Call Me AI
I was intending to talk about AI once the froth around a ChatGPT and the other Large-Language Artificial Intelligence models had died down but a couple of announcements this week made it seem like a propitious time to do so. More of those later
Most of the blizzard of coverage that has followed the recent releases of publicly-available AI tools has either been the pant-wetting excitement shown by teenage girls when their favourite boy band is in town or the cataclysmic predictions of impending doom that would make Turkey-Lurkey look like a level-headed optimist. Let’s hope they don’t train these models on the tidal wave of bilge their introduction has caused or they will actually get stupider.
We’ve seen this before. Have we already forgotten the metaverse hype? It’s been going on ever since we’ve been able to build machines to do what humans do. I’m not suggesting we should be Luddites about this but we should not be gullible fools either.
(To be accurate, the Luddites were not against technology, they were fighting for a share of the benefits and wrecking the machines was the only effective action available to them. Historian Eric Hobsbawm called it "collective bargaining by riot” and it happened against a backdrop of a cost of living crisis caused largely by the Napoleonic Wars. Whether there are any parallels with today’s financial situation in the UK and our current strife with Europe, I’ll leave you to decide).
At this point, I refer to the Gartner Hype Cycle for the introduction of new technologies. Now, there are criticisms that it is not scientifically accurate and uses subjective terminology but I think we can all agree that it represents a reality that we’ve all seen, and we can all figure out where we are with these Large Language AIs right now. Sadly, I doubt we’re at the peak yet but don’t worry, we’ll soon be through it (does ANYONE remember the metaverse? Really? So soon?).
I’ve seen several hype cycle in my years, they seem to come around faster and faster and I don’t think that’s just an effect of my age. Each one promises much and delivers very little. Take the slew of personal productivity tools that have come along since the 1980s. I’m not really sure how much more productive we are, but I do know that the work we’ve saved seems to have been replaced by even more work the tools have created (dealing with email or choosing the right font for a powerpoint, for example). The result is that we’re all working longer and getting more stressed. I don’t remember seeing that in any of the puff pieces that littered the media at the time.
I’m not saying AI won’t have an impact. In fact, it is already. It could be enormously beneficial, augmenting the skills we have to produce better outcomes at scale, such as in areas like medical diagnosis and drug development. It could also be deeply harmful, enabling propaganda and the surveillance state and ‘baking in’ existing biases in our society. Or it could just give us vending machines that can cook and deliver the perfect pizza every time. Let’s just wait and see, shall we?
Video Killed The Radio Star
This week as seen a couple of good example of organisations jumping the gun on this. IBM announced they were delaying hiring to positions that could be replaced by AI, mainly in back office areas. They believe this could be as many as 7,800 jobs in the next few years, a statement both suspiciously precise and also vague at the same time.
BT also announced that they would replace jobs with AI as part of a wider workforce reduction, with about 11,000 going by 2030. They say the main opportunity lies in Customer Service, which will not fill anyone who has had to deal with that function with any joy. Perhaps we can look forward to voice response systems that make you feel you are being dicked around by a real human being.
I think ‘AI’ is doing quite a bit of heavy lifting here and refers to a range of automation tools but, even so, you have to ask how feasible this is. And how wise. I can remember when the latest wheeze was to off-shore these functions to places like India and the Philippines and that hasn’t exactly been an unalloyed success, certainly not in terms of putting the customer first.
That also raises the question as to why they are doing this. It seems to be driven by financial considerations as much as anything, which is no real surprise but makes me wonder whether this is about making noises to mollify shareholders rather than a considered approach to embracing a new technology.
Maybe the first impact of AI will be a positive, if temporary, uplift in the share price.
I think we can look forward to a lot more of these types of announcements from organisations desperately seeking silver bullets to the problem of stagnant, if not declining, productivity. It’s a lot easier that addressing the real issues, after all.
Won’t Take It Lying Down
Then we come to the eye-watering predictions of how many jobs with be lost or ‘degraded’ (now that’s an interesting choice of word!) by AI. The World Economic Forum reckons 25% of jobs will be affected, Goldman Sachs says 300 million and McKinsey’s, never ones to shy away from plucking a big and impressive number out of the air, top that at 400-800 million.
Whilst everyone grabs the Valium and goes to lie down in a darkened room to get over the shock, they all helpfully trill that automation always leads to innovation and the creation of new jobs, so it won’t be as bad as that. Unless it’s your job that’s been replaced, of course, and you can’t do the new jobs. In which case, it will be disastrous. But never mind, that’s the cost of progress. And you’ll be able to get a perfectly cooked pizza from the vending machine in the hall, so it all evens out.
The issue is not really about jobs, however, it’s about who the benefits go to. Given that we’re all working longer hours under more stress because of all those personal productivity tools that mean we are on-call 24x7 and have an endless workflow, we could use this as an opportunity to reverse the tide and make jobs better and less demanding. Given that this will improve profitability, we could improve employee pay and benefits so that everyone had a better standard of living and quality of life. This is what John Maynard Keynes envisaged when we said we’d all be working 15 hour week.
Or we could give all the benefits to shareholders. It’s a choice and I think we all know which one these forecasts are predicated on. So we’re back to the Luddites, who just wanted a fair share of the pie and could only make their voice heard by smashing up the machinery.
You can’t smash up an AI but you can disrupt it and disrupt business. The last wave of automation hit the blue-collar jobs as factories were automated (in the west this was coupled with a move to lower-cost countries, made possible by globalisation). We’re still dealing with the consequences of that, socially, economically and politically. However, until populists like Trump came along, these voices were largely ignored.
This wave is not the same. It’s going to hit the middle classes and that’s going to work out differently for two reasons.
One is macro. Economies need a healthy middle class, they are the market for goods and services. If they are denuded, the economy will suffer. We’ve already seen some of this over the past decade as salaries gave been depressed and living costs have risen, especially for housing. This is going to accelerate that effect. This is cannibalistic capitalism and it’s starting to eat its own tail at a worrying rate.
The other reason is political. The middle class have a voice, a very loud one. They are not going to take being screwed over in this way and will make one almighty fuss. They vote, they organise, they speak out. They have influence and power, they know their way around the system and are used to getting what they want out of it.
The only thing we can be sure of is that all predictions of the future are wrong (by the way, where are our flying cars?!!). When it comes to technology adoption, they are often incomplete in their analysis too. I think there are a lot of factors that will come to bear on the adoption of AI that aren’t even on the radar yet.
Could It Be Magic?
So, just how smart is this technology?
We are very prone to anthropomorphism (the attribution of human characteristics or behaviour to a god, animal, or object). We talk to our pets as if they are people, we give our cars names and ask them to be nice to us, we negotiate with our devices (in the case of printers, it’s definitely Beelzebub in there, right?). We’re doing it with AI.
That’s why people are over-estimating the capabilities of ChatGPT and the like. We imagine it’s a real person responding and we overlook the deficiencies and fill in the gaps. We think ChatBots are people because that’s what we expect and it’s a very limited interaction and vocabulary. Then we project all the attributes of people onto them but if we attempt to have a discussion about philosophy or feelings with them, we quickly realise they are just dumb machines.
So we are beguiled by the idea of real intelligence in machines (supported by sci-fi creations like HAL in 2001 and the idea of the singularity) and are easily fooled by a facsimile of a human response. We think ChatGPT is intelligent because we have been primed to project the attributes of human intelligence onto it and over-estimate it’s capabilities.
However, as this excellent New Yorker article about what AI is by Cal Newport, there is no intelligence there. It’s really a very sophisticated and powerful sifting and sorting mechanism. It’s complicated but it’s not complex. As Newport concludes “in the final accounting it’s clear that what’s been unleashed is more automaton than golem.’
It put me in mind of the the infinite monkey theorem, which states that a monkey hitting keys at random on a typewriter keyboard for an infinite amount of time will almost surely type any given text, such as the complete works of William Shakespeare. What we have right now is the software equivalent of an infinite number of monkeys, but ones who take bits out of books rather than randomly hitting keys. What comes out looks like what humans produce but it’s not anything new. There’s no imagination, no creativity, no novelty. Just a facsimile, a collage of what’s gone before.
Maybe we’re the fools for allowing ourselves to be taken in by this sleight of hand. Perhaps that’s the real threat of AI, it shows how stupid we are and we’ll end up putting the chimps in charge.