Weekly recap: 2023-04-09

Posted by Q McCallum on 2023-04-09

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.

2023/04/03: Surviving April Fool’s Day in a generative AI world

So … the world didn’t fall into a hellhole of AI-generated April Fool’s images over the weekend.

Or maybe I just missed it all?

(I’m fine either way.)

2023/04/03: The zebra’s birthday

The humble barcode turns 50!

As it turns 50, how the barcode changed our lives” (The Times UK)

It’s hard to overstate the power of a standardized, printable encoding scheme that is designed for machine reading. (It’s sort of a reverse data viz, in a way.)

From grocery shopping to industrial applications, the barcode has served as a key bridge between the digital and physical. In turn, by making it easier and more reliable to get that information into systems, barcodes have made a lot of data analysis and AI possible.

QR codes and “digital twin” NFTs are all the rage now – understandably so, as they can pack in a lot more information than a barcode’s series of digits – but let’s not forget where it all started.

2023/04/04: Executive data literacy is a long-term investment

I rant – um, “talk” – about executive data literacy a lot. It’s a key element to a company achieving success with ML/AI. That said, I readily acknowledge one hurdle: the lack of instant gratification.

Data “literacy” is an apt term because it’s similar to learning how to read: doing so doesn’t mean you’ve already absorbed all of the knowledge that’s been written down. It means that you’re now able to absorb all of the written knowledge that you can get your hands on.

Similarly, data literacy is a skill that helps you understand when AI can (and cannot) help your business; that shows you what new risks you take on when using an AI-driven solution; that prepares you to evaluate the deluge of AI vendor pitches.

This is an investment with very strong returns. But those returns play out over the long run.

(It’s not as exciting as the vendor pitches that you – yes, you! – can be using AI, right now … but those sales pitches rarely live up to their claims anyway. It’s short-term excitement that fizzles out)

2023/04/05: This job is all about asking questions

What’s the most important attribute of a data analyst/data scientist/machine learning engineer?

Some might say that it’s “intelligence,” “technical know-how,” or “understanding of the business model.” Maybe even plain old “curiosity.”

Those are all very important, yes. But they take a back seat to:

A healthy sense of doubt.

Especially self-doubt.

This kind of job requires that you constantly ask yourself:

  • “How did this training data come to be?”
  • “What else could be the reason for this effect that I see? Did I miss something?”
  • “Are there other contributing factors for me to consider?”
  • “This model seems to be performing a little too well … did I introduce a feature leak here?”

If you accept your model’s performance without question, or if you go with the initial results of your analyses, it eventually will come back to bite you.

2023/04/06: A lesson from the robotics department

With all of the AI hype, I guess people have forgotten about robotics?

AI Is Running Circles Around Robotics” (The Atlantic)

I get it: between “ChatGPT,” “Midjourney,” and “people trying to forget about crypto,” AI is getting a lot of attention and investment these days. Sure.

But I see another reason why AI is getting more press than robotics: it’s the hand-wavey factor.

It’s easier to hand-wave over an AI model’s results and pretend that things are working when they aren’t.

A robot performing a task in the physical world doesn’t have that luxury. Questions like “Did it climb the stairs? Did it open the door?” don’t leave a lot of wiggle room.

Perhaps the field of robotics is more grounded in reality as a result. That leads to fewer flashy news stories.

(There’s a subtle AI safety lesson here: when your product has real-world impact, when there is possibility of tangible harm, it has to actually work. That usually means slowing down, adding more guardrails, and being honest when things aren’t performing as expected.)

2023/04/07: Are we really surprised?

Hmm. So when you sell marks of legitimacy, dissolve rules, and neutralize trust/safety efforts … things may get out of hand. Who knew?

Comment Twitter Blue est devenu un vecteur de désinformation” (Les Echos)

2023/04/07: A very organic barcode

Adding the biological equivalent of a barcode to verify movements through a supply chain:

How A.I. and DNA Are Unlocking the Mysteries of Global Supply Chains” (New York Times)