Agi As A Tool Maker

I’ve been thinking about Schopenhauer’s distinction between knowledge of perception and knowledge of conception, how that relates to AGI, and wanted to see what others think. Here’s a quick breakdown:

Schopenhauer’s Distinction

Knowledge of Perception (Understanding)

• Comes from our senses.
• Involves a mental process of forming a hypothesis about the cause as some external object behind our sensations.
• Results in an intuitive, rather than purely abstract, mental image.
• According to Schopenhauer, this is the only way to generate new knowledge.

Knowledge of Conception (Reasoning)

• Abstract, verbal, or symbolic knowledge.
• Serves mainly to store and transmit what we’ve discovered through perception.
• Doesn’t (on its own) create truly new insights about reality—rather, it refines or rearranges what we’ve already observed.

Examples from History

Galileo’s Telescope: Without building a new tool (the telescope) to extend his perception, could he have discovered Jupiter’s moons or mountains on the Moon? Abstract reasoning alone probably wouldn’t cut it.

Large Hadron Collider (LHC): A huge experimental apparatus that led to the discovery of the Higgs boson. Was there any way to confirm its existence by pure thought experiment alone? Most would say no—someone had to perceive (via sensors, detectors, etc.) new data.

Cosmic Microwave Background Radiation: In 1964, Arno Penzias and Robert Wilson were using a radio antenna to study the sky. They kept detecting persistent static, which they tried to eliminate by every method they could think of (even cleaning out pigeon droppings!). Eventually, they realized the noise wasn’t instrument error but a faint signal coming uniformly from all directions.

AGI as a Tool-Maker?

One intriguing extension of this idea is how it relates to Artificial General Intelligence (AGI). If we follow Schopenhauer’s line of thought, an AGI might be most valuable not just as a conceptual engine (running endless abstract computations) or a tool user but as a tool-maker—designing new instruments or experiments to expand our perceptions in realms we currently can’t observe.

Without novel tools that feed new sensory data to our scientific community (human or machine), we might be stuck re-hashing the same concepts indefinitely. The real breakthroughs happen when we push the boundaries of what we can perceive—like building bigger colliders, telescopes, or detectors that show us something truly new.

• Do you agree with Schopenhauer’s stance that all truly new knowledge stems from perception?

• Could humanity have discovered the Higgs boson or Jupiter’s moons purely via “thinking really hard,” without actually looking or measuring?

• How does this tie into modern AI research? If AI were to truly “discover” something, would it need the capacity to design experiments and gather new data?

Discussion, links, and tweets

I'm a Software developer. Follow me on Twitter; you'll enjoy my tweets. I take care to carefully craft each one.

Tweet Follow @a_sarentorbic