How to Speak

ELI5/TLDR

Patrick Winston, an MIT AI professor, spent decades studying what makes people effective speakers and distilled it into a single hour. His core claim: success depends more on your ability to communicate than on the quality of your ideas. Speaking well is not about talent. It is about knowledge and practice, applied in the right order, to a checklist of techniques most people never bother to learn.

The Full Story

The Formula

Winston opens with a provocation borrowed from military law: officers who send soldiers into battle without weapons face court martial. Students, he argues, deserve the same protection — they should not be sent into the world unable to communicate.

He frames speaking ability as a formula: Knowledge + Practice + Talent, where talent is the smallest term. To prove the point, he tells a story about out-skiing Mary Lou Retton — an Olympic gold medalist — on a novice slope. She had the talent. He had the knowledge and practice. He was the better skier.

“Your success in life will be determined largely by your ability to speak, your ability to write, and the quality of your ideas — in that order.”

That ranking is the quiet bombshell of the talk. Most academics spend their lives optimizing the third thing on the list.

The Starter Kit: Four Heuristics

Winston lays out four techniques he keeps in his head during every talk:

Cycling. Say it three times. Not because people are stupid — because at any moment, roughly 20% of the audience has fogged out. Three passes make it statistically likely everyone catches the idea at least once.

Building a fence. Distinguish your idea from similar ones. If you are teaching someone from Mars what an arch is, you show them arches, then show them things that are not arches. In technical talks, this means explicitly stating how your work differs from Jones’s work. Otherwise people will conflate them.

Verbal punctuation. People who fog out need landmarks to re-enter the talk. Enumerate things. Say “the third idea is…” Give the audience seams in the fabric where they can grab back on.

Asking questions. Winston demonstrates this live by asking the audience what the fourth heuristic might be, then waiting seven seconds of dead air — the standard maximum before the silence becomes unbearable. The question cannot be too easy (embarrassing to answer) or too hard (nobody answers).

Time, Place, and the War on Darkness

The ideal lecture time: 11 AM. Most people at MIT are awake by then and nobody has gone back to sleep yet.

The ideal room: well-lit. Winston is emphatic about this. Dim rooms signal sleep. When audio-visual staff suggest turning off the lights for better slides, his reply:

“It’s extremely hard to see slides through closed eyelids.”

He also recommends casing the venue like you would a bank you intend to rob — visit beforehand, find the weirdnesses, adjust. And on his own preparation ritual:

“I imagined that all the seats were filled with disinterested farm animals, and that way, I knew that no matter how bad it was, it wouldn’t be as bad as that.”

The room should also be right-sized. Ten people in a big hall, and everyone wonders what more interesting event they are missing.

Tools of the Trade: Board, Props, Slides

The blackboard is Winston’s weapon of choice for teaching. Three reasons: it has a graphic quality, it enforces the right speed (you write at roughly the pace people can absorb ideas), and it gives you something to do with your hands. That last one matters more than you would think. Novice speakers become suddenly aware of their hands, as if they were “private parts that shouldn’t be exposed in public.” Hands go into pockets, or behind the back — both of which are considered insulting in various cultures. A board solves this by making your hands useful.

He tells a story about watching Seymour Papert lecture and being riveted. On the second viewing — he went once for content, once for style — he realized Papert was constantly pointing at things on the board that had nothing to do with what he was saying. It did not matter. The technique worked anyway.

Props are memorable in a way nothing else is. Winston draws from Henrik Ibsen’s Hedda Gabler: a potbellied stove with embers that grow larger as the play’s tension mounts, alongside a manuscript the audience knows is doomed. Twenty years after breaking a pointer during this lecture, a colleague’s only memory of the talk was “that’s the one where you broke the pointer.”

He demonstrates the bicycle wheel gyroscope problem — a physics puzzle where mechanical engineers armed with right-hand-rule knowledge perform at coin-flip accuracy, while anyone who focuses on a single piece of duct tape on the rim gets it right instantly. The point: props let you think about problems the right way.

His explanation for why boards and props outperform slides:

“When you’re sitting up there watching me write on the board, all those little mirror neurons in your head become actuated, and you can feel yourself writing on the blackboard.”

He calls it empathetic mirroring. You cannot mirror a slide.

Slides are for exposing ideas, not teaching them. Winston’s rules are blunt. At an airport, a stranger asked him to critique his slides. Without looking:

“You have too many, and they have too many words.” … “Because it’s always true.”

The crimes: too many words, speaker standing far from the slide (creating a tennis-match effect), reading slides aloud, background junk, logos, titles that could just be spoken. He strips a sample slide down layer by layer until almost nothing is left. The remaining words are condiments, not the main course.

Font sizes below 35 points are a crime — not because they are illegible, but because small fonts enable the sin of cramming too many words on the slide. Laser pointers are a crime because using one forces you to turn your back on the audience. Instead, put arrows on the slide and say “look at arrow number one.”

He shows a real talk he attended: slide 80, which opens with the words “this is the first of 10 conclusions slides.” The meeting sponsor is reading his email. The co-sponsor is examining the lunch menu. A third attendee appears attentive in the photo but was, in the video, doing this: yawn.

Inspiring People

Winston surveyed freshmen, senior faculty, and everyone in between about what inspired them. The pattern:

Freshmen were inspired by a high school teacher who told them they could do it
Senior faculty were inspired by someone who helped them see a problem in a new way
Everyone was inspired by someone who exhibited passion about what they were doing

He demonstrates with a resource-allocation lecture opener: a brute-force map-coloring program that will not finish before the sun consumes the Earth, versus a slightly tweaked version that finishes in seconds. “Isn’t that cool?” is not filler here — it is the technique.

Teaching People to Think

When you ask MIT faculty what their most important job is, they say “teaching people how to think.” When you ask how they do it: blank stare.

Winston’s answer: we are storytelling animals. Thinking is story manipulation. So you teach thinking by providing stories, questions about those stories, mechanisms for analyzing them, ways of assembling them, and ways of evaluating how reliable they are.

The Job Talk

Over drinks with two fellow members of the Navy Science Board — one who had written 21 books (making Winston jealous of his own mere 17), and another who drove steel rods through tank armor with rail guns — Winston asked what they look for in a faculty candidate.

Within one microsecond: “They have to show us they’ve got some kind of vision.” Immediately followed by: “They have to show us they’ve done something.”

How long does a candidate have to establish these two things? Five minutes.

The structure: open with a problem somebody cares about and something new in your approach. List the steps needed to solve it (you do not need to have completed all of them). Demonstrate what you have built. Close by enumerating your contributions. The talk is a sandwich — contributions at the beginning and end.

Practice, but Correctly

Practicing your talk by showing slides to labmates is almost useless. People who know your work will hallucinate material in your presentation that is not there. Your faculty advisor is especially bad at this — they know what you mean, so they cannot tell you when you have failed to say it. Instead, find friends who do not know your work and open with:

“If you can’t make me cry, I won’t value you as a friend anymore.”

Also worth noting: the difficulty of questions is inversely proportional to the age of the questioner. Old professors have made peace with the world. Young ones are trying to prove themselves. Choose gray-haired examining committees when possible.

Getting Famous (or at Least Not Ignored)

Winston sat next to Julia Child at a fundraiser for saving Venice. After watching a steady stream of admirers approach her all evening, he asked: “Ms. Child, is it fun to be famous?” She thought about it. “You get used to it.”

His reflection:

“You never get used to being ignored.”

Your ideas are like your children. You do not want them going into the world in rags. So packaging matters, and it is not vanity to care about it.

His framework for making work memorable — “Winston’s Star” — has five S’s, drawn from his own accidentally-famous PhD thesis on arch learning:

Symbol — a visual anchor (his was the arch itself)
Slogan — a handle phrase (“one shot learning”)
Surprise — the counterintuitive result (you do not need a million examples to learn)
Salient idea — one idea that sticks out above the rest (the “near miss” concept)
Story — how you did it, how it works, why it matters

How to Stop

The final slide should not be a list of collaborators (put those on the first slide), a “Questions?” slide (wastes real estate for 20 minutes), a “Thank You” slide, or a URL nobody will write down. It should be a contributions slide — the mirror of the vision you opened with.

As for final words: do not say “thank you.” It implies the audience stayed out of politeness and secretly wanted to be elsewhere. Winston shows clips of Chris Christie and Bill Clinton both ending speeches with “God bless you, and God bless America” — a benediction, not gratitude. He points out that Clinton appears to physically stop himself from saying “thank you,” pressing his lips together before pivoting to a salute.

Other options: a joke (people are ready for one by the end), a salute to the audience, or simply a clear signal that the event is over — the way a conductor shaking the concertmaster’s hand means you can clap now.

Winston closes his own talk by saluting the audience for showing up:

“By being here, I think you have demonstrated an understanding that how you present and how you package your ideas is an important thing. And I salute you for that.”

He does not say thank you.

Claude’s Take

This is one of those talks that has been legendary on YouTube for years, and it earns the reputation. Winston is doing the thing he is teaching — the lecture is itself a demonstration of every technique he describes. He cycles through ideas, builds fences, uses verbal punctuation, asks questions, deploys props, and ends without saying thank you. It is a masterclass taught by example, which is rare and difficult to pull off.

The substance is solid. Most of his advice is practical, testable, and does not require charisma or natural talent — which is precisely his thesis. The formula (Knowledge + Practice + tiny T for Talent) is probably the most useful framing in the talk, because it converts public speaking from an identity (“I’m not a good speaker”) into a skill with learnable components.

A few points worth examining more closely. The claim that speaking ability matters more than the quality of your ideas is provocative and probably overstated as a universal truth — but within academia, where brilliant work regularly dies in obscurity because its creator could not present it, he is directionally correct. The “empathetic mirroring” explanation for why blackboards work is speculative neuroscience. Mirror neurons are real, but their role is more contested than Winston implies. Still, the observation that boards and props outperform slides is well-supported by student surveys and common experience, even if the mechanism is debatable.

The “don’t say thank you” advice is his most controversial point and the one people push back on hardest. He is technically right that it is a weak signal — but in many professional contexts, omitting it reads as rude rather than confident. Cultural context matters here more than Winston acknowledges. Use judgment.

What makes this talk endure is not any single insight. It is that Winston, a man who spent 40+ years at MIT thinking about artificial intelligence, chose to spend a significant portion of his legacy on teaching people how to talk to each other. He died in 2019, shortly before this recording was published. The talk has been viewed tens of millions of times. The irony would not be lost on him: his most famous contribution to the world was not about how machines think, but about how humans communicate.