|
Here, a several hacks & frameworks I use to get better AI results. Please steal them all!
The co-worker framework
🧑🏻💼 See AI as a co-worker - You can unlock better results if you perceive AI as a helpful co-worker - and not just as a tool. Don't expect it to be able to read your mind or to be always right. Instead, give it clearly formulated tasks.
🔊 Voice chat with AI like you chat with your co-worker - Using the voice feature can unlock a more flowing, creative, fun and impactful way of using chatbots. Try it out and drop your voice notes in it!
🎯 Be clear and precise - The clearer and detailed you prompt, the better the results are (similar to the guidance you need to give your co-worker). So, please take time to think through and to improve prompts.
🔄️ Use AI to get help with AI - Are you sometimes stuck with your prompts and can't get the results you want? Then, use AI to talk about how to use AI! Open another tab and ask your AI to reformulate your prompt that it will work better with AI.
❓ Ask AI to ask you questions - You can improve your prompts by asking AI to ask you questions to clarify your request. Finish your prompt: "Please ask me three questions if you need more clarity to understand this task."
📈 Be emotional or even rude - This might sound surprising: But prompting emotional ("Are you really sure?") or even being a bit rude ("Are you kidding me?") can help to get better results. Using urgency ("This is really important to me.") can also help.
Truth
🏋🏻 AI tools have a too strong bias to be a friendly helper - Be aware of this bias, if you want to find the truth. Challenge it, by forcing the AI to give sources, asking the AI if it is really sure or even prompting it to judge you like a "very honest skeptic".
⌨️ Build trust by prompting the same in different tools - If you want to get to an accurate answer, try to prompt several tools with the same prompt and compare the answer. This triangulation can help you to get to more truthful results.
💾 Question the underlying data quality - AI tries to give you good and helpful answers - even on slim amounts of publicly available data. Therefore, stay critical when you research questions with a narrow publicly available data set.
🪥 The data form factor matters - The "bullshit in - bullshit out" rule also applies for AI. To improve your outcomes, it can help to clean the data. For instance, try to ask AI how to improve the data form factor to get better results.
🌎 LLM's are trained on a certain language - So, based on what LLM you use, you will get better results for English, German or Chinese sources. HuggingFace gives you a good idea which other LLMs you could test for your most used language.
Always stay curios & creative
🔬 AI forces you to experiment and be curious - Working with AI kicks off a continuos learning and experimentation process. It makes totally sense to reserve 30–90 minutes a day to learn and experiment with AI, like, for instance, Seth Godin does it.
💬 AI can prepare you for hard conversations - Are you having a hard business chat coming up? Use AI in voice mode to train for it. Prompt the AI to be your chat partner and then interact with it. Then, ask another AI to judge your conversation to improve.
🪴AI is a "living organism" - Thinking of AI as an evolving organism can help you to work better with it. AI tools are continuously learning and are, therefore, changing over time. For instance, you often can get different results for the same prompt.
|
