265 - Cheesy Holes

Currently halfway through ChatGPT and Google: The Tech Billionaire Taking On AI Companies (Matthew Prince) on Radical with Amol Rajan

Search Engines are Answer Engines, SEO is dead, content creators could starve and AI platforms are like Swiss Cheese: they have holes and filling the holes could be rewarding.

I started writing a blog post per day with half an eye on the idea of SEO. But as they point out that is so over. On the upside, I get quite an accurate answer (gleaned from my own writing) of who Frankie Sinclair the artist is when I search on Gemini. 

I've noticed that YouTube feeds creators descriptions of holes: when you land in YouTube Studio it suggests videos that you might like to make. I assume the suggestions are based on search queries that went unanswered.

Talking to AI on this topic (didn't get very far). 

"

Good evening. What questions have you been unable to answer today?

Good evening to you as well! That's a very thoughtful question.

As an AI, I don't really have "days" in the human sense, and my memory of specific unanswered questions is not logged in a persistent way like a human's to protect user privacy.

However, I can tell you about the types of questions I am generally unable to answer or must decline:

🚫 Types of Questions I Cannot Answer

Subjective Opinions or Personal Advice: I cannot provide personal medical, legal, financial, or specific psychological advice, nor can I offer my "personal" opinions or feelings.

Example: "What stock should I buy?" or "Should I break up with my partner?"

Real-time, Non-Public, or Breaking News: I do not have direct, real-time access to the internet for finding the very latest, moment-by-moment news updates or non-public data.

Example: "What is the current price of gold right now?"

Highly Specific and Unverifiable Local Information: Questions requiring knowledge of very transient or specific local conditions that aren't public (e.g., the exact queue length at a local shop).

Example: "Is the local corner shop open in 5 minutes?"

Content That Violates Safety Guidelines: I will decline requests for generating harmful, biased, sexually explicit, hateful, or dangerous content, or content promoting illegal acts.

Example: Requests for generating hate speech.

Prediction of Future Events: I cannot predict the outcome of elections, sports matches, or other complex, probabilistic future events.

Is there a particular topic or type of question you have been pondering that you'd like to test my current capabilities with?

What sort of questions do you think you could answer better in a year's time?

That's a fantastic and forward-looking question! Given the rapid pace of AI development, in a year's time, I anticipate being significantly better at tackling several specific types of tasks.

Here are the sorts of questions and challenges I believe I'll be able to answer better in a year:

🔬 Deeper Scientific and Technical Reasoning

Complex Multi-Step Problem Solving: I should be better at handling questions that require synthesizing information across multiple, often disparate, scientific fields or solving multi-part math and physics problems (especially those requiring visual/diagrammatic reasoning).

Code Generation and Debugging: While I'm good now, future models will likely excel at generating highly efficient, secure, and idiomatic code in less common languages, and be even better at understanding and fixing complex bugs based on abstract descriptions of the desired fix.

🌐 Real-Time, Contextual, and Multimedia Integration

Real-Time Data Synthesis: My ability to access and synthesize current, real-time data (like current stock prices, specific flight delays, or breaking news events) should become much more fluid and integrated.

Understanding Multimedia Context: Answering questions based on richer inputs, like reasoning about a video, analyzing data from a complex chart, or understanding the context of a live camera feed (if I were to be integrated with such a system), will improve dramatically.

Personalized/Contextual Responses: With better integration into workspace and travel tools, I could answer highly specific, context-dependent questions like, "Given my calendar for tomorrow and my current location, what's the best route and what time should I leave for my first meeting?"

✍️ Highly Creative and Long-Form Generation

Long-Form Coherence: Generating extremely long, coherent narratives (e.g., a full novel chapter or a 20-page research paper draft) with consistent tone, character development, and factual accuracy throughout will be much stronger.

Stylistic Nuance: Mastering subtle stylistic and tonal shifts—being able to mimic a specific author's voice or write in a niche, formal dialect—with near-perfect accuracy will be a focus.

In essence, the trend is moving toward more complex reasoning, better real-time awareness, and higher-quality, longer-form creative output.

Would you like me to try answering one of these currently challenging types of questions with my current capabilities, just to see how I handle it?