« Dependencies at scals | Main | That was the year that wasn't »


hockey-stick.jpgWeb3, the push to decentralize the net, got a lot more attention this week after the venture capital firm Andreesen Horowitz published guidance for policy makers - while British software engineer Stephen Diehl to blogged calling web3 "bullshit", a "vapid marketing campaign", and a "rhetorical trick" (thanks to Mike Nelson for the pointer).

Here, a month ago, we tried to tease out some of the hard problems web3 is up against. Diehl attacks the technical basis, citing the costs of the computation and bandwidth necessary to run a censorship-proof blockchain network, plus the difficulty of storage, as in "who owns the data?". In other words, web3, as he understands it, won't scale.

Meanwhile, on Twitter, commenters have highlighted Andreesen Horowitz's introductory words, "We are radically optimistic about the potential of web3 to restore trust in institutions and expand access to opportunity." If, the argument goes, venture capitalists are excited about web3 that's a clear indicator that they expect to reap the spoils. Which implies an eventual outcome favoring giant corporate interests.

The thing that modern venture capitalists always seek with (due) diligence is scale. Scale means you can make more of something without incurring (much) additional cost. Scale meant Instagram could build a business Facebook would buy for $1 billion with only 13 employees. Venture capitalists want the hockey stick.

Unsurprisingly, given the venture capital appeal, the Internet is full of things that scale - social media sites, streaming services, software, other forms of digital content distribution, and so on. Yet many of the hard problems we struggle to solve are conflicts between scale and all the things on the Internet that either *don't* scale. Easy non-Internet example: viruses scale, nurses don't. Or, more nettishly, facial recognition scales; makeup artists don't. And so on.

An obvious and contentious Internet example: content moderation. Even after AI has automatically removed the obvious abuses, edge cases rapidly escalate beyond the resources most companies are willing to throw at it. In his book Social Warming, Charles Arthur suggests capping the size of social networks, an idea echoed recently by Lawfare editor Ben Wittes in an episode of In Lieu of Fun, who commented that sites shouldn't be allowed to grow larger than they can "moderate well". It's hard to think of a social media site that hasn't. It's also hard to understand how such a cap would work without frustrating everyone. If you're user number cap+1, do you have to persuade all your friends to join a less-populated network so you can be together?

More broadly - a recurrent theme - community on the Internet does not scale. In every form of online community back to bulletin board systems and Usenet, increasing size always brings abuse. In addition, over and over online forums show the power law distribution of posters: a small handful do most of the talking, followed by a long tail of occasional contributors and a vast majority of lurkers. The loudest and most persistent voices set the tone, get the attention, and reap the profits, if there are any to be had.

The problem of scaling content moderation applies more generally to online governance. As societies grow, become more complex, and struggle with abuse, turning governance over to paid professionals seems to be the near-universal solution.

Another thing that doesn't scale: discovery, as Benedict Evans recently pointed out in a discussion of email newsletters and Substack.

One of the marvels of 2021 has been the reinvention of emailed newsletters as a paying proposition. Of course, plenty of people were making *some* money from such things way back even before email. But this year has taken it to a new level. People are signing six-figure deals with Substack and giving up ordinary journalism gigs and book deals to do it.

Evans points out that in newsletters, as in previous Internet phenomena - podcasts, web pages (hence search engines, and ecommerce (hence aggregation) - the first people who show up in an empty space with good stuff people want do really well. We don't hear so much any more about first-mover advantage, but it often still applies.

Non-fungible tokens (NFTs) may be the latest example. A few very big paydays are drawing all sorts of people into the field. Some will profit, but many more will not. Meanwhile, scams and copyright and other issues are proliferating. Even if regulation eventually makes participation safer, the problem will remain: people have limited resources to spend on such things, and the field will be increasingly crowded.

So, too, Substacks and newsletters: there are not only limits to how many subscriptions people can afford, but also to how many things they have time to read. In a crowded field, discovery is everything.

Individuals' attention spans and financial resources do not scale. The latter is one reason the pay-with-data model has been so successful on the web; the former is part of why people will sacrifice privacy and participatory governance in favor of convenience.

So, our partial list of things that do not scale: content moderation, community, discovery, governance. Maybe also security to some extent. In general: anything that requires human labor to be added proportionately to its expansion. Incorporating solving problems of scale will matter if we're going to have a different outcome from web3 than from previous iterations.

Illustrations: A hockey stick.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Stories about the border wars between cyberspace and real life are posted occasionally during the week at the net.wars Pinboard - or follow on Twitter.


TrackBack URL for this entry:

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)