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.

2024/01/08: Diversifying your SaaS-style LLMs

Back in July, I noted that companies building on SaaS-style AI models were exposed to certain kinds of risk. I framed this as: What if OpenAI experiences large changes, so you lose access to ChatGPT and the like? (Original post here, which was reshared to the Gray Rhino blog a couple months later.)

I've since highlighted this risk to clients and developed mitigation plans.

A few months have passed and now other companies are having a think:

"OpenAI Turmoil Pushes Customers to Diversify" (WSJ)

Executives at companies that use OpenAI’s software say they are increasingly looking to also use others’ technology to protect themselves from the risks of problems at any one. OpenAI’s competitors are using the opportunity to sign up wary customers.

(While I'd love to take credit for these companies making the move, they aren't clients of mine. :-)

What's important is that the companies listed in that article aren't so much giving up on their generative AI plans, as they are looking for alternatives to OpenAI.

And that is what "risk management" is all about. It's not about saying no to everything. It's not about shutting yourself out from every opportunity. It's about identifying and addressing potential problems in advance, before they bite you, so you can stay on-track to meet your goals.

If you'd like an assessment of your company's AI-based risks, along with a list of steps for mitigation, please reach out: https://riskmanagementforai.com/

2024/01/08: Being data-driven

For every company that claims they want to be "data-driven":

I'm sure you've heard about "looking at the map versus experiencing the territory." When the two disagree, you don't want to insist that the map is correct. Yet, that's what companies do with data.

Here, Brandon Arvanaghi shares a lesson from Amazon.

(Website note: In case that LinkedIn URL no longer works, it references this excerpt from an interview: "Jeff Bezos called Amazon customer service | Lex Fridman Podcast Clips")

After you watch this clip, why not take a hard look at your metrics (map) and see how well they compare to reality (territory)? Or would you rather wait for the CEO to do it for you?

2024/01/09: When new tech meets old norms and laws

I've noted before that every emerging technology eventually collides with existing norms and rules. So while the technology itself gets all the attention in the short run, matters like law, policy, and insurance take over in the long run.

Today's example: generative AI used to clone people. Sometimes with their consent. Sometimes without. The latter falls into a legal gray area.

"A New Kind of AI Copy Can Fully Replicate Famous People. The Law Is Powerless." (Politico)

2024/01/10: Grabbing popcorn …

Sharing this here to see how the quants, analysts, and AI practitioners react.

[Grabs popcorn.]

Jokes aside:

A number of you data professionals will read this and nod approvingly. Maybe even give a wry smile as you reflect on the times these two tools have won the day.

For those who don't … especially if you see spreadsheets as inferior to Python … I'd encourage you to have another think.

Remember to focus on your end-goal (usually: "analyze / present / share data in a way that is useful for the business") and remain open to learning tools that will help you get there.

(Original tweet: https://twitter.com/netcapgirl/status/1744445848778424560 )

2024/01/11: More thoughts on tooling

Reflecting on yesterday's post about data tools (and tool loyalties in general) I'm reminded of some old product wisdom:

_People don't want to buy a hammer. _

They don't even want to drive a nail into the wall.

They want to hang a picture.

If your stakeholders want the data equivalent of hanging a picture, then … help them hang the damned picture. That's the job.

• Some days you'll need a hammer and a nail (for that heavy wooden frame)
• Other days you'll reach for adhesive (because the hammer would destroy the wall)
• Once in a while, you'll mount a monitor and use kiosk mode (so someone can update the image in real-time)

I'm stretching the idea, but you see what I mean.

I've had days where I've used a spreadsheet, Pandas, and Tensorflow – all for the same client and on the same project. It's a reality.

Focus on the goal.

2024/01/12: When simple data analysis tells you a lot

I'm catching up on some Odd Lots episodes. This one on salad chain Sweetgreen (2023/12/14) was particularly interesting:

"How Salad Chain Sweetgreen Figures Out Its Next Product to Sell" (Odd Lots / Bloomberg)

One point stood out in particular: the use of data.

I mean simple, BI-style, analytics data.

The hosts asked Sweetgreen co-founder Nicolas Jammet about prices, trends, seasonality, and so on. The kinds of questions that are 1/ very important to any business and also 2/ require zero AI.

To be clear: AI is useful! BI is also useful! Neither one is necessarily better than the other. But I've found that BI tends to get overlooked because AI has so much of the spotlight. Which is unfortunate when you consider the value BI brings to the table.

(Doubly so, when you realize that good BI is a key precursor to a company's AI efforts. I've written more about that here and here.)