Lightnews — Scholar-powered news

martin_e

@flourogoth.bsky.social

130 followers 540 following 310 posts

Just some bloke.
He/him.
These are my opinions, not those of my employer.
I manage people who write code.

Posts Replies Media Videos

martin_e

@flourogoth.bsky.social

Urgh. Job hunt is getting to me...
This is unlikely to help - but I do find it hard to be dishonest.
I blame @edzitron.com for finally flipping out a bit... (in a good way!)

Job application. Text of question reads: "How are you currently using AI (or thinking about using it) to improve your day-to-day work ? We're interested in how you apply AI to drive efficiency, improve experiences, or raise the bar in your work. Whether you're experimenting with tools, building automations, or shaping your team's approach - we'd love to hear how AI fits into your thinking."

My answer: "As it has been a while since I was an Individual Contributor, but rather have been working as an Engineering Manager; I am far more interested in the use of AI from both a people-centric perspective and the wider strategic landscape.

I have personally worked with Neural Network technology since the late 1990s and, in my opinion, the use of these tools needs to be balanced with the company's risk appetite.
Understanding the risks may allow you to implement controls and strategies for the newer LLM/Generative AI-based tools.

As I see it the risks which should be at the forefront of the business are:

1. Cost. Currently these tools are vastly underpriced, we are at the user acquisition stage of the providers of these tools.
The companies who are building and developing them are borrowing huge amounts of money to provide the service. This means that the cost to the user is going to have to increase dramatically in quite a short space of time. Currently the major providers would need to pay back about $2 Trillion dollars by 2030 collectively, whilst currently spending between 10-100 times their costs for the service they provide at their current rates. Reliance on these technologies does, therefore, have a high risk of becoming much more expensive in a short time period."

(answer continues in next image and alt text).

My answer continues: "2. Reliability. Any use of external services within a business would expect consistancy and foreknowledge of breaking changes.
As most LLM/Generative AI-based services are closed "black boxes" (which is also the nature of the underlying technology - at present); changes to the model used by the service provider can create unexpected results - at any time. If it is impossible to effectively test the systems reliant on the service, then large amounts of work can be created for your teams at any breaking change.
Even with self-hosted, self-trained models; if the task it is being used for cannot be monitored, or tested, for unexpected results; you may be creating incomprehensible tech-debt for your teams.

3. Damage to your workforce. Recent studies have shown worrying reductions in the ability to learn new things from those working with these technologies.
It appears that outsourcing complex congnitive tasks to LLM-based tools actively reduces the ability of people to learn, and also to concentrate. After sustained use of these tools it looks likely that this damage may be cumulative. As someone primarily responsible for coaching, mentoring and helping people to grow and learn within their careers - this is a high risk for the business.

Having said that, it's pretty good at summarising machine logs or writing boilerplate code (as long as a professional is able to investigate it's results effectively)
But most automation can be managed in determinative ways with far less risk of error - and consistancy in deployment tests or monitoring."

December 11, 2025 at 10:25 PM

martin_e

@flourogoth.bsky.social

Medical image of a man with 3 images of Trump. Text reading "His posture is the classic fronto-temporal dementia stance. Forward leaning, arms down like balancing rods. He's at risk to jerk and fall backwards. He can't stand hands on hips or in the pockets. He would topple."

December 10, 2025 at 5:37 PM

martin_e

@flourogoth.bsky.social

Meh.

Graffiti on wall stating "Be gay, do crimes"

July 28, 2025 at 1:45 AM

martin_e

@flourogoth.bsky.social

July 9, 2025 at 12:32 AM

martin_e

@flourogoth.bsky.social

Small boy raising a middle finger at a billboard with Elon Musk standing on a Tesla with the slogan "Buying a Tesla? The far right salutes you"

April 22, 2025 at 8:35 PM

martin_e

@flourogoth.bsky.social

Honestly. The kids are fucking cool. After the mad punk song where the passionate songstress complained about billionaires stealing their future, I saw this next time I needed a wee...