AI muscles into Churchill

We are all bemused by the machi­na­tions of AI (Arti­fi­cial Intel­li­gence). As Hilaire Bel­loc said about the intel­li­gence of women: Men come to look upon it “first with rev­er­ence, then with stupour, and final­ly with terror.”

Any­way! Last week I updat­ed the AI ver­sion of Churchill’s com­ments on bananas. And won­dered, since I first pro­vid­ed what he actu­al­ly said, whether Chat­G­PT had picked it up. Evi­dent­ly not! But Chat GPT now offers some­thing he nev­er said.

Nev­er­the­less, AI con­tin­ues to progress, offer­ing an amus­ing pas­time. Take your favorite remark by any­body. The only require­ment is that it be an exact, well-known quo­ta­tion. Send it to Chat GPT. Ask them to repro­duce it as a scene in a Shake­speare play. I promise you will be amused.

Drunk and ugly: the actual encounter

First, here is the orig­i­nal tran­script of the famous con­fronta­tion, con­firmed by an eye-wit­ness. The venue was the House of Com­mons, the year 1946. The speak­ers were Win­ston Churchill and the Labour MP for Liv­er­pool Exchange, Bessie Brad­dock.  From my book, Churchill by Him­self, page 573:

Brad­dock: “Win­ston, you are drunk, and what’s more, you are dis­gust­ing­ly drunk.”

Churchill “Bessie, my dear, you are ugly, and what’s more, you are dis­gust­ing­ly ugly. But tomor­row I shall be sober and you will still be dis­gust­ing­ly ugly.”

This famous exchange was con­firmed to me by Ronald Gold­ing, Churchill’s Scot­land Yard detec­tive. He was on the spot to observe it—no bet­ter wit­ness. Churchill was leav­ing the House of Com­mons after a long, late evening debate.

Lady Soames, who said her father was always gal­lant to women, doubt­ed the sto­ry, but Gold­ing explained that WSC was not drunk, just tired and wob­bly, which caused him to fire the full arse­nal. Churchill was, how­ev­er, rely­ing on his pho­to­graph­ic mem­o­ry. In the 1934 movie It’s a Gift, W.C. Fields’s char­ac­ter, told he is drunk, responds, “Yeah, and you’re crazy. But I’ll be sober tomor­row and you’ll be crazy the rest of your life.” So Churchill was real­ly para­phras­ing W.C. Fields.

The AI Shakespeare version

We fed this tran­script into Chat­G­PT and asked for an AI ver­sion as a scene from Shake­speare. The result took only seconds:

Char­ac­ters:

• Sir Win­stonus Churchillius, a stout, wit­ty states­man, full of jests and gravitas.
• Lady Bessi­ca Brad­dock­ia, a fiery noble­woman with a sharp tongue.

Set­ting: A grand hall in West­min­ster. Tapes­tries hang from the walls, and the faint clink of gob­lets echoes through the air.

Enter:  Sir Win­stonus Churchillius, gob­let in hand. Lady Bessi­ca Brad­dock­ia approach­es, fan­ning her­self dramatically.

Lady Bessi­ca:
Hail, Sir Win­stonus, thou art returned
From feast­ing, drink­ing, or some sport absurd?
Thy face is flushed, and eyes like moons do glow;
Dost thou drown England’s cares in wine’s deep flow?

Sir Win­stonus:
Lady Bessi­ca, sharp of wit and tongue,
Thy voice dost pierce like arrows deft­ly flung.
Yet in thy charge, some truth may yet reside,
For­sooth, my goblet’s brim hath oft been plied.

Lady Bessi­ca:
O, gal­lant knight, though words like hon­ey flow,
Thy gait doth tell what sober truth may show.
Thou reek’st of vintner’s wares! Thy steps betray
A man whose gob­let oft doth go astray.

Sir Win­stonus (with a wry smile):
And thou, fair dame, art not of angels born,
Thy tongue’s keen edge doth all who cross thee warn.
Yet in thy zeal to strike with jest and jeer,
Thou dost for­get the mir­ror draw­ing near.

*

Lady Bessi­ca (point­ing dramatically):
Thou speak’st of mir­rors? Look upon thy girth!
A ves­sel vast, no ship of slen­der berth!
Thou art, Sir Win­stonus, round as a barrel!
Dost thou com­pete with Bac­chus in this quarrel?

Sir Win­stonus (plac­ing hand on his belly):
A bar­rel, say’st thou? O, cru­el jest indeed!
But bar­rels hold the finest wine and mead.
Yet thou, good dame, art no del­i­cate dove;
The pot and ket­tle may yet learn of love.

Lady Bessi­ca (gasp­ing in mock outrage):
Out, vile retort! Thy jest doth stab too deep.
Yet truth be told, I can­not silence keep.
Thy wit, though sharp, hides wisdom’s kind­ly balm;
A tem­pest fierce, yet ever seeks the calm.

*

Sir Win­stonus (rais­ing his goblet):
Lady Bessi­ca, thou art a noble foe.
Thy fire doth warm where icy winds might blow.
Let us, then, toast, not bick­er in this hall,
For Eng­land needs us both to stand, not fall.

Lady Bessi­ca (tak­ing a gob­let from a pass­ing servant):
To Eng­land, aye, and to the bonds we weave,
Though tem­pests rise, we shall not take our leave.
For­sooth, Sir Win­stonus, thou art a rogue,
Yet one I’d fight beside, ‘gainst any vogue.

Both raise their gob­lets high and drink deeply….

Sir Win­stonus:
A rogue, thou say’st? That title I shall wear!
Yet call me rogue, and not a bore, I swear.
Let his­to­ry judge, when both are but a name,
Which spark burned brighter in Britannia’s flame.

Lady Bessi­ca:
Then let us spark togeth­er, fiery bright,
And burn away the shad­ows of the night.

Exe­unt, laugh­ing and ban­ter­ing, arm in arm.

“If we shadows have offended,
“Think but this, and all is mended:”

—Puck, in A Mid­sum­mer Night’s Dream

Soft­ware engi­neer Ian W.D. Lang­worth explains how AI repro­duces the famous encounter in Shake­speare­an form:

A large lan­guage mod­el (LLM) focus­es on a cor­pus of text, such as all of Wikipedia. Or an even larg­er cor­pus, such all of the scannable Inter­net. The train­ing hap­pens by feed­ing the mod­el a large amount of text, ask­ing it to pre­dict the next word in the sequence. Doing this bil­lions of times pro­duces many lev­els of refine­ment. The mod­el learns the pat­terns in the text, and can then use these pat­terns to gen­er­ate new text.

The result is a spe­cial­ly for­mat­ted mul­ti-giga­byte file which can be used to gen­er­ate text. This is called a “mod­el.” You can use this mod­el to gen­er­ate text by feed­ing it a prompt.

In this case, the mod­el was trained on Wikipedia and the Inter­net. Obvi­ous­ly, it learned of the Brad­dock-Churchill exchange some­where on the web—in addi­tion to the works of Shakespeare.

Accu­ra­cy is irrel­e­vant. The mod­el doesn’t nec­es­sar­i­ly have a con­cept of “truth.” It’s just a sta­tis­ti­cal mod­el that has learned the pat­terns in the text. It is not reasoning—just pat­tern match­ing. Giv­en the mod­el, the prompt, and a ran­dom num­ber gen­er­a­tor, it will gen­er­ate text that match­es the pat­terns it learned dur­ing training.

Related reading

“Update: AI Churchill Quotes- Yes, We Have No Bananas,” 2025.

“Mary Soames Cen­te­nary 1922-2022: A Remem­brance by a Friend,” 2022.

“Drunk and Ugly: The Peren­ni­al Quo­ta­tion Chase,” 2022.

“Churchill the Drunk. Or: Fas­ten Seat­belts on Bar Stools,” 2022.

“Jibes and Insults: Churchill Took as Good as He Gave,” 2024.