Skip to main content

AI is gaining ground but these 2 boundaries should not be broken

Submitted by Lennart on

Past predictions about the limitations of AI have consistently been proven wrong.

The exponential growth in AI capabilities has brought it from research labs into everyday life, accomplishing many feats once deemed impossible. While limitations still exist, I would not bet against AI.

In the past

  • Reasoning and Complex Problem-Solving: Once considered impossible for computers, systems like IBM's Deep Blue beating chess grandmaster Garry Kasparov in 1997 proved this wrong.
  • Natural Language Processing (NLP): The nuance, idioms, and humor in human language were thought to be insurmountable barriers.
    • Early chatbots like Eliza (1965) showed initial steps.
    • Modern chatbots exhibit remarkable understanding of natural language instructions, inferring user intent and even anticipating needs, a key aspect of generative AI.
  • Creativity: The assertion that computers can't create has been challenged by generative AI, which can produce art and music. The speaker draws a parallel to human creativity, which is also influenced by prior experiences.

On the horizon

The speaker then shifts to current and future challenges for AI:

  • Artificial Superintelligence (ASI): AI that surpasses human intelligence in all domains, currently in the realm of science fiction.
  • Sustainability: Current AI systems are incredibly power-hungry and expensive to run.
  • Understanding: The question of whether AI truly understands the meaning of what it "says" or if it's merely simulating thought remains.
  • Judgment (Wisdom): AI currently struggles with making nuanced, ethical, or subjective judgments.

The Human Role

The speaker concludes by defining the complementary roles of humans and AI:

  • Humans should focus on "What" and "Why": Defining the overall macro-level goals, objectives, purpose, and meaning behind actions. This is where human intuition and higher-level understanding excel.
  • AI (particularly agents) should focus on "How": Once the "what" and "why" are established, AI can figure out the "how," automate tasks efficiently, and perform them in an optimized way.