Pongo Points 2/15/24

US Commerce Secretary installs AI leadership | Coinbase's latest state of the industry report | AI disinformation is a good thing? | Microsoft AI in journalism | How one file rules the internet

1. Executive Leadership of US AI Safety Institute Announced

Why it’s interesting: Pursuant to President Biden’s Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, the US Secretary of Commerce has announced Elizabeth Kelly as Director, a Special Assistant to the President, and Elham Tabassi as CTO, an associate director at NIST.

What stands out: Within one week of 20 state attorneys general contesting Biden’s Executive Order, the US Commerce Secretary seeks to cement the administration’s role as the de-facto regulator of AI by inserting politically aligned leadership into its dubiously created initiative.

What’s next: AI will likely become a politically divisive technology, as it has already proven to have the potential to create massive amounts of disinformation and can be easily abused for political or economic gain.

2. Coinbase’s Latest “State of Crypto” Report Paints Rosy Picture

Why it’s interesting: Based on research from Morning Consult, Bovitz, and The Block, Coinbase posits that American consumers could have saved over $74 billion in credit card transaction fees and merchants over $126 billion had cryptocurrencies been used instead.

What stands out: Coinbase is promoting the ideal of cryptocurrencies as mediums of exchange, whereas they’re mostly used as speculative tools today, but neglects to consolidate the fact that there is very low adoption for cryptocurrency payment rails today.

What’s next: Coinbase is making valiant efforts to establish cryptocurrencies as viable alternatives to the traditional finance tools we use today, however the crypto industry itself if struggling to find an enduring product-market fit.

3. Is AI Disinformation Overblown?

Why it’s interesting: Finance and technology writer Byrne Hobart proposes an interesting take on why AI-drive disinformation campaigns might not be as big a deal during election cycles due to “trial balloons” and partisanship.

What stands out: A trial balloon is an intentional information leak meant to gauge public reaction to a potential announcement, and Hobart argues that AI-deepfake productions might end up being used by politicians to determine the popularity of a given policy choice - rather than simply smearing opponents.

What’s next: Hobart’s argument is as counter-intuitive as it is fascinating, especially alongside the claim that the AI “deepfake effect” might actually enable more efficient information markets.

4. Microsoft Partnering with News Companies on AI

Why it’s interesting: Microsoft announced a number of partnerships with various media outlets to collaborate on new generative AI products that help journalists research, translate, and more.

What stands out: One of the outlets, Semafor, acknowledged some financial benefit received from Microsoft to use its AI tools, which will be used to create an international newsfeed of “the best of the world’s reporting.”

What’s next: Microsoft’s stated aim is to help those outlets grow audiences and streamline time-consuming tasks, but it also forces one to wonder “how far” do these AI tools go - do they write? Do they have political bias? What data are they trained on?

5. The Social Contract of the Internet

Read it on The Verge here: The text file that runs the internet

Why it’s interesting: The robots.txt file is a universally accepted way of allowing or disallowing internet crawlers like Google, Bing, etc. from viewing your website and serving it to their users, but now AI companies are creating challenges to this 30 year old system.

What stands out: Because of the rift between creators and AI companies using their content without permission to train new models, the robots.txt system is under scrutiny as individuals, companies, and regulators debate on a path forward that’s fair to all parties.

What’s next: Website operators have a choice between allowing AI companies to use their data or not; simultaneously, AI companies have a choice in whether or not to acknowledge those preferences when sourcing data for their models.

Meme Digestif

Thank you for reading! This is likely to be the last Pongo Points curated newsletter for the foreseeable future. I appreciate you being along for the ride!

🦧