Building Change On Preferences And Tech Shifts

Forbes

a day ago

Carefree young woman smiling happily while playing music using a smartphone and earbuds. Cheerful ... More woman with curly hair enjoying her favourite playlist while standing against a blue background.
There's a common phenomenon I've observed pretty frequently with people talking about the AI revolution is now upon us.
They're basically comparing today's tech world to that of the past waves of innovation that gave us the Internet, big data, the cloud and everything else.
To be sure, none of this sprang fully formed out of nothing. It came incrementally – and was built piece by piece. There's a good argument that the Internet laid the groundwork for artificial intelligence work to flourish. At the very least, it provides the playground or environment for digital activity that AI can direct – as when Claude or some other model starts to actively use the Internet as an agent.
So from time to time, we might look in a more granular way at what has happened, and why.
The 15-Year Flywheel
This came up in a recent TED talk by Sungjoo Yoon who began by talking about his romantic exploits early in life, and how he understood the fickleness of a particular partner who decided to leave him for his best friend.
He made the analogy to consumer trends and market changes, where companies have their 15 minutes of fame with a product or service before something else takes over. He called this 'tectonic shifts' in the market, for example, the eventual supplanting of the 1969 command line interfaces that were fundamental to the development of the early personal computer.
'We saw, in 1969, the creation of the Thompson Shell, the first Unix shell that you could just prompt if you knew the terminals, and it gave rise to the modern terminal commands,' he said. 'You could type in stuff, and it would do way more calculations for you. This was the base case of out with the old, in with the new.'
He referenced the rise of GUIs in the 80s, a web-native interface in the 90s, and then, the evident advances of the twenty-first century.
What Makes Change Happen?
Throughout this, Yoon referred to something that he called the preference principle, which suggests that if you have more information about what someone likes, you can create more powerful technology around that, or in other words, more appealing technology.
'I spent time as a member of the research staff at Caltech in a behavioral economics lab, and this is actually an underlying principle that exists all across the world,' he said.
When you look at welfare distribution, it's just a product of what people need, finding out what they need, and trying to distribute and allocate resources in an efficient manner. In order to do this effectively, you need to have honest information about what those needs actually are.'
He covered auction theory and behaviors based on preference principles, like the fear of overbidding. He also cited the job world, and hiring decisions:
'If we want to have efficient labor allocation all across the world, what we need is a lot of information on what people's skills actually are, such that for a skill, a job that requires really high amount of skills,' he said. 'We don't assign someone who doesn't have those skills, and vice versa.'
Then there's interface theory.
'This is something that is incredibly understudied, but crucial in consumer technology,' he said 'What do I mean by understudied? Well, think about the way that people talk about consumer tech form factor. It's very strange because it's not just about the hardware dimensions, right? This is a definition that a lot of people like to use. What are the hardware dimensions of any given product?'
He gave the example of an iPhone and how that device changed over the years.
'It's not actually strictly about the dimensions itself, or else we'd see a more consistent pattern over time. It's also not just the end experiences that those dimensions create.'
Developers, he suggested, are paying attention to these ideas, in trying to market better to a growing audience of users, and new generation with – you guessed it – changing preferences.
'Participation just means you have more aggregated preferences among a wide group of people,' he said, 'and more aggregated preferences means higher information quality, and therefore product quality.'
Yoon cited Google, YouTube and LinkedIn as platforms that are doing this research, and making the most of the information they aggregate.
He also pointed out that there are better ways to source the information itself.
Language Matters
Yoon talked about natural language, and unnatural language.
Basically, he suggested, if you just ask a person to give information, it's awkward and unnatural. by contrast, if you use natural language, and social processes, that's going to be more effective in general.
Natural language, he noted, equals trust.
The Agentic Age
How does this apply to AI in practice?
We're starting to see AI agents take initiative and start to do things on their own. They made to be loosely directed by a human to reach an objective, but they come up with the tasks themselves.
That means we're soon going to have all of these non-human actors in a world that was always being directed by human participants.
That's probably a lot of what Yoon is pointing out with his preference principle, and incremental changes in technology that we have to reckon with if we're going to build a better way forward.