If you’ve yet to jump on board the ChatGPT bandwagon, you may have found yourself at one of those rare junctures in humanity where the people who wait the longest make the biggest advances. You’ve done nothing to earn it but congratulations nonetheless.?
For the sake of all of us, I voluntarily jumped into the AI snake-pit as soon as those charmers started their melodies. I’ve been on the front lines of AI, learning the lingo, building apps, and speaking to AI developers, to understand who’s behind it and what it does. Long story short, it’s a murky world and we should all tread carefully.
Here then are 7 human-created reasons for slowing down your entry or avoiding the world of ChatGPT altogether:
- Fake: The way they bang on about AI - this whole conceit that AI is going to take over - you’d think the A stands at least for “advanced”. Quite the opposite. The “A” stands for Fake. DeepBlue was never a chess prodigy devising a whole new way of playing the game, they simply fed thousands of endgames into the machine. That’s cheating. If DeepBlue was fake, at least it excelled at one thing. ChatGPT doesn’t even try. The clue again is in the initials. The ‘G’ stands for general. In practice, that means you’ve got a very basic version, not suited to any particular task. Call yourself a barista, with your fancy $100 home Espresso machine, but your coffee isn’t all that. No offence.
- Unreliable: All of the current wave of chat models are works in progress and whilst the capability may be improving, there’s constant tinkering or “fine-tuning” in the background which means that the output for the same prompt can be entirely different. I was 2 weeks away from being crowned the Picasso of AI when Stable Diffusion changed its code and my art is back to looking like it was drawn by a spider wearing roller-skates. They call it “behaviour drift”, I call it “a pain in the butt”. Either way, it makes it very difficult to develop with this software. AISnakeOil reports that the fact OpenAI provides snapshots of their earlier models “only underscores how hard it is to build reliable products on top of them.”
- Slow: ChatGPT, Bard, and the like, were trained on all the data in the world. Among that data you’ve got the kind-but-ignorant folks on Yahoo Answers pumping disinformation into the mix. ChatGPT can’t discern the difference - that’s your job and it takes time. There's now a meme from developers that what used to take 4 hours to programme and 8 hours to debug now takes 5 mins to programme on ChatGPT and 24 hours to debug. In other words, it takes twice as long. Why? Because being the general purpose tools they are, they only do 80% of the job and many times trying to do the remaining 20% takes longer than doing it all yourself.
- Illegal?: We’ve seen this behaviour before in the likes of Airbnb and Uber - big companies entering the frame with "regulatory arbitrage”, taking advantage of legal grey areas and then ploughing money into lawyers? when the inevitable pushback arrives. The law impacts you too. Last month you were going to travel the world by renting out your Airbnb and this month you’re being sued by the local council for breaking hospitality laws. ChatGPT, Bard, and the people who brought you this technology railroaded any privacy concerns. There’s a realistic possibility that the competition authorities will catch up and some of the AI tools you start using will be pulled. Do you really want to invest in a technology with such shaky legal foundations?
- Environment: People who haven’t tried ChatGPT still inhabit lush gardens. Those who have been asking Chat “which is the most environmentally friendly car” now live in deserts. At the same time as we’re washing out our baked bean cans for recycling and looking for any desperate way to become carbon neutral, we’ve introduced yet another environment-trashing technology. During a 20 to 50 question conversation with a chatbot, they estimate AI could “drink” a 500ml bottle of water.
- Cost: And then there’s the subscription fees. Before you get giddy thinking this a free-for-all, it’s not. Nor was Google. The 2 largest travel companies spend £10bn a year on Google advertising. Google may be free to you at that point of use but check the line of your hotel bill that says “hidden fees”. ChatGPT gets you more directly. It’s free to play but all the useful stuff is subscription based. I’m now paying £600 per year on AI - 2 years ago that was zero.
- Unnecessary: Some argue there has been no measurable increase in productivity despite massive investments in and adoption of computers. In what’s known as the “productivity paradox”, growth slowed in the 1970s and 80s as computer use expanded, challenging the notion that they would boost efficiency. And we all know productivity ground to a halt with the invention of PowerPoint. Suddenly everyone could force us to sit through an endless stream of pointless pie-charts. If you don’t need yet another distraction in your life, you may want to give ChatGPT a wide berth.
The Golden Rule: If something sounds too good to be true, it isn’t. And if something is telling you that it’s fake, take it at its word.
Rest assured, I will remain on the front lines, keeping a close eye on this brazen AI upstart. You can remain as you were, with your calculators and your printed maps to hand, and I’ll be the first to let you know if and when it’s time to start dancing to the tune of ChatGPT.
Postscript: I asked ChatGPT to proofread this article and got back “I'm unable to produce a response”.