OpenAI recently launched its new ChatGPT models, o1-preview and o1-mini, as part of the highly anticipated “Strawberry” release. These models are designed to excel in complex logical tasks like quantum physics, gene sequencing, and software coding.
However, the question arises—how do they perform in solving everyday problems, like riddles, relationship advice, and even movie plot holes?
As a former member of a middle school riddle club, I decided to test this AI on my turf: solving puzzles and riddles. To my surprise, ChatGPT’s o1 models handle both simple and complex riddles with impressive accuracy.
Although the o1-mini is slower than its more advanced counterpart, both models are incredibly quick when compared to human problem-solving. Observing the AI breaking down the logic step-by-step was fascinating, revealing its thought process for each solution.
One notable example is a riddle from The Hobbit. The AI broke down the problem into logical parts, though its grammar sometimes slipped, making it sound less polished. Still, the end result was correct, if not a bit robotic.
Can the AI Create New Riddles?
While ChatGPT is brilliant at solving existing riddles, its ability to invent new ones is more limited. I challenged the o1-preview model to create a riddle based on an answer I gave it—two dogs.
After about 30 seconds of logical deduction, the model came up with: “What has eight legs, four ears, two tails, and loves to bark?” While not incorrect, the riddle lacked creativity and humor. This highlights a key limitation: the AI excels in logical tasks but struggles with inventiveness.
Expanding Beyond Logic: Practical Life Advice
Next, I decided to take the AI beyond riddles and see how it handles practical problems. I asked it to diagnose a common car issue—what could cause a popping noise every 20 seconds while driving?
The AI provided logical and sound advice. It suggested checking the tires, engine, muffler, and brakes, offering repair tips like how to replace a tire.
The model explained its thought process, saying things like, “I’m piecing together causes of engine misfires, like faulty spark plugs.” This almost human-like logic made the interaction feel engaging, but the solutions were by-the-book and lacked creativity.
Relationship Advice: Logical but Bland
As someone who finds the dynamics of human relationships more complex than quantum physics, I asked the AI about flirting. It gave a solid, though rather dull, list of behaviors to watch for, such as asking questions and showing interest.
The headers of its response, like “Understanding flirting dynamics” and “Recognizing playful intimacy,” felt like something out of a robotic self-help manual.
The model succeeded in offering sound advice but fell short in making it personal or insightful. It was more like reading an instruction manual than having a conversation.
Missing Bells and Whistles
The o1-preview and o1-mini models don’t come with advanced features like image uploads, document analysis, or web browsing.
However, their speed and accuracy in solving logical problems make them stand out. For now, these AI models seem to remain out of reach for creative tasks like inventing riddles or offering witty relationship advice.
The new ChatGPT o1 models excel in logic-based tasks, but creativity and humor are not their strong suits. While they can solve riddles, diagnose car problems, and give basic relationship advice, they lack the inventiveness needed to truly replace human creativity.
For those looking for a highly logical AI assistant, these models are great, but don’t expect them to surprise you with creative solutions.
