Our World of Abundance Needs a Digital Personal Curator

Our World of Abundance Needs a Digital Personal Curator
Generated by Stable Diffusion - from lexica.art

It would take 468 years to listen to the entire music catalog on Spotify, 163,116 years to read every book on earth, and 95 years to watch all feature films ever made. On YouTube, every minute, 500 hours of video are uploaded. It would require 82 years to watch just the added material from the last 24 hours. Without considering other sources such as podcasts, news, blogs, magazines, and social media, the immense volume of content available would necessitate multiple lifetimes to ingest.

During my childhood, the landscape was entirely different. There was only one national TV channel that started at 6 pm and ended at 11 pm. The typical programming began with kids friendly content, followed by the daily news, a TV show, and finished with a movie or documentary. Books were limited to the inventory carried by the local libraries and music came from mix tapes and a few radio channels; It was tranquil, predictable, and didn’t require much mental energy to decide what to consume.

What may sound like a blessing in today’s overwhelming world with abundance felt very different back then. It was a feeling of content starvation. You were out of luck if you were curious about particular topics and had distinctive musical preferences.

Over the years, the situation improved. Initially with VHS players and later with Digital Satellite TV that blasted the offering by floating hundreds of channels accessible 24/7 by pushing a remote-control button. Today, we have clicking and scrolling; back then, it was called zapping.

I thought we had reached a peak level of media lavishness, but I had no clue about the massive permanent storm that would hit us in the coming decade.

Fast forward, the Internet came along and enriched our lives in extraordinary ways. The web browser unlocked access to knowledge and information about any topic we desired, a massive library of books, and a huge catalog of media, all accessible from a pocket device.

It felt magical and wonderful. But over the years, the sensation of satisfaction from an outstanding level of instant optionality slowly ceded to a sense of indigestion. Throughout my life, I gradually moved from a state of content starvation to a feeling of content satiety to a condition of content gluttony.

Today, the usual suspects that get pointed at are social media apps. Those are the ones that mastered the game of attention and digital overfeeding, which makes them the most problematic. But the issue is much broader than that. Beyond social apps, we still have to deal with news, blogs, newsletters, personal messaging groups, work messaging channels, discussion forums, emails, digital magazines, podcasts, audiobooks, eBooks, online classes, video streaming, and music services. The pull is triggered by the various topics we are curious about and by the trending “current thing”. The broader our interests are, the more severe the problem gets.

One recent example is the new episode of the Lex Fridman Podcast with Balaji Srinivasan. Two brilliant minds I enjoy listening to. There’s only one minor glitch; it’s almost eight hours long!

Our brains are not meant to process that much data throughput daily, and the effort spent on deciding what to consume adds decision fatigue on top of the already challenging matter.

So how can we escape this golden trap?

There’s no permanent solution yet. The only fix we have, for now, is to manage the situation manually using band-aids such as time-boxing Internet usage and enabling tools that limit phone habits.

The ultimate answer has to come from a technology that would allow us to engineer a healthy kind of artificial scarcity. The kind of scarcity that is created from abundance rather than shortage. Two fundamental shifts need to occur to lay the foundation for such a solution.

1.    A shift in the current business model

We do have to rely on recommendation engines as it's impossible to keep up with the immense media catalogs and the phenomenal amount of user-generated content. The existing algorithmic engines theoretically surface the best content for our interests but in reality, serve the interests of the platforms by keeping us hooked for as long as possible.

Both the ad-sponsored and subscription-based models lead to non-aligned incentives between the platforms and the end users. Ad-sponsored platforms need to keep us around to serve ads that represent their source of revenue. So, their prime customers are the advertisers. Subscription-based platforms get their income from consumers, which may seem like a more aligned model. The reality is that they also need to keep us addicted for as long as possible. If the monthly consumption drops drastically, customers will cancel their subscriptions.

For an algorithmic recommendation engine to truly serve the interests of consumers, it has to be:

  · Unbundled from the platforms

  · Paid for directly by the end users

2.    A transition from centralized recommendation algorithms to local ones

Unbundling the recommendation algorithms from the platforms is necessary but not sufficient. An independent central business running our unified interest graph across all the media would, in principle, better serve the end-user's interests that directly pay for it. But that would only be true if we assume that the central service provider would always act with integrity, would never be influenced by more powerful entities and, in times of global conflicts, would not side with its origin country and target other nations with “info wars.” In other words, that entity would effectively control the tap of what people see and what they are blind to. It would expose us to mind manipulation risks, an order of magnitude more significant than what's occurring today. Even if there is more than a single provider, consumers will be stuck with the engine they have been using as it’s the one that learned their interests. Switching would be impractical.

What also needs to occur is an evolution from centrally run engines to locally run ones.

For example, Stability AI chose that approach when they released Stable Diffusion as a pre-trained open-source image generator anyone could run on their computer. In contrast, Open AI decided to keep central control over their similar system.

"I saw a vision of an intelligent Internet where everyone has their own AI as opposed to centralizing AI…. I thought it needs to be expanded not only through the open source, but across the world"  - Emad Mostaque, founder of Stability AI.

Similarly, the interest graph models could be initially trained based on massive global data sets on a powerful central computer and then made available for each individual to download and run locally. The models would then expand as they learn about specific people's habits and interests while using them.

With such a setup, the attention maximization dilemma could be solved as each person would get their own Digital Personal Curator (DPC). A curator that would have their best interests in "mind." and not be controlled by a major corporation.

The DPC would tone down the overflow of information to the optimal level by considering our objectives, available time, and, most importantly, our well-being.

What would such a system look like?

The Digital Personal Curator could be an AI web browser that acts as a filtering layer between users and the current platforms. It would connect to a locally stored interest graph owned by each individual and then show an automatically generated daily page with only the right dose of content from the various sources that should get our attention. It would also consider our goals, priorities and what’s already on our daily calendars.

Furthermore, the DPC could solve, for example, the challenge of the eight hours podcast mentioned earlier by automatically summarizing it into multiple short versions. Consumers could choose between a fifteen-minute summary and a one hour condensed version depending on the desired level of depth and available time. We have already started seeing language summarization AI models emerge with promising early results, and the rate of improvement will likely be exponential.

Even with all the challenges of the current world of abundance, it is still a much better world than a world of scarcity. We have to navigate the present phase the best way we can. The short-term answer could be narrowing our choices by adopting a minimalist approach. It would require self-discipline, but possible through repetition and consistency.

Once the locally run AI curation technology matures to the optimal level, we can enjoy the benefits of abundance while being shielded from its side effects. Those benefits include better quality of life, more time for meaningful activities, and less fear of missing out.