Social media has turned decidedly uglier amidst the Israel/Hamas conflict, researchers raise new AI safety and transparency concerns, and former FCC chair Tom Wheeler has a new book out on regulation in the "digital gilded age."
As compared to the complexity of election related content moderation, terrorism has historically seemed a relatively straightforward issue for social media platforms. But the current conflict between Israel and Hamas is calling all that into question. Meta has taken the approach of limiting comments on posts from regional users (without defining the geographic parameters), users have accused Meta of censoring pro-Palestinian views (shadow banning), and Hamas' Telegram channel has experienced a surge in users. Platforms (including Meta, TikTok - notably excluding X) have released statements concerning their approach to the conflict, but disinformation continues to proliferate. With 50% of Americans reporting that they often or regularly get news from social media and vanishingly few media outlets trusted by Americans on both side of the aisle, it's hard to imagine turning back the tide in the disinformation war.
The War in Gaza Is Getting Remixed in Real Time | The Atlantic
This paper from Princeton, UVA and IBM researchers raises concern that OpenAI's paid feature that allows users/businesses to fine tune the model for specific downstream use cases effectively allows them to remove digital guardrails. The conclude that "fine-tuning aligned LLMs introduces new safety risks that current safety infrastructures fall short of addressing."
Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! | arXiv
Stanford just released an index of the 10 largest foundation models, ranking each on its transparency. The results illuminate trends in declining transparency in AI models, with troubling implications for regulation and research.
The Foundation Model Transparency Index | Standford
We're looking forward to reading Former FCC chair Tom Wheeler's new book “Techlash: Who Makes the Rules in the Digital Gilded Age?” -- featured in this Vox write up. Characterizing digital assets as soft, shareable and effectively inexhaustible, he urges a regulatory approach that mirrors the agile digital management style of BigTech.
We’re in a new Gilded Age. What did we learn from the last one? | Vox
CSIS' Benjamin Jensen testifies about the scope of Chinese cyber espionage campaign and what we can expect to see as they target AI developers and deploy AI powered tactics for IP theft.
How the Chinese Communist Party Uses Cyber Espionage to Undermine the American Economy | CSIS
Lawfare interviews Dean Jackson on the state of election disinformation, highlighting the findings of a recent Center for Democracy and Technology report.
What are the challenges facing efforts to prevent the spread of election disinformation? | Lawfare