Rekayasa pakon
Rekayasa pakon (basa Inggris: prompt engineering), yaiku tumindak natasungsun utawa nyipta prentah kanthi ancas ngasilaké wetonan luwih becik saka modhèl kacerdhasan gawéyan mangriptayasa (AI). Tumindak iki kalebu ngreka pitakon sing genah, muwuhi wirasa (konteks) sing cocog, lan mrayoga pitembungan kanggo nuntun modhèl murih bisa saya tlesih, miguna, lan ngasilaké respons luwih ajeg. [1]
Pakon yaiku teks basa alami sing njlèntrèhaké bubuhan (tugas) sing kudu linakon déning AI. [2] Pakon kanggo modhèl basa teks-marang-teks bisa arupa pitakon, prentah, utawa pranyatan luwih dawa sing mawa wirasa, prentah, uga babad pacelathon. Rekayasa pakon manawa uga kalebu ngrumusaké pitakon, nemtokaké gaya, pilihan tembung lan paramasastra, [3] nyumadyani wirasa sing salaras, utawa njlèntrèhaké sipat sing kudu ditiru AI. [1]
Nalika mursita (berkomunikasi) karo teks-marang-gambar utawa modhèl teks-marang-audio, pakon biasané wujud wedharan saka apa-apa sing dikarepaké dadi wetonan. Tuladhané, "foto mutu dhuwur saka antariksawan sing nunggang jaran." [4] utawa "Lo-fi alon BPM elektro santai kanthi sampel organik". [5] Makon modhèl teks-marang-gambar bisa uga kalebu muwuhi, mbusak, utawa nandhesaké tembung kanggo piwoleh jejer (subjek), gaya, tata panggon, pancahyan, lan kaéndahan sing dikarepaké. [6]
- 1 2 Genkina, Dina (2024-03-06). "AI Prompt Engineering is Dead: Long live AI prompt engineering". IEEE Spectrum. Dibukak ing 2025-01-18.
- ↑ Radford, Alec; Wu, Jeffrey; Child, Rewon; Luan, David; Amodei, Dario; Sutskever, Ilya (2019). "Language Models are Unsupervised Multitask Learners" (PDF). OpenAI.
We demonstrate language models can perform down-stream tasks in a zero-shot setting – without any parameter or architecture modification
- ↑ Wahle, Jan Philip; Ruas, Terry; Xu, Yang; Gipp, Bela (2024). "Paraphrase Types Elicit Prompt Engineering Capabilities". Ing Al-Onaizan, Yaser; Bansal, Mohit; Chen, Yun-Nung (èd.). Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. Miami, Florida, USA: Association for Computational Linguistics. kc. 11004–11033. arXiv:2406.19898. doi:10.18653/v1/2024.emnlp-main.617.
- ↑ Heaven, Will Douglas (April 6, 2022). "This horse-riding astronaut is a milestone on AI's long road towards understanding". MIT Technology Review. Dibukak ing 2023-08-14.
- ↑ Wiggers, Kyle (2023-06-12). "Meta open sources an AI-powered music generator". TechCrunch. Dibukak ing 2023-08-15.
Next, I gave a more complicated prompt to attempt to throw MusicGen for a loop: "Lo-fi slow BPM electro chill with organic samples."
- ↑ Mittal, Aayush (2023-07-27). "Mastering AI Art: A Concise Guide to Midjourney and Prompt Engineering". Unite.AI (ing basa Inggris Amérika Sarékat). Dibukak ing 2025-05-09.