We Should Have Seen ChatGPT Coming
OPINION
My brother-in-law is a professor at a leading UK University who teaches innovation studies in a business school. After learning about ChatGPT my husband asked him what he thought about it:
Well, there was a time when people were worried about what would happen to all the horses
… after the invention of the car.
I don't know where you were when you found out about ChatGPT
I was on the beach, on vacation, bantering with my teenagers about a seemingly hair-brained business idea around the fashion industry.
The teenagers were like, nah, that's a stupid idea, and my friend said: "Oh yeah, let's see how stupid this idea actually is!"
He took out his phone, typed a simple phrase into the search bar on his phone, and hit the paper airplane icon.
Immediately, the flashing cursor began to explain this "stupid idea" with eloquence and accuracy. It had examples and details citing actual people and items. It then proceeded to outline a fairly cogent business plan.
I'm no investor; I write ideas and pitch them as projects and stories to people and institutions. But this explanation, with the adjoining workflow, with a rationale and demographic analysis, was good enough that I could have taken the text, chopped it up, styled it into a slide deck, laid a fairly thin Don Draper layer to it, and walked into any boardroom to pitch it.
It gave me a weird feeling.
Should we have expected ChatGTP?
For about a decade I've been feeding my own voice interviews into AI transcription machines…anything to not have to do this by hand, or pay someone to do this at the then going rate of $1.00–2.00 per minute of audio…so a 45-minute interview could be as much as $90.00 to transcribe, until about 2017. In my line of work, one hour is just a drop in the bucket of the number of hours of interviews that need to be transcribed for a documentary film or audio project. Until recently, this has been a line item to consider.
The earliest one that I found circa 2013 was connected to the IBM Watson project….I recall it required some odd login protocols, and then you had to go to the demo, click the JSON tab, upload an MP3 file, and then wait for the transcript.
Watson didn't hear accents well; it preferred Midwestern English with correct grammar and pronunciation. When I began to work on a project that required transcribing a heavy Australian accent, I discovered the transcription results were humourous but useless; cleaning up the transcript took as long as it would to transcribe it from scratch. Watson also definitely couldn't detect any languages other than English.
When I first began with Watson, I found it mildly interesting to consider where all these hours of Machine Learning files would end up, but I didn't go much further than that. As I Google that same question today, I learn that the Watson-transcribed interviews that I uploaded may have tangentially helped propel AI in some video games.
Way back, I would casually use the term "machine learning" as a definition as to why these transcription tools were getting better; although, I didn't understand what that really meant, nor where it could all go, despite the fact that I had already read and basically ascribed to the ideas outlined by Donna Haraway in the Cyborg Manifesto.
When I was able to recognize that these ‘machines' were getting better at their job, for a brief moment I congratulated myself that I had done my work here too; all that helpful data from interviews with whomever from wherever. Maybe it was my feminist response to the work.
From where I was sitting, it was free help. I pressed Upload and then Download. Thank you very much.
More recently, I've used Otter.ai, Descript and to a lesser extent, Sonix.ai All of these programs were a godsend to someone like me: free transcriptions (there's a threshold of time or words for "free" accounts each month) for audio interviews that otherwise would have taken me hours upon hours with the existing tools of the day like Wreally, or cost hundreds of dollars to pay someone else to do it.
But with the arrival of ChatGPT, along with other AI writing tools, the penny has dropped for me. I now see where all that work went; these transcription software platforms have all contributed to NLP, or Natural Language Processing.
All those hours of interviews that I uploaded that allowed "machines" to sort through pronunciation and emphasis, to determine the difference between diction and dialect, and then to watch and calculate repeat turns-of-phrase…they've all been very helpful towards the joint effort of building a massive database that has allowed machines to learn how humans speak and think.
And guess what, you've been contributing too.
Give yourself a small tick beside the NLP things that you have used recently:
- Siri speech-to-text requests
- Alexa-anything
- Hey Google…show me a recipe for…
- Apple.com technical support chatbot
- Google Translate
- Grammarly
- Predictive texts
- Netflix recommendations
- Follow suggestions
- Transcription software
That's just the obvious stuff. There are many more layers to this if you want to go deeper into this universe.
If I look only at transcription software, I can say that within only a few years I've watched the accuracy rate improve massively. They've gone from only being able to transcribe words if. they. were. said. clearly. and. separately. To now: where they can transcribe with about 90 percent accuracy a four-way interview with people who speak with different accents including non-native English speakers.
I've been guilty of this head-in-the-sand approach before
In these "olden days" of the Internet, by which I mean the early naughts, the general public interacted with Technology and the companies that control and create technology, quite differently.
Take Google as an example: before it went public in 2004, to non-Silicon Valley insiders, Google was just a no-frills, zero advertising, one-page search engine where you could ask basic questions, and get basic answers. Google didn't start out as the end-to-end solution from doorbells to Drives that it is now.
The early days of a Google search would only yield information from websites (a new concept back then). Or, from existing texts that had been either uploaded to the Internet, or newly created for the Internet (also not plentiful at that time). Plus you had to factor in data rate…who remembers the dial-up machine sound? The Internet was niche when it began.
Academics stuck to actual libraries with hardbound books, or sat in dark rooms scanning microfiches for research for another decade. Wikipedia was just getting rolling from its conceptual beginning in the 1990s to when it launched online in 2001.
The early Internet was good enough to discover names of people, published items, travel information and print out long rambling driving instructions on Mapquest…it was not the vast pool of anything and everything that exists today.
There's another thing to consider…computers were not "personal" back then. Computers were used for work, generally at work, and if you had one at home they often sat on a desk in a corner somewhere with a dust cover on them. "Laptops" looked more like collapsible desktop machines, unless you could afford a fancy Macbook, which were five times the price of anything else.
In 2018, the website 24/7 Wall St published an analysis of how much computers cost based on the year you were born and then adjusted for inflation. If you were born in 1978, the notable IBM 5110 cost almost $10,000, which after you adjust for inflation, is over $38,000.
In 2001, an Apple Powerbook G4 cost $3,500, or almost $5,000 adjusted for inflation; in 2002, a Toshiba Satellite 1995 cost $2,499, or about $3,500 after you adjust for inflation. In 2018 (the last year cited using this same methodology) a Huawei Matebook Pro was just $1,200…and today if you were to order a new Macbook Air with a new M2 chip, the entry-level price is $1199.
Seeing computers as anything other than an expensive but necessary tool took vision; this is something most people did not have in the early 2000s.
To explain to you how limited some people were, I need only look in the mirror. In 2001, while at the SXSW Interactive Festival to present my cutting-edge interactive project, I skipped a Keynote delivered by Eric Schmidt. I even worked inside this world, the confusing fledgling days of computing just after the first dot-com burst of 2000, and yet I actually remember saying aloud to my colleagues:
What could be interesting about a presentation about a search engine?!
Here's how I rationalize this critical error at this point in my life: The world was loosely divided into Geeks and non-geeks. Geeks understood the power, relevance, and importance of computing. Geeks could see that computers would eventually touch everyone and everything; non-geeks were still out in movie theatres, reading books, binge-watching DVD sets of 24 and The Sopranos, or recording appointment television on their PVRs. Some of those types like myself even created new ways to tell stories using computers…
Yes, I went for free coffee at the conference while Eric Schmidt was delivering his keynote address to a small auditorium in 2001 in Austin, Texas. There. I've said it. I'm making peace with this fact.
This story is revealing somewhat of my age – almost 50 – but it also reveals something else: while I've been working away in the media world, which is inextricably linked to technology, I must also recognize that I've also helped to build them.
I am, and chances you are too, one of the humans who have been feeding the bots for years. I've even been paid, in a sense, for doing this (by taking the free transcriptions). I've done this willingly and half-knowingly.
Only now, after the fact, do I see the importance of it all. I also see, with the benefit of hindsight, that based on all the voice commands, interview exchanges, and preference selections, that we all have fed the "machine" for a decade. And what has been birthed is just exactly what we should expect:
An AI chatbot that has natural and native speaking ability, which include patterns and connections, ideas that have depth and innovation, supported by ample customer data…and an entire social media marketing plan.
Doesn't that just about sum up the modern human/media/entrepreneur condition now?
Should we be scared?
Should we fear this evolution of computing and technology? Is it radical to call it life-changing, or is it just hyperbole to catch a headline?
Let's circle back to that insight, from my brother-in-law at the beginning of the article:
Well, there was a time when people were worried about what would happen to all the horses
… after the invention of the car.
Ask yourself this: Are you willing to go back to riding horses to commute to work? Is your next horseback vacation going to be enjoyable?
Clearly, the arrival of the car has completely re-ordered everything we know about our civilization: It governs how we live, where we live, how we work, where we work, and what we do for fun. Cars have also helped to burn a hole in our atmosphere and accelerated climate change.
But should we fear it? Should we fear the arrival of the AI bots to help us do our work for us?
This is a more complex question, but not one to completely ignore and treat as a Luddite's revenge.
Now that we now have more than a century of information to work from to analyze the effects the arrival of the car has brought upon our society, I firmly hold on this position:
Yes, we should be aware, if not slightly fearful, of new inventions that displace a previous, or existing systems.
But that doesn't mean we should wish them gone; to imagine our civilization without cars is basically preposterous.
Perhaps the way to look at tools like ChatGTP (including its own GTP-3 which it is said will outstrip the ability of the 1.0 version handily) is to be forward-looking…not looking back at the barn door where the horses live.
There will be others, and soon. Google live-streamed an event on February 8, 2023…not typically a date on the Google annual calendar….and without surprise, we learned that Google's competitor to ChatGTP, Google Bard, is on the horizon. Clearly, they feel threatened by the fact that they were beaten to market in the AI bot challenge.
AI bots call for a different approach to technology – not just an excited grab and run
Maybe it's more about creating a culture of safe usage, and then building a body of rules and codes of conduct to govern its use. These bots will compel us to get smarter as a result, both to recognize the difference between human and bot…and then also when to use them to ramp up proficiency and use them in a way to increase our own productivity. They will also make us lazy in the ways that technology has already (I used to know about 100 phone numbers by memory…I'm down to about five now).
Just like the car built suburbs, the AI bots are definitely going to make us form new life habits and work protocols. AI bots will most certainly impact education, entrepreneurship, the information and copywriting industries…to name just a few that I can see in my sphere of experience.
Somehow, perhaps even against our own better judgment, we are going to be compelled to view AI chatbots as an obvious evolution in the knowledge and information economy that we live in.
Indeed, it's daunting to consider this is just the beginning; the seatbelt was not part of the original design of a car. But cars got faster, roads got longer and straighter, speed limits went up and the usage pattern of cars changed. Car accidents have always been a reality; but after someone crunched the numbers and realized that a small invention – the seatbelt – could save lives, laws were changed, and usage patterns were modified. As a result, many lives have been saved.
What will be the seatbelt test of ChatGTP (and the others that follow it?)
So, yes. Yes, we should have expected ChatGTP, because we've all been helping to create it for years. We've done this through our various time-saving or convenience-based interactions with technology.
But rather than trying to imagine that it doesn't exist, or that we can change it now that it's here, perhaps we should adapt with it, make it work for us and with us, not against us. And somehow figure out how to get smarter, or work less, because of it.
Samantha Hodder is an audio producer and writer. If you love narrative podcasts as much as I do, subscribe to my Substack Bingeworthy.
