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.
This is a scathing, yet not exactly unfair, take on how companies use AI:
“Would you ask Facebook AI Snoop Dogg for frozen yogurt suggestions?” (Washington Post)
Specifically, it’s an indictment of companies’ claims about what their AI can do compared to the actual results thereof.
I get it: part of the sales game is to get people excited about a future state. Sure. But as I’ve noted before, you get into trouble when you sell a possible, distant-future state as a certain, present-day state. That’s what’s happening with a lot of AI these days.
And when things don’t pan out, there’s only so much hand-waving you can really do.
Many of these AI technologies are experiments. And it’s fine to be guinea pigs for unproven technology — as long as the stakes are low and we get some benefit from imperfect technology.
But a company has to properly set your expectations.
If Google says its AI can summarize email, it shouldn’t make up fictional messages. If Zoom promises AI can make us more productive, it shouldn’t be a creepy eavesdropper that drains your attention with information flotsam.
Interesting FT piece on remote work:
Anyway, homeworking should keep getting more productive. It emerged at the worst possible moment: unplanned, during lockdowns, when many workers had children at home. To borrow a metaphor from Dutch writer Joris Luyendijk, the day the Wright brothers took their first flight in 1903, they couldn’t yet have designed an aviation industry. Solutions emerge over
A mere three years into the mass homeworking experiment, inefficiencies remain. For instance, remote workers still waste endless hours managing their managers. Bloom reports that US hybrid workers with degrees spend half their day in meetings, twice as long as office workers, probably to please bosses who worry they are slacking. No wonder homeworkers struggle to switch off and suffer burnout. But then so do office workers.
People typically think of a supply chain in terms of its mechanics: delivery of goods. “Move Item A from Point B to Point C within Time D.”
It’s also helpful to think of a supply chain in terms of what it makes possible: keeping promises.
Specifically, a well-run supply chain lets you keep promises to sell items that you don’t have on the shelf right now. You tell a buyer they can pick up the item on Friday because you know the delivery truck is due at your shop on Thursday.
Supply chain disruptions, then, lead to broken promises.
This is where risk management, simulations, and predictions come into play:
- Risk management: By taking the time to explore what could go wrong, you can lay out plans for mitigation.
- Simulations: Running simulations – large-scale “what-if?” scenarios – of your supply chain activity can spot additional problems and demonstrate their impact. This feeds back into your risk management plans.
- Predictions: You can predict when certain problems might occur, and prepare. Maybe you stock extra goods because your forecasts call for a surge in consumer demand.
How is your business using risk management, simulation, and predictions to improve its operations?
I’ve been thinking about this as of late as I read about difficulties in this latest vaccine rollout. People schedule vaccine appointments, which then get canceled because the pharmacy doesn’t receive its shipment in time. All of which feeds back into the system as those people have to schedule new appointments … and so on.
(This article is two weeks old. But the problems it describes are still in play.)
Several of the nation’s largest pharmacy chains acknowledged a challenging rollout. CVS, Walgreens and Safeway pharmacies have had to cancel and reschedule some appointments because of delayed shipments of the vaccines, the companies said. Other providers delayed making the shots available. Rite Aid said new vaccines would be in stores by this weekend; Kaiser Permanente said it would largely not administer Covid shots until next week.
We often talk about how developers and data scientists need domain knowledge in order to be effective in their work.
If you’re building systems that people will use … then your domain includes “people.”
You really want to think about how people might (mis)use what you build, and sort out how to handle it.
Here’s the latest example of Generative AI Feature Doesn’t Mix Well With The Real World:
“Facebook’s new AI stickers can generate Elmo with a knife” (Ars Technica)
Yes, yes, enjoy the Schadenfreude. Sure. But also take the time to look at your own company:
- Now would be a good time to reflect on your AI plans. Do they reflect actual business need? What do you stand to gain if they work out? And what do you lose if they don’t? Because if you’re just trying AI for the sake of trying it, you might be carrying a lot of unintended risk.
- If you’re working on a public-facing, generative AI offering, I invite you to check out my Radar article “Risk Management for AI Chatbots” for ideas on how to limit your downside risk while remaining open to the upside gain: https://www.oreilly.com/radar/risk-management-for-ai-chatbots/
Interesting use of AI in a business case:
French newspaper Le Monde is using AI to assist in translations. The purpose? To port some of their content to English, thereby growing their total addressable market (TAM).
I emphasize the word “assist” in there. Their team uses AI to bootstrap each translation, but that output then passes through human translators and then an editorial team.
(Speaking from experience: sometimes translating a document can feel more like writing an entire document from scratch. Building on a machine translation – even if it’s rough around the edges, even if you need to fill in some local idioms – sounds like a big step up.)
Pour cela, le journal mise notamment sur la montée en puissance de sa version anglaise, « Le Monde in English », lancée en avril 2022. Jusqu’à 50 % des contenus publiés chaque jour par la rédaction y sont traduits, grâce au logiciel d’intelligence artificielle DeepL, supervisé par des agences de traducteurs et, en bout de chaîne, par une équipe de sept éditeurs anglophones. Depuis la rentrée, l’application du « Monde », devenue bilingue, se charge automatiquement en anglais dans les pays anglophones. « Aujourd’hui 13 % de notre audience provient de l’étranger, donc nous visons 15 % d’abonnés en langue étrangère », affirme Louis Dreyfus.
The other day I mentioned that a well-run supply chain helps you keep promises. At the time I focused on the promise of goods being in a specific place at a specific time.
I should have also mentioned the promise that the goods moving through the supply chain are safe and legitimate.
I’m thinking about this now in relation to a recent discovery about airplane parts:
Let’s remember that the notion of a “supply chain” extends beyond physical goods. If your company deals in money or data, you’re also a participant in supply chains.
As you look at the supply chains in your business, what steps are you taking to ensure security and otherwise reduce your risk?