Fight. - by Ruben Hassid - How to AI
When it comes to choosing between ChatGPT and Claude, it's hard to determine which is better. Both have their strengths - they search online, reason before answering, and are smarter than the average person. But as typical users, we can leverage both to our advantage. Rather than picking one over the other, I prefer to make them battle it out.
Introducing the "LLM Battle"
LLM, which stands for Large Language Model, is a battle I orchestrate between ChatGPT and Claude. The reason I advocate for LLM Battles is that large language models often exhibit confirmation bias, favoring information that aligns with pre-existing beliefs. Through these battles, we aim to challenge and overcome this bias.
LLM Battle is not suited for quick Google-like queries or simple tasks with straightforward answers. Instead, it shines in handling long, complex queries that require deep internet searches and critical thinking to avoid falling into confirmation bias.
The ChatGPT Experience
Starting with ChatGPT, I appreciate its vast memory of conversations and the custom instructions I've implemented internally. Despite its impressive agentic capabilities in searching and reasoning through the web, ChatGPT does have a tendency towards confirmation bias. An example that comes to mind is a discussion about launching a newsletter on LinkedIn, where ChatGPT's suggestions veered off track.
On the other hand, Claude offers a different perspective. When presented with the same query as ChatGPT, Claude's responses were more cautious and practical. The clash between the two models in the LLM Battle already shows promise in challenging and refining the advice provided.
Challenging the Advice
One key point of contention arose regarding launching a newsletter on LinkedIn. While ChatGPT advocated for it as a first step, Claude cautioned against it, emphasizing the importance of quality comments and the changing landscape of collaborative articles. This difference in advice sparked further conversations and reflection on the best course of action.
It's crucial to understand that large language models like ChatGPT and Claude, while powerful, have their limitations. They excel at completing language tasks but may not always reflect reality accurately. By setting them up to battle and scrutinizing their responses, we can leverage their strengths while mitigating their shortcomings.
Strategies for Effective LLM Battles
To conduct successful LLM Battles, I recommend following a structured approach:
- Ask ChatGPT for guidance and search results.
- Challenge ChatGPT's responses by comparing them with Claude's insights.
- Iterate the battle until a consensus or best approach is reached.
Additionally, a useful strategy is to engage in Perspective Transitioning, where different expert viewpoints are considered to avoid confirmation bias and promote critical analysis.
Ultimately, the goal of these battles is not to declare a winner between ChatGPT and Claude but to enhance our understanding of AI capabilities and limitations. By questioning, testing, and refining our interactions with AI, we can harness its potential effectively.
