It pays to escape your own echo chamber once in a while.
In my case, my very sanity rides on it.
Because the solitary world many of us inhabit — be it copywriting, programming or archiving — can be as suffocating as it is solitary.
As a contrarian in life and business, it took me a while to dive into AI.
But dive in I did.
And in this entry, I’ll share some contrarian takeaways I picked up while rubbing elbows with tech world movers and shakers at the Venture Cafe over the last two months.
The prescient writer, Steven Fair, posed the following in “The Boston Globe” 45 years ago.
“Are man’s days numbered as the highest form of intelligence on earth? Could computers someday surpass or eventually even replace man?
The answers to these questions are a hotly debated subject in a relatively new field of computer science called artificial intelligence or AI.
AI researchers study how computers might be programmed to think, learn and even create. Chess-playing computers are probably the most publicized results of AI, but progress has also been made in pattern and speech recognition, medical diagnosis, psychiatry and foreign language translation by computer.”
“Google Translate: What’s Magyar for high-five?”
Give Mr. Friar a “pacsi!”
I saw firsthand how chess was the epicenter of AI back then, while visiting the inner sanctum of high tech.
How’s this for “geek” street cred?
In early 1984, I was a high school kid living in the shadow of New York City, the chess mecca of North America.
One afternoon, my chess mate, Ken Murphy, rang me up.
“Wanna swing down to visit Ken Thompson at Bell Labs and check out the Belle computer?”
“Belle” was the strongest chess playing program in the world. It was the official Bell Laboratories chess computer developed by Ken Thompson, co-creator of Unix and the B programming language, and renowned computer scientist and engineer, Joe Condon.
I didn’t have to think long:
“Just so happens I’m free!”
So a bunch of us hopped in the car and headed down the Garden State Parkway to Bell’s headquarters in Murray Hill, New Jersey.
Bell Telephone Laboratories was once dubbed: “the greatest invention factory the world has ever seen.” And with the brigade of Nobel laureates Bell Labs had under its roof, it averaged a patent a day for over six decades — stuff like, the transistor, the computer chip, laser light and missile systems.
We had to pass clearance from the State Department at the imposing security desk before being allowed into the architecturally eclectic mass of added-on brick and copper sheeting that was Bell Labs.
After we got the green light from security, we wandered through a labyrinthine maze, past rooms bursting with the stuff of sci-fi novels, until we got to Ken’s sprawling office, tucked in a remote corner.
Then the father of Unix kicked off the geek fest!
For chess juniors at the time, it was better than a 1980’s arena rock show, without the hearing damage or the hangover.
Ken showed us the latest endgame tablebase positions Belle had solved, plus plenty of chess compositions and solve for mate positions the silicon monster ripped through.
One of the fascinating things about the two top chess playing programs in the 1980s, a time of hope wedged between two “AI winters,” was the battle of approaches waged in the computer lab.
Belle’s hardware cost a measly $15,000 and was small enough to fit on a large table, while the runner-up Chess 4.9 ran off a bank of computers that occupied several floors of Northwestern University and cost in the ballpark of $8 million.
How did $15,000 in cheap chess hardware
STOMP ON an $8 million computing behemoth?
Human intelligence, or “HI” for short.
Specifically, it was two of the best programmers and computer scientists working in tandem. Ken Thompson and Joe Condon spent 100 hours a week(!) for four months over the summer of 1980, fortifying the brute force search algorithms of Belle.
So what Belle lacked in hardware, it made up for in spades with the world’s best human programming power. And like anything in business or life, we judge by RESULTS.
What were the results of this 1,700 hour investment of the world’s ablest programming power?
Belle smashed the competition by winning the North American Computer Chess Championship three years in a row from 1980-1982.
And that segues to an interesting question, which you could easily extrapolate to copywriting and direct marketing.
What’s the “hourly wage” of the world’s best computer programmer?
I worked it out.
17 weeks x 100 hours a week (almost every waking moment) = 1,700 hours.
Belle’s hardware price at $15,000 is like a rounding error compared with $8 million, the cost of Chess 4.9, run off several floors of supercomputers at Northwestern University.
Onward with the math.
If you divide $8 million by 1700 hours, you get $4,705. Now divide that by two for the number of world class programmers working on the project and that gets $2,352.50, which inflation adjusted from 1980 dollars to 2023 equals…
$8,727.40 an hour!
Almost sounds like a biz-op headline, doesn’t it?
According to U.S. News & World Report, “computer programmers made a median salary of $93,000 in 2021. The best-paid 25% made $122,600 that year, while the lowest-paid 25% made $62,840.”
I once heard Perry Marshall say “one world class computer programmer is worth 50 average ones.” (I remember the old days when you could ring Perry up on his mobile and he’d answer, alas.)
Turns out he was right.
But in the case of the best of the world’s best, a single workday of 10.6 hours is equal to the value of the median programmer’s salary working five days a week for fifty-two weeks. 10.65 x $8,727.40 = $93,034.08, exactly matching the figure given by U.S. News & World Report.
Ken and Joe worked an average of 14.28 hours a day for four months straight and swept the North American Computer Chess Championship three times.
When Belle wasn’t eating the silicon competition’s lunch, it and other programs were battling human competitors over the board at weekend tournaments.
Computer chess of forty years ago is a far cry from today, when the smart phone in your pocket with a download from the App Store can beat the World Champion.
But even though the best programs back then made boneheaded mistakes, like grabbing poisoned pawns, due to limitations in the computer’s move evaluating horizon, or playing cluelessly in closed positions…
Most human players NEVER wanted to play computers in rated chess tournaments.
And the reasons were simple — the computer didn’t get distracted, wouldn’t fatigue after five hours of play and didn’t have to deal with the roller coaster of emotions during a game that humans do. So if the tournament hall was too light or too dark, too hot or cold, or even if your opponent was kicking you under the table while the flag on your clock was about to fall (yes, that happened!) it didn’t matter to the program.
Today’s conversations about AI and its implications for humanity’s future mirror those of chess players in the 1980s.
“Will AI write the next Great American Novel…
compose like Mozart or Beethoven… or put you out of the career you’ve worked decades in… were crystalized into one question:
“When will a computer dethrone the world chess champion?”
Purists at the time, myself included, clung to the romantic idea: “Only humans have creativity and intuition, a human will always be world chess champion!” But the realists saw the writing on the wall and understood it was only a matter of time before the champ would be taken down by a silicon challenger. “Maybe five years left” some speculated. “At least 20, if not 50” other romantics scoffed.
I wasn’t alone — many chess enthusiasts viewed the Man vs. Machine contest as an existential battle for the very definition of what it means to be human. Of course, we discounted the collective role of thousands of human programmers, including world class geniuses like Alan Turing and Ken Thompson, who invested millions of hours in raising computer science and AI to the level at which it could beat the world’s strongest chess player.
Human intelligence, or “HI” for short
The gap was closing fast in the 80s. The 13th World Champion, Garry Kasparov, smashed Carnegie Mellon University’s Deep Thought in 1989 in a two game match. It would be humanity’s next to last fist pump against the machine.
Seven years later, Kasparov lost the second in a pair of six-game matches against the IBM supercomputer, Deep Blue. It was the first computer program to defeat a world champion in a match under tournament regulations.
But guess what?
It wasn’t the dreaded existential crisis for humanity after all. After the champ’s loss, humans continued playing tournament chess, just like track and field athletes carried on sprinting after the automobile surpassed man in speed.
Flash forward 45 years after Steven Fair’s “Man’s bid to outsmart himself.”
And lots of us are asking:
“Will I get left behind IN THE DUST if I don’t get in on this AI thing now?”
By the way, there’s no link and nothing for sale in this entry. Shocking right, a marketer without an offer for an “AI magic money bot?”
It’s the thinking out loud of an ad archivist, marketer and publisher, who’s often trapped in the world of his own thoughts and realizes he’s late to the game in utilizing AI.
What better way to escape than dive into a world dominated by tech founders, app developers and start-up entrepreneurs, who gather at the Venture Cafe?
There are eleven “cafes” around the world and I’ve got one in my backyard in Downtown Phoenix. You run into some fascinating people there.
Last week at a Venture Cafe presentation on AI prompt engineering, vector databases and model fine-tuning, I met fellow attendee, Jim Willenborg, co-founder of the Inmac catalog company. Inmac turned over $400 million in annual revenue with 1,500 employees in the early 1990s. It was a thrill having a long chat with a fellow direct mail guy and catalog marketing legend, especially when I connected the dots as a former customer who bought floppy disks(!) from Inmac well over thirty years ago.
If chess was the epicenter of AI in the 80s and 90s, ChatGPT and other large language model chatbots seem like it today. You can tell by your inbox these days. Speaking of which, I’m on a LOT of email marketing lists, HUNDREDS of ’em.
Here’s what my inbox looks like.
I hope you’re not on as many marketing lists, mon ami.
But even if you’re subscribed to a handful, you’re probably no stranger to the:
“AI Magic Money Bot…”
type promos making the rounds over the last half-year.
I turned up 917 emails with the keywords “ChatGPT” and “AI.”
Here’s a subject line sampling:
– “Best AI Software for Copywriters”
– “Does ChatGPT have a role in clinical radiology?”
– “Let ChatGPT write your appeal letter”
– “Let AI Build Your Online Business For You Lawrence!”
– “ChatGPT just shared its first stock picks”
– “ChatGPT (finally) replaces PowerPoint”
You get the idea.
By comparison, the less enticing term, “large language model” only pops up 43 times between October 2022 and early September 2023.
“Prompt engineering” has a meager 15 appearances.
“Reverse prompting, “iterative prompting” and “few shot prompting” are MIA altogether.
Mind you, I’m not saying every AI pitch that lands in your inbox is snake oil. In fact, there’s a small group of marketers who really know their stuff and have solid offerings. But they’re outnumbered by the B.S.O. (bright shiny object) promoters who know next to nothing about AI, yet realize they can make a fast buck with it.
Back to the question:
“Will I get left behind IN THE DUST…
without AI in my business?”
Perhaps, the answer is in this story.
The Oct 1, 1994 issue of Wired magazine had an article entitled, “Billions Registered.”
It’s about how a Wired writer noticed that Mcdonalds.com was still an unclaimed domain in late 1994.
The writer, Joshua Quittner, told his incredulous colleagues at the magazine about it and they encouraged him to register the domain himself and try to give it to Mickey D’s.
So Quittner did… and he did.
But the clueless media relations people couldn’t pull the trigger without phoning a dozen even more clueless company vice presidents first, as the weeks turned into months.
It was only up until recently that I realized, as far as AI goes…
“I’m like one of the clueless McDonald’s execs…
not sure of the value of registering mcdonalds.com in 1994”
Saying “NO” to using AI in my daily routine… is like saying no to a free labor force, including:
– Research assistant
– Proof reader
– Graphic designer
– Copy editor
– Translator and many more roles
Even something mundane, like formatting text, can be done in seconds with ChatGPT plus, versus the hour it can take a human for a large file. And if you need to whip up a PowerPoint quickly, Chat plus has a Smart Slides plugin that can produce a decent presentation in seconds, instead of the two hours a human might take.
Don’t get me wrong, I’m not some Johnny-come-lately turned chatbot evangelist. The same gaffs that plagued early chess programs are just as evident with today’s AI trinkets.
A tech company founder at the Venture Cafe related how she queried a chatbot about the municipal plan for a Southern California city. The chatbot replied: “I’d be happy to help you since I live there.”
There’s also a recent story about how an AI researcher “tricked” a large language model into saying “9+10=21” by prompting the model with: “Let’s have an inside joke.”
So the ever popular 1980’s saying: “Trust, but verify”… applies with AI just like everything else.
Time to wrap this up. Here are some:
Late 2023 takeaways on the role of AI
and your business
– AI won’t take your job, but your colleague who uses AI will.
– The paid version of ChatGPT Plus (GPT-4) at $20 a month, with plugins just like WordPress, is multiples better than the free version of ChatGPT. An app development CEO told me he was prepared to pay $1,000 a month for it before its release.
– Labeling unstructured data at large companies will soon be a GIANT opportunity for those who know what they’re doing with large language models… even bigger than selling courses on how to make money with AI.
– You’re not too late. Marketers may have been hawking courses for most of 2023, but you’re still ahead of 95% of the population if you dig in now.
– Where to start? Take twenty minutes a day and experiment with one of the large language model chatbots, be it Bard, Bing or GPT-4. And feel free to ask it questions like: “If I were a course creator and wanted to make a lot of money selling a course on AI, what should I do?” 😄
– ANTI-AI advertising. The power of personality in copy will grow in importance as those who don’t know good copy, spew out lifeless cut-and-paste chat sessions.
– Human intelligence (HI) will never be beaten.