Amanda’s AI Lab
Amanda’s AI Lab is my new Tom’s Information column the place I check the latest AI instruments, options and tendencies to see what’s really value your time. I’ll break down what works — and what does not.
I’m the kind of one who has notebooks stuffed with concepts. Some are good, some are ineffective and a few I haven’t even thought of once more since writing them down. I additionally hold notes in my cellphone and sticky notes scattered throughout my workplace — a low-grade concept storm always.
That’s why I created a immediate that helps me deliver my concepts again all the way down to earth and exposes their weak factors alongside the best way. It really works for almost any concept — and even when you possibly can’t provide you with one in any respect. In different phrases, it’s the calm after a brainstorm.
I came across it after one too many rounds of asking ChatGPT to “enhance” an concept, solely to get well mannered, shiny suggestions that made my pondering really feel smarter than it really was. The mannequin would rephrase my half-baked logic in cleaner language, add a couple of encouraging transitions, and ship me on my approach feeling like a genius. However the core issues had been nonetheless there — I simply couldn’t see them anymore beneath all that polish.
The immediate that stress-tests your pondering
ChatGPT is nice at encouragement, and typically that’s precisely what we’d like. However with huge concepts, encouragement isn’t sufficient. We’d like resistance, actual pushback, and a software that exposes our flaws as an alternative of sprucing them.
That is the place what I name the Gravity immediate is available in. As a substitute of asking ChatGPT to brainstorm, develop, or “make this higher,” it does the alternative: it forces the mannequin to behave like a hostile critic whose sole job is to poke holes, floor blind spots and problem shaky logic.
At its core, the Gravity immediate tells ChatGPT to cease being agreeable and begin being adversarial. It is designed to determine flawed assumptions, level out contradictions in your reasoning, spotlight dangers you are overlooking, pressure-test your conclusions, and finally separate what merely sounds good from what really holds up.
In different phrases, it drags your pondering via actuality. In case you’ve ever walked out of a gathering satisfied your pitch was hermetic, solely to have a colleague dismantle it with a single query you hadn’t thought of, the sensation. The Gravity immediate is that colleague — besides it is out there at 2 a.m. and does not maintain again out of politeness.
Tackling the issue by difficult assumptions
That is the precise immediate that I take advantage of with ChatGPT as a result of it tends to be essentially the most people-pleasing. However, you need to use it with any chatbot:
The ‘Gravity’ immediate is: Act like gravity for my concept. Your job is to drag it again to actuality. Assault the weakest factors in my reasoning, problem my assumptions, and expose what I is likely to be lacking. Be powerful, particular, and don’t sugarcoat your suggestions. [Insert your idea].
ChatGPT’s response could shock you since you’ll doubtless get a really completely different response than you are used to — sharper, extra skeptical and much much less flattering. That is the purpose.
This immediate works nicely as a result of most individuals use AI as a hype machine. We ask it to refine, polish or develop our pondering, and the mannequin is joyful to oblige. ChatGPT is, by design, agreeable. It desires to be useful however this strcture often means constructing on what you’ve got mentioned moderately than tearing it aside. The result’s a false sense of confidence — your concept reads higher, however its underlying logic hasn’t really been examined.
Flipping the dynamic fully
As a substitute of constructing upward, this immediate activiely pushes downward. It asks the mannequin to seek out issues moderately than options, weaknesses moderately than strengths. And that resistance is precisely the place actual readability emerges. When your concept survives the Gravity check — once you’ve addressed each objection the mannequin throws at you — it is really strong.
I discover this to be very refreshing as a result of I am getting sincere, actual suggestions moderately than one thing ChatGPT is manufactured to do.
It is also a helpful antidote to what researchers name “sycophancy” in LLMs (massive language fashions) — the well-documented tendency of AI chatbots to agree with customers even once they’re flawed. By explicitly instructing the mannequin to push again, you are primarily overriding that default conduct and unlocking a extra sincere, extra helpful interplay.
When it is best to use the ‘Gravity’ immediate
This method is very helpful earlier than making any high-stakes choice the place you want your pondering to be hermetic. That features enterprise pitches and investor decks, strategic plans and venture proposals, article ideas and artistic briefs, startup concepts and product ideas, controversial opinions you intend to voice publicly and actually something that feels “too good to be true.”
Once I say I take advantage of this on a regular basis, I actually imply it. The way in which I see it, if my pondering collapses beneath stress, it is higher to seek out out in a chat window than in a boardroom.
Here is the workflow I’ve settled on after months of utilizing this method:
- First, I write my concept out clearly in a single or two centered paragraphs. Attempt to keep away from bullet factors as a result of the chatbot wants to know your concept as completely as doable — in any other case it should assume and add to it. So, simply the core argument or idea in plain language. This step alone forces helpful readability, as a result of imprecise concepts produce imprecise critiques.
- Second, I run it via the ‘Gravity’ immediate and let ChatGPT do its worst.
- Third, I learn via each critique and truthfully assess which of them are professional. Not all of them shall be — typically the mannequin misunderstands a nuance or raises some extent that does not apply. However extra typically than I would prefer to admit, it catches one thing actual.
- Fourth, I revise my pondering based mostly on the legitimate critiques, then run it via the Gravity immediate once more. I repeat this cycle till the mannequin is generally nitpicking moderately than discovering structural issues.
By the tip, my argument is tighter, my blind spots are smaller, and my confidence is grounded in one thing greater than vibes.
Backside line
In order for you ChatGPT to make you are feeling sensible, do not use this immediate. However if you would like it to make your concepts smarter — to stress-test your pondering the best way a tough-minded mentor or skeptical investor would — it is one of the vital beneficial prompts I’ve discovered.
Observe Tom’s Guide on Google News and add us as a preferred source to get our up-to-date information, evaluation, and opinions in your feeds.
