Variant Fund: How the Crypto World Shapes More Sophisticated AI Models?
TechFlow Selected TechFlow Selected
Variant Fund: How the Crypto World Shapes More Sophisticated AI Models?
Models with taste have a large and growing potential market.
Navigating Web3 tides with focused insightsWritten by: Alana Levin
Translated by: TechFlow
Over the past two years, new AI models have emerged at a rapid pace. These models are capable of performing many types of tasks—from retrieving information and answering questions to providing customer support, proofreading documents, generating content, and more.
Many such tasks are objective, with clear optimization functions: finding the correct answer, identifying the most relevant information, detecting errors or anomalies, and so on.
But there are also models whose outputs are highly subjective—such as creating “good” art or developing “interesting” videos. I refer to these as “models with taste.” Taste-based models are often harder to optimize because they represent a blend of collective and individual judgment; there is no single correct answer or output. Frequent feedback is therefore especially valuable in helping these models understand evolving cultural preferences.
Today, there are broadly two ways to cultivate a model’s “taste”:
-
Using user-generated content/data (such as feeds from Twitter or Reddit), which in theory can reveal what humans currently care about (thus serving as a proxy for taste).
-
Leveraging a community of human “tastemakers” to actively train the model around their preferences.
The first approach comes with several undesirable issues. Data may be siloed (e.g., Reddit shutting down its API) or biased (e.g., only partial data being shared). Models might also overfit to the algorithms of specific platforms, especially if their data sources are limited. This might not seem important until one imagines vast amounts of new media generated based solely on trending Twitter content. That would not be ideal.
The latter approach—a network of humans providing feedback—avoids many of these risks. There may still be bias, but only insofar as it reflects the preferences of those who choose to participate in training the model. The key, then, is ensuring that these community members—the ones shaping “taste”—have a genuine stake in the model developing good taste.
Crypto can help align these incentives. Granting ownership in the model—or offering economic rewards tied to model outputs—can motivate participants to engage meaningfully. Cryptocurrency also makes participation more open and accessible: anyone, anywhere in the world, can contribute as long as they have a wallet and internet access.
A notable example is the Botto project. Botto is an autonomous artist, and $BOTTO token holders have the ability to help train the model each week. The training process is simple: participants vote for or against various images, and Botto learns from the community’s preferences. At the end of each week, the most popular artwork is auctioned, and participants who helped train Botto that week are rewarded.
Art is just one category of taste-based models. Others could include film, television, other forms of storytelling (novels, short stories), comedy, and advertising/brand campaigns. Just a few years ago, such taste-driven models would have been impossible. The tools were less expressive, slower, and couldn’t reliably produce coherent or (in the case of video) realistic outputs. Only today has this become feasible.
Crucially, taste-based models have a large—and growing—potential market. The art market alone is worth billions of dollars. Online content consumption captures trillions of hours of attention annually. If people are already spending time and money on these forms of entertainment, it seems reasonable to give them a stake in the creation process too. This wouldn’t just create a more engaged user base—it would create a more satisfied one. Imagine audiences who helped train and shape storylines being central figures at the Oscars for Best Picture, or an entirely new award category for films created by communities—that would be truly exciting.
I see this not as replacing existing creative work, but as creating a new category of content. It’s similar to how smartphones and Instagram turned everyone into photographers—these new tools didn’t eliminate professional photography; if anything, they may have increased appreciation for it. Taste-based models are the same: they use new technology—in this case, crypto rails for consumer ownership and economic alignment—to create a new form of participation, thereby expanding each of the categories mentioned above.
In the past few years, we’ve seen thousands of new models emerge. In the coming years, we may see millions—or even more. At least some of these should aim to involve stakeholders in new ways, from greater openness and accessibility to novel ownership structures that experiment with incentives. Taste-based models are a particularly fitting area for such innovation—but they certainly won’t be the last.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
Variant
