Hard Fork Podcast | New York Times
We've been enjoying the New York Times's Hard Fork podcast lately. It's a fun and conversational overview of recent tech news and trends and what they mean for society. This latest episode covered the recent snafu over Zoom's terms of service and a growing backlash against data scrapping.
The use of robotics and other technology in policing brings with it concerns about privacy and bias. Further troubling is the finding that much of this technology is purchased with federal and state forfeiture funds, leading to concerns about a "vicious cycle where police purchase more surveillance software to seize more assets to fund even more surveillance."
Even with federal legislation still in nascent stages, it's worth keeping an eye on AI uptake in the private sector. McKinsey's global survey on the state of AI finds that corporate uptake of AI has focused on product/service development and service operations (customer support, back office functions). Concerning risk, businesses were primarily concerned about AI's risk of inaccuracy (followed by cybersecurity and infringement) but only 21% of respondents said "their organizations have established policies governing employees’ use of gen AI technologies."
We've been doing some work around how LLM training data feeds into biased outputs (more to come!) so this article and the underlying study about the political biases of various LLM really resonated. Social media already serves to generate echo chambers, but this raises the question as to whether LLMs with continue that trend.
Relatedly, the Washington Post took a look at LLMs' pro-eating disorder outputs, illustrating the complexity and diversity of harms that can result.