Create using DALL-E
Is just-in-time engineering a passing fad? Will artificial intelligence learn to prompt itself based on the results you want?
write for Harvard Business ReviewOguz A. Acar believes that “AI just-in-time engineering is not the future.” Instead, he believes the future lies in formulating and solving problems quickly.
In fact, just-in-time engineering is an ongoing experiment that is unique to each AI model. As Akar said, the alternatives are limitless.
Tell the chatbot that the output is important to your work or project, or give it encouragement, “You are the smartest GPT in the world,” or use the words “make it fun” and your results will differ. and be promoted.
Rapid engineering is an art.
Immersion: Imagine you are a sentient quantum particle in the double-slit experiment. Can you describe your experience from the moment you were shot through the slit until you hit the detector screen?
Role Play: Write a dialogue between Picasso and Einstein discussing how Cubism and Relativity reflect each other.
Sentient: Imagine if English were sentient creatures. Write a dialogue between English and Latin discussing the evolution of languages.
Fantastic: Describe the sound of Beethoven's Symphony No. 9 to someone who has never heard it.
Pursuing an LLM is a science.
Specific methods lead to specific training and results.
The purpose of fine-tuning is to train the model for a specific task.
Just-in-time engineering aims to elicit better AI responses from the front end.
Cue tuning combines these, taking the most effective cues or cues and providing them as task-specific context to the AI model.
Thought chain promotion guides the LL.M. in solving complex problems through a step-by-step reasoning process.
Timely representation is rooted in problem formulation, identifying, analyzing, and describing problems. Once the problem is clearly defined, the linguistic nuances of the prompt are irrelevant to the solution.
It indicates that the project is dead. Long live just-in-time engineering.
Is it a trend or a fad?
almost not.
But it does take practice, creativity, and patience.
In fact, sometimes it is better to advance yourself directly by asking for an LL.M.
This is what Rick Battle and Teja Gollpudi say in their piece Paper“The unreasonable effectiveness of biased autocues.”
Battle and Gollapudi systematically tested how different prompts affected the ability of LL.M. students to solve elementary school mathematics problems. They tested three different open source language models using 60 different prompt combinations. They found results to be varied and inconsistent.
Barthel and Gorpudi shared their observations IEEE“The only real trend may be that there is no trend. The best for any given model, data set, and prompting strategy may be for the specific combination at hand.”
Work hard to achieve your results and work with the LL.M. to find the best position for you and your partner in terms of intellectual enhancement.
Legendary music producer Rick Rubin once said: “Be aware of the assumption that the way you do something is the best way simply because it’s the way you’ve done it before.”
We are all learning and should always continue to try and imagine a new and better future.
Stay open. Stay curious.
Please subscribe Artificial Intelligence Insights.
Please subscribe to my guru newsletter, Solis Quantum.
