It's been 10 years since AlphaGo's Move 37. Would 2016-you be impressed or disappointed by where AI is today?
A decade after AlphaGo's historic win, AI can write code and pass exams, but daily life looks remarkably unchanged.
Ten years ago, DeepMind's AlphaGo made Move 37 against champion Lee Sedol, a moment that signaled AI could perform creative, expert-level tasks. Today, the AI landscape is defined by a paradox: capability has skyrocketed, but daily life remains stubbornly familiar. Systems like OpenAI's GPT-4 can write code and pass professional exams, while models like Sora generate photorealistic video from text. Yet, transformative applications like ubiquitous self-driving cars, advanced robotics, and mass-market AR/VR have not materialized as many in 2016 predicted. The economy and our primary device, the smartphone, are structurally similar.
The path to this point has been the most surprising element. The dominant breakthrough wasn't a continuation of AlphaGo's reinforcement learning techniques, but the massive scaling of large language models (LLMs)—essentially making 'autocomplete really, really big.' As DeepMind co-founder Demis Hassabis reflects, the creative spark from Move 37 catalyzed a different set of breakthroughs now converging toward AGI. The reflection posits that a 2016 observer would be simultaneously awed by AI's raw abilities in narrow domains and disappointed by the lack of pervasive, tangible change in work, transport, and daily tech. The hope is that the next decade will see AlphaGo's brilliance finally leave the game board and revolutionize scientific fields.
- Capability vs. Integration: AI models like GPT-4 and Claude 3 exhibit stunning abilities in reasoning and generation, yet have not triggered predicted societal shifts in transportation or robotics.
- The Scaling Surprise: The dominant AI advance post-2016 was not AlphaGo's reinforcement learning, but the unexpected power of scaling large language models (LLMs), a path few foresaw.
- A Decade's Reflection: DeepMind's Demis Hassabis marks the anniversary, noting Move 37's creative spark is now part of a broader convergence of techniques paving the way toward AGI.
Why It Matters
It recalibrates expectations for AI's societal impact, highlighting the gap between rapid technical progress and slow real-world adoption.