2024): Tools like Suno and Udio create full songs from prompts
  • ✅ NVIDIA Blackwell GPU (2024): Breakthrough in AI hardware acceleration
  • ✅ Tesla Optimus Gen 2 (Late 2024): Humanoid robot walks and performs tasks autonomously
  • ✅ Meta’s Llama 3 (2024): 400B+ parameter open-source LLM
  • ✅ Stable Diffusion 3 (Early 2024): Advanced open-source image generation
  • ✅ Microsoft Copilot (2023): AI-powered assistant widely integrated into Office
  • ✅ Google DeepMind’s AlphaDev (2023): AI discovers faster sorting algorithms
  • ✅ AI-powered satellites (2023): Onboard AI for real-time Earth observation
  • ✅ AI co-piloted fighter jet flew 17 hours (2023)
  • ✅ ChatGPT: 100 million users in just two months (early 2023)
  • ✅ AlphaFold: Predicted structures for 200 million proteins (2022)

Projection: GPT-5.1 (Fall/Winter 2025): Might beat human-level reasoning on MMLU

Global AI Landscape
  • Model production (2024): U.S. 40 models; China 15; Europe 3
  • Investment: U.S. firms poured in $109 billion (2024)
  • Hardware efficiency: +40 % performance‑per‑watt gains
  • Public optimism: 80 % positive in China vs. 40 % in North America
  • AI workforce growth: +32 % YoY increase globally (2024)
  • Top AI patent filers: China leads with 60 % of global AI patents (2023)
  • Training costs: Frontier models now exceed $100 million in compute spend
AI Forecasts
  • Economic impact: Could add +$15.7 trillion to global GDP by 2030
  • Jobs & skills: 170 m new vs. 92 m displaced by 2030; 39 % of core skills set to change
  • AGI timeline: 50 % chance of human‑level AI somewhere between 2040 and 2061
  • Deepfake detection: may become undetectable (even by AIs) before 2030
  • Drug discovery: Expected to speed up dramatically, cutting timelines and costs by 2030
  • Quantum & AI: Breakthroughs likely in the early 2030s, boosting training & simulation speeds
Regulatory Developments
  • EU AI Act: In force 1 August 2024; full rules phased in by 2026
  • U.S. approach: Executive orders & voluntary frameworks; still no comprehensive law
  • China’s rules: Real‑name verification & mandatory labeling, effective 1 September 2025
  • Global trend: Over 40 countries now drafting or updating AI legislation
  • Compliance risk: Estimated 25 % of AI deployments in 2025 could face regulatory issues

Quantum advancement

believed % progress to main break-trough point (1000 logical QuBT)
0%
One of the biggest wild cards for AI’s future is its intersection with quantum computing. Quantum computers harness the quirks of quantum physics and could one day solve problems far beyond the reach of today’s fastest supercomputers. Researchers have already begun exploring ways to use quantum machines to supercharge AI – and the early signs look promising. In 2024, for example, IBM scientists showed a quantum computer dramatically speeding up a data‑clustering task. Looking toward the late 2020s, we expect more breakthroughs like this. If quantum computing matures, by 2030 AI models might train much faster or tackle enormously complex simulations (think climate models or drug discovery) in real time. Two revolutionary technologies could end up amplifying each other.

Everyday AI in the 2030s: Over the next few years, expect AI to become as commonplace – and as invisible – as electricity. By 2030, nearly every new piece of software is predicted to include some AI component. Your home will get smarter: thermostats will learn your habits, fridges will suggest recipes, and virtual assistants will coordinate your day. Self‑driving vehicles could be a regular sight on highways, cutting accidents and easing traffic. In healthcare, wearables might flag issues before we notice symptoms, enabling earlier treatment. On the flip side, AI‑driven cyberattacks may rise – which means we’ll need AI‑powered defenses, too. Governments could turn to AI for public services, from automated tax filing to AI tutors in schools and predictive city planning. By decade’s end, AI is expected to weave even deeper into daily life, often behind the scenes, making chores easier and opening doors we haven’t yet imagined.
P(doom) Speculated ~1% in the next 100 years span