AI is a Hail Mary pass and other links & notes

6 February 2023 – Baldur Bjarnason

Generative AI is the tech industry’s Hail Mary pass #

Friday’s newsletter wrote about how the tech industry’s desperate need to keep its valuations up is leading it to go all in on generative AI.

Over the past few decades, tech companies have been priced based on their unprecedented massive year-on-year growth that has kept relatively steady through crises and bubble pops. As the thinking goes, if you have two companies—one tech, one not—with the same earnings, the tech company should have a higher value because its earnings are likely to grow faster than the not-tech company. In a regular year, the growth has been much faster.

This has been great for the tech industry and for investors. Even shitty, poorly-run tech companies benefit because people assume tech will grow enough to make up for poor management. If you can demonstrate growth, investors will ignore that you’re pissing it away with reckless abandon. Tech companies—generally—have a baked-in assumption of future profitability that most other sectors don’t have.

The problem is that things that can’t go on forever don’t, and over the past year some investors have changed their tone on the tech industry. Gone is the assumption that endless growth will make up for the inefficiencies and incompetence.

“Stable Attribution” #

This is an interesting experiment in trying to find the original images that ‘inspired’ the AI output. (Wonder if GitHub’s experiments in finding attributions for verbatim code works similarly?)

The discussion on this over on Hacker News is, unusually enough, informative, pointing out quite a few limitations in the approach taken.

“The Market for Lemons - Infrequently Noted” #

Always post links that reference Akerlof.

I am not a fan of the Single-Page-App approach to web development and the discourse surrounding them in web development has always been exceedingly unpleasant. But it looks like the tide is turning, finally.

“OpenAI’s Whisper is another case study in Colonisation” #

For our organisation, the way in which Whisper was created goes against everything we stand for. It’s an unethical approach to data extraction and it disregards the harm that can be done by open sourcing multilingual models like these.

“Getty Images v. Stability AI – Complaint - Copyright Lately” #

This is in the US, in addition to Getty Images’ suit in the UK that was filed earlier.

“4chan users embrace AI voice clone tool to generate celebrity hatespeech” #

AFAICT, this is exactly the sort of thing that the EU’s proposed AI regulations would stomp on, hard. With GDPR-style fines.

“You wise up. – Hi, I’m Heather Burns” #

Much chatter amongst the chatterati, then, with yesterday’s addition to the saga that is the Online Safety Bill.

“Introducing the Slickest Con Artist of All Time” #

ChatGPT and its ilk are the con artist. If you couldn’t guess from context.

“SOLVED! The Case of the Bing Ban Theory - daverupert.com” #

Bing was marking a bunch of normal blogs as spam. Because of course it was. These are the geniuses the industry thinks will ‘disrupt’ Google with a ChatGPT-based search engine.

“Against risk-based authentication (or, why I wouldn’t trust Google Cloud)” #

Add to that the risk of having auth linked to social media behaviour (Youtube). Get banned from Youtube; get locked out of Google Cloud. Tying work credentials to that of a social media network is fundamentally a bad idea and Google needs to do more work to separate Youtube from Google Cloud and the Google productivity apps.