My Recent Podcast
My latest guest on Brave New World was David Halpern, former professor from Cambridge University who went into public service and directed the “Nudge Unit” (formally known as the Behavioral Insights Team) at Downing Street during the Tony Blair and David Cameron administrations. David is the author of the book titled Inside the Nudge Unit: How Small Changes Can Make a Big Difference. It’s a great read.
David’s work through Britain’s Nudge Unit has brought a scientific data-driven approach to policy making, where interventions are carefully crafted and their behavioral impacts are measured at population-level scale. It involves designing nudges and running large-scale randomized controlled trials to understand how people behave in response to them. An example of a nudge might be one where the default choice is to opt into pension plans instead of the opposite: the goal is not to restrict your set of choices, but to nudge you into the more desirable one. Other nudges can be things like subsidies, incentives, or even simple messages like “Most people in your hood pay their taxes on time.” As David puts it, if you want to nudge people towards a desired behavior, encourage entry. Another strategy could be to discourage exit from the desirable path. I can imagine a future when AI could assist in designing the right nudges, although that sounds mildly dystopian.
What is clear is that nudges are an important part of the emerging world of AI-driven public governance. As David claims in his book, nudges have had tremendous policy impact globally, leading to millions of healthy life-years saved, hundreds of thousands of people finding work faster, and millions in revenue being brought forward.
So, check out my conversation with David. I learned a lot reading his book and during our free-flowing conversation:
https://bravenewpodcast.com/episodes/2024/02/22/episode-78-david-halpern-on-nudging/
Resistance is Futile: Embrace the Machine
AI is on a tear. Last week Open AI announced that AI can now create videos from text. It has only been a year since AI took the world by storm with ChatGPT, AI’s first killer app toy. It marked the first time we could converse with the machine, however imperfectly, about anything, as we do with humans. Now that the machine seems to understand what we say to it, it can create all kinds of things that we specify in natural language. It wouldn’t have been able to create things for us if it didn’t understand us. In a sense, text-to-video AI, however imperfect it might be at the moment, would have been difficult if not impossible without the ChatGPT capability.
This new and improving ability of the machine to communicate fluently has far-reaching consequences on all aspects of our lives. It provides the AI with an unprecedented amount of high-quality training data about humans, that it acquires in parallel with its normal operation. So far, it has learned about the human world only indirectly, by analyzing the publicly available collection of human expression on the Internet. Now that the lines of communication are truly open, the machine has a huge amount of training data about our thoughts, desires and feelings to learn from. Such a machine can do all kinds of new things for us, like creating movies or legal documents and helping us live better lives. An individualized version of “Her,” as in the brilliant Sci-Fi movie, doesn’t seem that distant.
What is becoming clear is that resistance to AI is futile. There’s no place to hide from its fast-growing capacities. It is already redefining how we live and work. Understandably, there’s fear that it will take away our livelihood. Last year, the Writers Guild of America, which represents over 20,000 writers for film, television, radio and new media, went on strike for almost six months to demand “guarantees that artificial intelligence technology will not encroach on writers’ credits and compensation.”
Unfortunately, that’s wishful thinking. It is difficult to muzzle an innovation that is good for business or the consumer.
SAG-AFTRA, a larger union which represents over 150,000 actors, announcers, broadcast journalists, dancers, DJs, news writers, news editors, program hosts, recording artists, singers, and other media professionals, went even further, asking for two percent of revenues from streaming shows. Its president Fran Drescher, star of the TV show “Nanny” said “We demand respect! You cannot exist without us!”
Again, that’s wishful thinking. From a business perspective, AI will help humans do much more with less effort. It will boost productivity and profits. It will especially gut professions such as traditional set design, which is incredibly costly. Existing talent will be replaced by more AI-savvy people across the economy who are able to leverage AI.
What Will Humans Do?
Creative people aren’t the only ones who are worried. In my class last week, students asked me whether there will be any jobs for them considering the rate at which AI is advancing. For example, what is the future of systematic investing, they asked. Will AI do that for us? Will the tech behemoths who “observe everything” have the data advantage over traditional players? Will Google or Microsoft offer better investment tools or advice than Fidelity of Morgan Stanley? And why do we need so many financial analysts and money managers if AI knows more than them? Is the bulk of expertise of these professionals also a step away from elimination, like members of the WGA?
These are great questions.
We should keep in mind that AI makes mistakes or can produce less than desirable outputs. For this reason, it will replace the only the low-risk jobs for the foreseeable future, but it will some time before we trust AI with the risky ones. By low-risk, I mean jobs where the machine will do better than humans despite its mistakes. Human traders, for example, are prime candidates for displacement if the machine generates more revenue than them for the same level of risk despite its mistakes. On the other hand, traders who are savvy dealing with data and can help design such machines will be valued. Similarly, lawyers who fill out boilerplate contracts are replaceable, but those who can ensure that there are no loopholes in a complex contract will be highly valued. Would you be willing to accept AI’s verdict on a complex legal contract? Unlikely. Likewise, screen writers who can offload the grunt work to AI and create original high-quality scripts that the AI can’t produce will be valued.
How will this impact the aggregate level of employment? It is hard to say at the moment. If there is a lower amount of aggregate work to be done by humans, we will need government to step in to help people during this time of transition, as many people will need retraining and a robust safety net. If, on the other hand, there is more work to be done due to the enhanced productivity that enables new kinds of things, that’s great news all around. Regardless, we should remember that every disruptive innovation has caused pain but also eventual gain.
Moreover, my sense is that the future labor market is already adjusting to the new reality. I’m probably encountering a selection bias, but I see that entering students are much better equipped to work with data than they were ten years ago. More of them have learned how to program as kids, and they’ve had access to much better learning materials that keep improving. It’s a great time to learn if you have the motivation.
Talking about motivation, I tell my students that the future presents tremendous opportunity for entrepreneurs. I tell them that if I were thirty years younger, I’d be looking to create another AI company. Now is the prime time for AI startups. Opportunity abounds.