Seeing Like an Algorithm — Remains of the Day: What Happened to Yahoo

Seeing Like an Algorithm — Remains of the Day: What Happened to Yahoo

In a world where machine learning algorithms are becoming increasingly prevalent, understanding how these algorithms achieve their accuracy is crucial. TikTok, a popular short video platform, has gained attention for its algorithm's ability to recommend personalized content to users. However, experts in the field believe that TikTok's success lies not in a groundbreaking algorithm, but rather in its closed loop of feedback. This feedback loop enables the algorithm to continuously train itself on user-generated videos, improving its recommendations over time.

When designing an app, it is important to consider how best to help an algorithm "see." This concept of algorithm-friendly design involves every action and interaction within the app being a signal to the algorithm. For example, before a video is even sent to a user's phone, it is watched and labeled by a human on TikTok's operations team. Additionally, TikTok's camera filters are designed to track human faces, hands, and gestures, utilizing vision AI from the point of creation.

The default user interface of many social networks, including Facebook, Twitter, and Instagram, is the infinite vertically scrolling feed. While this design allows for easy scanning of content, it also poses challenges for algorithms in accurately judging user sentiment. By primarily relying on positive engagement mechanisms, these platforms may miss out on negative signals from users. Furthermore, content derived from a user's social graph may not align with their true interests, leading to a divergence between personalized recommendations and user preferences.

Algorithm-friendly design does not have to be user-hostile. Instead, it requires a different approach to serving the user's interests. The goal of any design should be to help the user achieve their desired outcome, which may sometimes involve introducing friction or tradeoffs. In the software era, true competitive advantages are increasingly illusory, as most features and designs can be easily copied. The real magic lies in creating a cohesive ecosystem where every element and process aligns with a single purpose and goal, as exemplified by TikTok's success.

On the other hand, the downfall of Yahoo can be attributed to two key factors: easy money and ambivalence about being a technology company. Unlike Google, Yahoo did not fully capitalize on the value of its traffic. Advertisers were already overpaying for Yahoo's services, and if the company were to extract the actual value, their revenue would have decreased. This lack of focus on maximizing revenue led to Yahoo becoming the beneficiary of a de facto Ponzi scheme in the late 1990s.

During the dot-com boom, investors were eager to invest in Internet startups, and these startups, in turn, used the money to buy ads on Yahoo to generate traffic. This cycle of investment and revenue growth further fueled the excitement around the Internet as an investment opportunity. However, Yahoo's emphasis on brand advertising and lack of targeting capabilities prevented them from taking search seriously, unlike Google. While Yahoo portrayed itself as a media company, it failed to recognize the importance of being a technology company.

One of Yahoo's fatal mistakes was treating programming as a commodity. User-facing software was primarily controlled by product managers and designers, with programmers serving as mere translators of their work into code. This approach led to the hiring of subpar programmers, and as the quality of programmers declined, Yahoo entered a death spiral from which it could not recover. In contrast, Google maintained a hacker-centric culture that attracted top talent and propelled their success.

In conclusion, the success of algorithms like TikTok's and the downfall of companies like Yahoo highlight the importance of understanding and incorporating algorithms in design and decision-making processes. To create algorithm-friendly designs, every action and interaction should be considered as a signal to the algorithm. Additionally, reducing friction is not always the goal; instead, the focus should be on helping users achieve their desired outcomes. To avoid pitfalls like Yahoo, companies must align all elements and processes with a single purpose and goal, while also fostering a hacker-centric culture that attracts top talent.

Actionable advice:

1. Understand the significance of algorithm-friendly design and consider how every interaction within your app or platform can serve the algorithm's needs.
2. Prioritize user interests over minimizing friction, as reducing friction may not always lead to the best outcomes for users.
3. Foster a hacker-centric culture to attract and retain top talent, as the quality of programmers plays a crucial role in a company's success in the technology industry.