Discover more from Vasant Dhar's Brave New World
The Battle for Attention
Also, Bots are Kicking Human Ass!
My Recent Podcast
My most recent guests on Brave New World were NYU colleagues Hanna Halaburda and Yannis Bakos, discussing the world of Crypto, Blockchain, and Defi. My two prior conversations on these topics with Antoinette Schoar and David Yermack provide a nice background for this conversation, so check them out if you haven’t heard them.
I started by asking Hanna and Yannis whether there is any positive financial innovation in crypto, and let the conversation flow. The question was prompted by a letter written by 1,500 computer scientists to the US Congress warning that this innovation presents a big risk and does not serve society well:
… we write to you urging you to take a critical, skeptical approach toward industry claims that crypto-assets (sometimes called cryptocurrencies, crypto tokens, or web3) are an innovative technology that is unreservedly good. We urge you to resist pressure from digital asset industry financiers, lobbyists, and boosters to create a regulatory safe haven for these risky, flawed, and unproven digital financial instruments and to instead take an approach that protects the public interest and ensures technology is deployed in genuine service to the needs of ordinary citizens.
We strongly disagree with the narrative—peddled by those with a financial stake in the crypto-asset industry—that these technologies represent a positive financial innovation and are in any way suited to solving the financial problems facing ordinary Americans.
So, tune in to listen to what Hanna and Yannis have to say.
The Social Media Shit Storm
Social media platforms are front and center these days, with large scale layoffs happening as I write this. Several reporters reached out to me last week asking whether we should expect a change in the landscape of social media platforms going forward.
The answer is yes, we should, but what’s driving it, and how could things unfold?
Clearly, the global economy has been a wreck, with declining ad revenues, which is shining a spotlight on the cost side of social media platforms. But the real war is over long-term attention, which is essentially a zero-sum game. Amazon, Google, Facebook and Twitter cornered most of our attention over the last decade because they were early, and “network effects” did the rest, meaning that they became more valuable to people as more people joined the platform. Most of the ad dollars flowed into them, away from conventional media.
But Meta/Facebook and Twitter have had a murky mission, and have gradually lost their way. In good times, mission didn’t matter much, as long as their “objective functions” did a decent job of creating engagement, typically measured by time spent or value added on the platform. In my conversation with Stuart Russell, we considered to what extent the pursuit of such singular objectives – such as time spent, might have led to the unintended side effects we’ve seen, such as political polarization and teen depression. In tough times, murky missions can lead to the wrong objective functions that harm long-term value.
Twitter’s mission is to “give everyone the power to create and share ideas and information instantly without barriers.” It did. Mission accomplished. But some people have been kicked off the platform, so it’s clearly not for “everyone.” It has inherited a content moderation nightmare, described in this funny piece that begins with “Hey Elon.” It could just as well have said “Hey Zuck.”
Meta’s mission is a similar mission impossible, to “give people the power to build community and bring the world closer together.” It did give people the power to build community, but the second part is mission impossible. It also faces a content moderation nightmare.
Consider Tiktok’s mission: to inspire creativity and bring joy. There’s no grand attempt to bring power to the people. (Besides, that’s probably the last thing a Chinese platform would attempt!) Inspiring creativity by giving people the best possible tools to express themselves brings in a large quantity of free content. Joy is about matching people with this content. As long as you can measure joy, the AI will do the hard work. A business model doesn’t get much better.
The proof is in the pudding. Dan Howley of MarketWatch reports that between Q1 of 2021 and Q2 of 2022, TikTok grew by 45% compared to 5% for Facebook. Twitter fared equally miserably.
In our conversation, Dan asked whether social media platforms are like fashions that come and go out of style. I had not considered Twitter as a fashion. If it is, it would imply that the powerful network effects that propelled it to dominance are in peril of unraveling. In which case, Musk needs to clarify Twitter’s mission before it goes out of fashion, and align its objective function accordingly.
AI is Creating Headaches for Teachers
AI bots seem to get smarted by the day, just like Tiktok’s machine. These “language model” bots learn by ingesting all available content on the Internet, which has made them very good at question answering. A few months ago, I shared my dialog with Meta’s Blenderbot about the meat industry, which was incredibly coherent.
Some of these bots can even write better code than many data scientists. They are only going to get better.
The trouble is that these AI models have become so smart that they’re already answering university level questions in some subject areas quite well. One of my colleagues gave the GPT language model a few multi-part data science questions, described in a few hundred words. GPT’s answers were roughly twice the length of the questions, and blew my mind. The answers included all the important things you would expect in data science, including what the objective function could be, the corresponding “target variable” to the optimized, and the actions that the firm could take. It even solved some expected value equations, which were expressed better than many of my students who answered a similar question. The answers looked very human, including the errors.
In the past, we used systems like “Turnitin” to detect plagiarism. What know? Ask everyone to check in their phones and laptops before an exam? Jam access to the Internet? Or answer questions the old-fashioned way: in person, or on paper. How crazy is this brave new world we’re creating?