Weekly recap: 2023-04-02
2023-04-02 | tags: weekly recap

What you see here is the last week's worth of links and quips I have shared on LinkedIn, from Monday through Sunday.

For now I'll post the notes as they appeared on LinkedIn, including hashtags and sentence fragments. Over time I might expand on these thoughts as they land here on my blog.


2023/03/27: Humans in the loop

Key point in AI safety: A law firm has deployed an internal chatbot for its attorneys. But it still knows to keep a human in the loop:

"Tech summit: Allen & Overy turns to the magic of AI for savings" (The Times UK)

(Remember my main AI safety rule: "never let the machines run unattended.")

2023/03/27: It's all about the basics

According to this newsletter (and the embedded tweet it references), OpenAI – the company behind ChatGPT – is hiring a "Killswitch Engineer." Someone to unplug the servers in case the AI bot runs amok.

Laugh if you will, but this meets my first rule of AI safety: "never let the machines run unattended."

(It's also a hell of a marketing ploy, if OpenAI indeed posted this: why fight the criticism and fears around AI chatbots, when you can lean all the way in?)

2023/03/28: AI moves into the real world

Per this article, "traditional legacy industries — such as manufacturing, transportation, logistics and healthcare" are poised to "be deeply transformed by technology." Including AI.

"A new technology boom is at hand" (FT)

That's all well and good, but there's a catch: once you're using AI in the physical world – managing factory systems, say – it has to actually work. There's no room for hand-waving and papering over poor results. A real-world incident of model failure can have real-world consequences.

Here's to hoping that every group that implements AI in these as-of-yet-untapped fields first implements a strong AI risk management function. They'll need it.

2023/03/30: Apple an BNPL

Apple strolls into the buy-now-pay-later (BNPL) space with "Apple Pay Later":

"Apple Starts to Roll Out ‘Pay Later’ Service After Delay" (Bloomberg)

Key point:

In a new approach for the company, Apple will be funding the loans via a new subsidiary called Apple Financing LLC. The company will also be handling its own credit checks and running the other underlying technologies needed to power the service [...].

Apple getting into the credit #risk game? Interesting. They've long been in the business of selling pricey devices… It's quite possible they have a good handle on how to gauge a person's creditworthiness.

2023/03/30: AI is not an add-on

New blog post: why it's a mistake to just treat AI as a quick add-on to a company.

(If you want the most out of AI, you need to be strategic about how you employ it.)

"AI isn't something you just add to a company"

2023/03/31: Truth in an age of generative AI

Interesting article on the synthetic (AI-generated) image of the Pope in a winter coat:

"Künstliche Intelligenz: Fake-Franziskus geht viral" (Der Spiegel)

There's a key point in there, on the widely-available access to generative AI. Quoting someone who had commented on the image:

»Das Zeitalter der Wahrheit ist vorbei.«

("The era of truth is over.")

Now, we can debate whether the era of truth ever started, but that's a story for another day …

We talk about how the pandemic changed our appetite for in-person interaction.

In a world where we can't believe what we see on-screen, face-to-face contact will become even more valuable.

Anyway, can't wait to see all of the synthetic April Fool's images tomorrow…

(Though I see that MidJourney has suspended free trials … so there's at least some hope.)

AI isn't something you just add to a company

If you want the most out of AI, you need to be strategic about how you employ it.

Weekly recap: 2023-04-09

random thoughts and articles from the past week