šŸ§˜ā€ā™€ļø TAKE A BREATH

Chatting with ChatGPT is more human than you think

Fv

Culture, tech, travel, business + marketing in 5 min. or less

Got 60 seconds?

With three separate rooms, Etihadā€™s ā€œThe Residenceā€ is the most spacious seat in the sky (Etihad Airways / Robb Report)

šŸ”¼ Tesla. The firmā€™s stock jumped 10% after hours following an outlook upgrade by Morgan Stanley Ā»Ā»

šŸ”½ AI musicā€™s industry cred. The AI-generated Drake/The Weeknd song is no longer eligible for Grammyā€™s consideration Ā»Ā»

šŸ’¬Ā ā€œMy biggest fear is losing my lifeā€™s work to people who view me as nothing more than a lottery ticket.ā€ More and more entrepreneurs are funding their own startups via consulting revenue Ā»Ā»

šŸ›« The 22 best Caribbean resorts, from boutique literary haunts to exclusive private-island escapes Ā»Ā»

šŸ‘— The 23 best perfumes of all time Ā»Ā»

šŸ’Ž The 7 most extravagant premium cabins in the air right now Ā»Ā»

Make Fv better in 5 seconds!

How often do you want this?

Login or Subscribe to participate in polls.

Got 2 minutes?

Citta di Veneziaā€™s outrageously stylish football kit (The Drum)

The AI-assistant wars heat up with Claude Pro, a new ChatGPT Plus rival Ā»Ā»

Also: OpenAI confirms that AI writing detectors donā€™t work Ā»Ā»

Why does everyone on TikTok talk like that? Ā»Ā»

When did football get fashionable? Ā»Ā»

More: Citta di Veneziaā€™s impeccable online shop Ā»Ā»

Coco Gauff says US Open win 'still doesnā€™t feel real at all' Ā»Ā»

More: Coco Gauff won the US Open, but the real crown is her discontinued Rolex ā€˜Red Grapeā€™ Ā»Ā»

Djokovicā€™s reign of greatness doesnā€™t look like ending anytime soon Ā»Ā»

Appleā€™s iPhone 15 event kicks off on Tuesday. Hereā€™s what to expect Ā»Ā» The newsletter's writer owns shares of Apple

iPhone 15ā€™s shift to USB-C has hidden potential to streamline computing Ā»Ā»

New York City is about to screw up congestion pricing Ā»Ā»

It's time for MySpace to make a comeback Ā»Ā»

Researchers gave 200 people US$10,000 each to study generosity. Did they spend the money on themselves or others? Ā»Ā»

Ivy League linguistics expert shares the No. 1 common phrase to never use Ā»Ā»

Mount Fuji in trouble: How Japanā€™s highest peak fell victim to overtourism Ā»Ā»

Got 5 mnutes?

Chatting with ChatGPT is more human than you think.

Not just for humans anymore. Apparently, telling ChatGPT to ā€œtake a deep breathā€ before solving a hard problem makes it better. Why?

TAKE A BREATH

Kinda fascinating: Researchers have (somehow) discovered that the phrase ā€œTake a deep breath and work on this problem step-by-step,ā€ significantly improves the effectiveness of AI's output.

Repeat: literally typing the words ā€œtake a breathā€ into a chatbot makes for better answers.

But why? How?

COMMON ADVICE

Sidebar: The advice to ā€œtake a deep breathā€ is so common that we hardly take a moment to consider just how good it is.

But it really does encapsulate the very essence of how we humans can best approach a tricky challenge ā€”for physiological reasons as well as emotional and psychological ones.

Deep breathing can stimulate something called the parasympathetic response, which promotes relaxation and actually counters the stress response. (Deep breathing is known to lower heart rate and blood pressure, too.)

Whatā€™s more, breathing deeply enhances oxygen delivery to the brain and other vital organs. This can sharpen cognitive function, improve concentration, and foster a clearer mind.

So how could typing that phrase into a chat window possibly help a generative AI system return better answers?

Letā€™s unpack.

AI! THEYā€™RE JUST LIKE US

Start with the words themselves. In our view, three key elements stand out in that prompt:

1 / Resetting really works. ā€œTake a deep breathā€ is more or less another way of saying: ā€œPause. Now reflect.ā€

Even though the words ā€œtake a deep breathā€ are as human-centric as it gets, and donā€™t feel like they should apply to AI at all, they actually do serve as a ā€œresetā€ or a context setter no matter who (or what) is on the receiving end.

Put another way, telling the system to ā€œtake a deep breathā€ is like saying, ā€œfocus completely on this next task and donā€™t get distracted by whatā€™s come before.ā€

Good advice for anyone, human or machine.

2 / Structure matters. ā€œWork on this problem step by step.ā€ This phrase is all about intention, and focus, and promotes clear and deliberate action. The goal is to nudge the AI towards concentrating on the issue at hand. The phrase also underscores the importance of sequential progress, or a methodical, step-by-step approach.

It makes logical sense that a computational system (as generative AI is) would benefit from an instruction to be as structured as possible with its ā€œthinking.ā€

3 / Always manage expectations. Saying ā€œwork on this problem step by stepā€ also subtly communicates that the question to follow may be multifacetedā€” and that a structured breakdown is expected in the response. Knowing exactly what you want, and telling AI as much, is a classic prompting best practice.

CLARITY. ALWAYS KEY.

Most people recognize that clear communications are a bit of a business and life hack.

In our view, the reason is because ā€”at its coreā€” clarity is about connection.

Stay with me.

Clear communication is not merely about transmitting information; it's an active task. Yep, ā€œtaskā€. Eliminating barriers of misunderstanding and ambiguity is hard work.

But itā€™s high reward too.

Before diving into a conversation or task with another person, it's beneficial to set the context up. Phrases like ā€œlet's take a moment to thinkā€ or ā€œbefore we proceed, let's recapā€ help us as humans establish a mental framework for the upcoming discussion.

They also ensure that both parties are on the same page, reducing potential misunderstandings.

Seen this way, clarity signals a respect for the recipient's time and understanding.

Now we know that it also improves ChatGPTā€™s responses, too.

OK, WHATā€™S THE POINT HERE?

As AI becomes an ever more integral part of our lives, itā€™s strangely satisfying to discover that the best way to speak to, understand, and enjoy the potential of AI, is the same as the best way to speak to, understand, and work with our fellow human beings.

More:

ā€œDeep in a new paper showing that AI comes up with more effective prompts for other AIs than humans do, there is this gem that shows how weird AIs are...ā€ Ā»Ā»

Written by Jon Kallus. Any feedback? Just reply!

Join the conversation

or to participate.