Digital Platform Policy Highlights - Digest 10
Yes, you guessed it - most of the highlights in this digest are ALSO on LLMs, but that’s not all. I’m excited to share that this is the 10th digest I’ve written for you.
This post is part one of a series documenting policy changes and feature improvements introduced by platforms in April 2023.
TL;DR → Policy Changes to Attract New Users include:
Policy Changes to retain existing users on the platform include:
Policy changes to extract surplus from a subgroup of users who value some specific features include:
Reddit changing the monetization of its API for enterprise users
Twitter experimenting the pay-per-article model with publishers
AWS releasing multiple foundational models for its enterprise consumers
In April 2023, there was one policy change to attract new users to the platform.
NPR, Twitter and the “State Affiliated Media” tag
NPR, the U.S. public broadcaster, has stopped posting on Twitter after the social media platform labeled it as “state-affiliated media”. The label, which Twitter uses for official propaganda outlets in countries like Russia and China, undermines NPR’s reputation as a source of independent journalism. After NPR’s exit, Twitter has removed these labels, which meant event outlets such as RT and Xinhau are no longer tagged “state-affiliated”. While the original move appeared to attract conservative consumers to the platform, it inadvertently equated public broadcasters with state-controlled media. The removal of all tags in response to NPR’s exit may make it more difficult for users to assess the credibility of all organizations (link 1 and link 2)
In April 2023, there were a few policy changes that platforms introduced to “retain” existing consumers from moving to an alternative platform.
Amazon Makes AI Coding Assistant Free for Developers
Amazon has made its AI coding assistant, CodeWhisperer, free for individual developers. CodeWhisperer, which helps write code based on text input and scans for security vulnerabilities, is part of Amazon’s strategy to compete with Microsoft’s GitHub Copilot AI tool, which costs $10 per month. This technology is based on Large Language Models, a class of generative AI system (The most popular example of a product which uses Large Language Models is ChatGPT). Tech companies overall are rushing to incorporate Generative AI features into their products, hoping the transformative technology will attract new users and better monetize existing ones (link)
Alibaba launches Tongyi Qianwen, an AI chatbot for cloud consumers
The name, which roughly translates to “seeking truth by asking a thousand questions” is Alibaba’s Unified Q&A chatbot for Alibaba’s cloud services. Trained on multiple domains and languages, it can generate coherent responses in both Chinese and English. Right now, it is intended to be used with their current products, such as Tmall Genie smart speakers and DingTalk workplace messaging platform. However, Alibaba plans to offer it as a customizable LLM service, making it domain and firm agnostic. Given that this is released within 3 weeks of Baidu’s solution, it would be interesting to study whether the competition dynamics change in China because of LLM superiority (link)
SenseNova Chatbot from SenseTime, a Chinese AI company
SenseNova can create text, images, or video based on user prompts. However, SenseTime intends to provide service to other businesses, including e-commerce services and videogame developers. SenseTime claims that SenseNova has better performance and security than ChatGPT, an important selling point as organizations worry ChatGPT could incorporate sensitive data from prompts into its database. With SenseNova, SenseTime joins other Chinese tech companies like Baidu and Alibaba in launching their own chatbots in response to ChatGPT’s success. It will be interesting to see how SenseTime navigates the dual challenges: China’s heavy-handed regulation of generative AI and the U.S. restrictions on chips crucial for powering such operations. (link)
Twitter blocks and then unblocks Substack a day later
Twitter blocked Substack, the newsletter platform, after it launched a Twitter-like feature called Notes. Twitter marked Substack links as potentially unsafe and limited the distribution of Substack links. The move promptly received backlash from authors who rely on Twitter as a major source of traffic and audience engagement. Even SubStack leadership and founding team criticized this move, calling Substack as a complement, not a competitor to Twitter. Twitter reversed this policy a day later. Understandably, Twitter seeks to limit competing platforms from tapping onto its user base. However, the boundaries of competition are not clear-cut in the networked economy, putting Twitter under frequent scrutiny, given its stated commitment to transparency (link)
In April 2023, there were a few policy changes that platforms introduced to extract surplus from a subgroup of users who value some specific features.
Twitter’s blue checkmark program becomes a paid verification program
Forever, Twitter granted a blue checkmark to verify accounts of public interest. Twitter has changed this blue checkmark into a paid verification system in April 2023. However, the revocation has been inconsistent: some accounts retained their checkmarks without verifying. So, consumers had no way to distinguish legacy verified users with accounts that pay for Twitter Blue. Furthermore, despite efforts and ban warnings, many fake verified accounts, including that of Nintendo of America or LeBron James, have crept up, undermining the veracity of the verification process. Charging for aspirational “signals” without a careful rollout presents a trade-off: short-term revenue vs. long-term trust (link)
Reddit Changes API Terms to Monetize Data for AI
Reddit has announced new API changes that will affect developers who use its data for commercial purposes, such as training artificial intelligence models. The changes will require developers to enter into a separate agreement with Reddit and pay a fee for accessing its data. As Reddit eyes an IPO, opening a new revenue stream would help the platform reduce its dependence on advertising. However, this monetization policy faces accusations of being hypocritical, since the entire data on Reddit are user generated over many years. This may affect the platform’s long run since some users and moderators have called to boycott Reddit (link)
Twitter’s move to enable Pay-per-article option for media publishers
Announced in April but not yet implemented, the new feature will allow publishers to monetize their content without requiring users to subscribe to their platforms but to just use Twitter’s payment system. As twitter aims to become the “everything app”, a matchmaker between service seekers and providers, it hopes that consumers will use Twitter to manage payments, presenting a win-win proposition for both the platform and publishers. However, many media organizations have already been antagonized by the platform’s past policy changes. This may explain why the feature that was planned for May has not yet been implemented (link)
Amazon enters Generative AI race with AWS Bedrock
To provide AWS customers with more choice and flexibility to use the best foundation models for their specific needs, Amazon has announced a new service called Bedrock. Developers can access different types of customized generative AI models that can produce text based on user prompts specific to their needs. Unlike Microsoft and Google, who have launched consumer-facing chatbots and software products with generative AI, Amazon bets on the B2B enterprise market and its ability to select the best model for each use case. This gives these enterprise customers a flexibility and choice to build AI solutions for their own customers. This policy could be a differentiator in a highly competitive field (link)
Research help from Anantesh Mohapatra and Yiran Liu (Thanks a ton, folks!)