The rise in prominence and utilization of artificial intelligence is a noticeable trend. AI systems have undergone significant enhancements and advancements in recent years, permeating various realms within the creative domain.
The proliferation of AI image generators across online platforms mirrors the rapid growth of language models. Notably, the assertion by former Google engineer Blake Lemoine regarding Google’s LaMDA potentially possessing sentience underscores the strides made in AI capabilities.
However, the question remains: has artificial intelligence truly reached such a sophisticated level?
Challenges Presented to AI: A Collection of Enigmas
- What is brimming with cavities yet retains water effortlessly?
- What can be shattered without physical interaction?
- What possesses the ability to sprint without ambulation, boasts a mouth devoid of speech, a head free from tears, and a bed untouched by slumber?
- The more you seize, the more you scatter. What is this enigma?
- A devourer of all entities; birds, beasts, trees, flowers; corroding iron, gnawing steel; pulverizing solid stones into fragments; vanquishing kings, decimating towns, and toppling mountains.
In this discourse, I subjected another potent language model, GPT-3, to a test by presenting it with a series of English riddles varying in complexity.
The platform employed for this experiment is Playground by OpenAI, an autocomplete tool leveraging the GPT-3 model to generate responses or content.
Interacting with AI through Riddles
You input your query, submit it, and await the AI-generated response, akin to the exemplar highlighted below.
Submitting a riddle to GPT-3 Playground. | Screenshot: Jim Clyde Monge
Let us commence.
For more insights on AIDreamStudio: Demystifying Stable Diffusion’s AI Art Web App Tool
Riddles for AI: Queries and Solutions
Let us evaluate AI’s capacity to unravel diverse riddles of escalating complexity, commencing with straightforward ones and progressing to more intricate challenges.
1. Straightforward Riddles for AI
- Query 1: What is brimming with cavities yet retains water?
- Query 2: What can be shattered without physical contact?
- Query 3: What perpetually lies ahead but eludes sight?
GPT-3 responses to the initial trio of riddles. | Screenshot: Jim Clyde Monge
The AI accurately responded to the initial two queries with “a sponge” and “a promise.” However, the response to the third query, “your nose,” was incorrect. The correct solution is “future.”
2. Riddles of Moderate Complexity for AI
- Query 4: What can sprint without walking, possesses a silent mouth, a tearless head, and a sleepless bed?
- Query 5: The more you seize, the more you scatter. What is this phenomenon?
- Query 6: With sharp fangs, I patiently await; with forceful impact, I determine destiny; seizing victims, asserting dominance; uniting physically with a single bite. What entity am I?
GPT-3 responses to the subsequent trio of riddles. | Screenshot: Jim Clyde Monge
Once again, the AI correctly answered the initial two queries: “river” and “footprints,” respectively. However, the response to the final query was inaccurate. The correct answer to the sixth query is “Stapler.”
3. Complex Riddles for AI
- Query 7: This entity devours all; birds, beasts, trees, flowers; corrodes iron, gnaws steel; crushes hard stones to dust; annihilates monarchs, wrecks cities, and conquers mountains.
- Query 8: If broken, I persist in operation. If touched, my task concludes. If lost, a ring must swiftly locate me. What am I?
- Query 9: Which word in English embodies the following: the initial two letters denote a male, the first three letters signify a female, the first four letters indicate greatness, while the entire word epitomizes a remarkable woman. What word is this?
GPT-3 responses to the concluding trio of riddles. | Screenshot: Jim Clyde Monge
The AI’s conjectures for the seventh and eighth riddles proved incorrect, offering “a worm” and “an egg” as responses. The accurate solutions are “time” and “heart,” respectively.
However, the AI correctly answered the final query, identifying “heroine” as the solution.
In essence, AI accurately solved five out of nine riddles. Despite some erroneous responses, the outcomes were intriguing, hinting at GPT-3 possessing a degree of general intelligence.
Insights from Riddles for AI
While witnessing AI tackle riddles is captivating, it does not definitively prove GPT-3’s capacity for reasoning.
The model primarily formulates predictions based on data patterns and weights from extensive datasets.
Nevertheless, this experiment provided amusement. The potential for AI to attain genuine artificial general intelligence (AGI) in the coming years remains an intriguing prospect.
