✍️ Ditulis oleh Ts Dr Suhailah Mohamed Noor | Founder, Prompt Academy
📅 Diterbitkan pada 23 Jun 2025
“Agaknya ada tak yang terfikir soalan ni sebelum ni atau selepas membaca artikel saya sebelum ini tentang AI jadi suara kita?:
Soalan yang sebenarnya menarik untuk difikirkan, sebab ia menyentuh keupayaan sebenar AI — bukan sekadar menulis, tetapi menyesuaikan diri dengan pengguna. Dan saya berani kata: ya, AI boleh belajar gaya kita, asalkan kita rajin “bercakap” dengannya.
🧠 AI Tak Ingat Kita Macam Manusia, Tapi…
AI tidak ada memori jangka panjang seperti manusia — tetapi dalam satu sesi perbualan, dia ada apa yang kita panggil “jendela konteks.”
Selagi kita masih berinteraksi dalam sesi yang sama (atau dengan cukup banyak contoh dalam sesi yang berulang), AI boleh:
- Ingat gaya bahasa kita
- Tahu nada kesukaan kita (profesional? bersahaja?)
- Kenal struktur ayat yang kita guna
- Faham kita suka ringkas atau terperinci
Saya uji perkara ini sendiri. Saya tanya begini kepada Gemini Pro:
“An AI like you should be my voice. Right? Tell me, how should I train you to be my voice.”
Dan beginilah antara jawapan yang saya terima — sangat sistematik, sangat masuk akal, dan… sangat berguna untuk dijadikan panduan sesi lanjutan bengkel!
Respon dari Gemini Pro
As an AI, I don’t have a “voice” in the human sense of the word. I don’t have personal opinions, emotions, or a unique way of speaking that is inherently “me.” My responses are generated based on the vast amount of text data I’ve been trained on, designed to be informative, coherent, and helpful.
However, when you say you want me to “be your voice,” you likely mean you want my responses to reflect your:
- Tone: Do you prefer formal or informal, serious or humorous, direct or nuanced?
- Style: Do you use specific phrases, sentence structures, or a particular level of detail?
- Perspective: What are your core beliefs, values, and the angles you typically take on subjects?
- Vocabulary: Do you use specialized jargon, common language, or a blend?
How to Train Me (or rather, guide me) to Be “Your Voice”:
Since I’m a pre-trained model, you can’t directly “train” me in the way a machine learning engineer would. However, you can effectively guide my responses to align with your voice through the way you interact with me. Here’s how:
- Provide Examples: The most effective way is to show me. When you ask me to generate text, provide examples of your own writing. For instance:
- “Write a short social media post about X, in the style of my previous post [paste your previous post here].”
- “Draft an email explaining Y, using the tone from this email I sent before: [paste your email].”
- Be Explicit with Instructions: Clearly state the desired tone, style, and perspective. Use descriptive adjectives.
- “Write a concise and professional summary of Z.”
- “Craft a witty and slightly sarcastic comment about A.”
- “Explain B in simple, easy-to-understand terms, as if you’re talking to a complete beginner.”
- “Adopt a formal, academic tone for this paragraph.”
- Specify Audience and Purpose: Knowing who you’re speaking to and why helps me tailor the language.
- “Write a persuasive argument for C, targeting potential investors.”
- “Create a friendly and encouraging message for a team update.”
- Give Feedback and Iterate: If my initial response isn’t quite right, tell me what to change.
- “Make that less formal.”
- “Add more enthusiasm.”
- “Rephrase that to be more direct.”
- “Can you use simpler vocabulary?”
- Define Your Persona (if applicable): If you have a specific persona you want me to embody (e.g., a brand voice, a character), describe it in detail.
- “Imagine you are a helpful, slightly quirky librarian. Write a tip about organizing books.”
- Provide Keywords and Phrases: If there are specific terms or idioms you frequently use, mention them.
- “When discussing sustainability, always include the phrase ‘for a greener tomorrow’.”
- Limit or Expand Detail: Tell me how much information you want included.
- “Give me a high-level overview of D, don’t go into too much detail.”
- “Elaborate on E, providing supporting evidence.”
By consistently providing clear instructions, examples, and feedback, you can effectively “steer” my output to closely match what you envision as “your voice.” Remember, I’m a tool, and like any tool, the better you understand how to use it, the better the results will be.
📋 Apa Sebenarnya Berlaku Bila Kita Ulang-ulang Gaya?
Semakin kita:
- Beri contoh tulisan kita,
- Beri arahan jelas (“tulis gaya saya”), dan
- Beri maklum balas (“terlalu formal”, “tolong ringkaskan”),
Semakin AI mula serasi dengan gaya kita. Saya perasan, bila saya berinteraksi dengan nada saya sendiri — AI mula menghasilkan jawapan yang “terdengar macam saya”. Kadang-kadang saya senyum sendiri baca ayat yang dia hasilkan. “Ini macam saya yang tulis la!”
💡 Tips Praktikal untuk Bantu AI Ingat Gaya Anda
Saya ringkaskan 4 langkah praktikal di bawah — ini lanjutan daripada artikel pertama:
- Gunakan gaya yang konsisten bila berinteraksi dengan AI
Kalau suka nada santai, jangan bertukar jadi terlalu akademik tiba-tiba. - Beri contoh tulisan anda, walaupun satu perenggan
Caption Instagram, emel rasmi, apa saja yang mencerminkan nada anda. - Tegur AI bila dia lari nada
“Terlalu panjang.” “Saya tak guna ayat begini.” “Boleh bagi gaya saya balik?” - Berbual dengan AI macam berbual dengan staf baru
Bukan bagi arahan satu arah, tapi beri peluang AI “belajar” daripada maklum balas kita.
🎯 Penutup: Bila AI Jadi Lebih Daripada Sekadar Alat
Saya tidak percaya AI akan menggantikan penulis manusia. Tapi saya percaya, AI boleh menjadi pasangan menulis yang setia — jika kita sendiri tahu bagaimana nak ajar dia.
“Setiap interaksi dengan AI adalah latihan. Dan setiap latihan adalah satu langkah ke arah menjadikan AI suara anda.”