A Guide to Human-Like Gen AI Behavior for Better Results

Keep in mind these human-like behaviors of generative AI (gen AI) as you engage with it.? Understanding these traits can lead to improved results from gen AI!?

1) Gen AI needs your context and orientation.

Have you noticed how designers need proper briefing before embarking on a job??

In the same manner, to achieve better gen AI results, give proper briefing!? The more knowns from you in the conversation, the less chances of deviation from gen AI.

What you can do:

• When using gen AI for image creation, describe what you’re imagining.
• When using gen AI as an idea generator, brief gen AI on the format that you need and the audience you are reaching.?

Here’s how gen AI images are only as good as your briefing: see the difference when you instruct gen AI to create an image of “people in a business meeting” versus “stock photo of business meeting in the woods”.

2) Gen AI relies on examples.

When trying to make a point, do you find yourself talking about examples of something done well in the past???

Do the same with gen AI, because it will rely on examples you use.? Be careful not to use sensitive data as your examples.

What you can do:

• Be clear when giving instructions to gen AI.?
• Specify your desired output.
• Take advantage of seed words in your gen AI conversations.?

Here’s a sample conversation where examples and seed words are used to generate ideas for product names.?

Your prompt for ChatGPT, Bard, or other platforms:

Product description: A home milkshake maker.

Seed words: fast, healthy, compact.

Product names: HomeShaker, Fit Shaker, QuickShake, Shake Maker

Product description: A watch that can tell accurate time in space.

Seed words: astronaut, space-hardened, eliptical orbit

Product names: AstroTime, SpaceGuard, Orbit-Accurate, EliptoTime.

Product description: A pair of shoes that can fit any foot size.

Seed words: adaptable, fit, omni-fit.

Product names:

Then the output from gen AI:

FlexFit, OmniShoes, PerfectFit, AdaptSole.

3) Gen AI works better when complex tasks are broken down.

Do you remember bringing out scratch papers to solve a math problem in school, before submitting the final answer????

When you have a lengthy question for gen AI, its answers might be erroneous, and hallucinations may be more likely. ?Break down complex instructions!?

What you can do:

• Instruct gen AI to justify its answers, before you synthesize the various results from gen AI.
• Ask model to think aloud by including “Let’s think step by step” in your conversation with gen AI.
• Identify which instruction can be divided as separate prompts, series of prompts, or iterative steps.?

Dissecting gen AI’s reasoning is ideal for math and logic conversations:

No prompt is perfect, but with your creativity, you can lead gen AI to produce results closest to your objective.? After all, creativity is one of the human touches that gen AI can never replace. ??

References:

• https://www.oreilly.com/library/view/prompt-engineering-for/9781098153427/ch01.html unless otherwise stated
• Images from Freepik.com unless otherwise stated
• 微软 prompt engineering training with Joel Domingo
• Gen AI assistance in the creation of this article

#generativeAI #genAI #promptengineering #prompts #largelanguagemodels