This is actually terrifying. The more I read about it the more horrified I am at the prospect of putting this inside people. I'd be curious if anyone has any references to additional primary source accounts about the state of things over there.
https://futurism.com/neoscope/terrible-things-monkeys-neuralink-implants

I'd theorized a couple of methodologies for generating fine tuning datasets from a bigger #LLM to train smaller models and it looks like the technique finally has a name. Google calls it distillation.

https://blog.research.google/2023/09/distilling-step-by-step-outperforming.html

It's always a treat to watch toddlers interact with something as genuinely intuitive as #ipad.

It's also hilarious to see toddlers with no concept of the world get absolutely frustrated by full screen ads.

https://fortune.com/2023/09/14/mike-bloomberg-golf-remote-work/

Apparently remote work is bad because all the golf courses Mike Bloomberg has heard about in the last 3 years have had record summers... ⛳🏌️

GPT is state of the art technology they said. It'll change everything they said...

"Make it work, make it right, make it fast."[1]

I got to share this with my team this week most of whom had heard it for the first time. It's something I've been thinking a lot about lately. Building good software that lasts is an iterative process and one that's tied to creating business value. Not understanding the right priorities can be a make or break decision for a project.

"I always interpreted this phrase in terms of business needs, not coding standards.
1. Meet the minimum requirements for the business to call project a success. (Make it work.)
2. Add bells and whistles to make the program less prone to error and more feature rich. (Make it right.)
3. Find and eliminate waste in the process. Some assumtions from the start will have been incorrect. Remove unecessary business logic. Included in this step is to improve code for better performance. (Make it fast.)"[1]

[1]: https://wiki.c2.com/?MakeItWorkMakeItRightMakeItFast

A lot of new #gpt powered #AI chatbot systems, particularly those targetting some vertical work via Retrieval Augmented Generation (RAG).

The tech that underpins them is an embedding model that's able to encode features of the text into a numerical vector. Using #OpenAI's embedding model as an example, here's an embedding for the word "red".

Fun fact is that every embedding I've ever seen has those peaks. Not 100% sure why they are there but have some suspicions.

I'm pretty interested in human-computer interaction, particularly how people interact with mobile apps/products. While passing by my local #Starbucks, I got to see how mobile orders influenced their traffic volume during a peak traffic period.

The experience is pretty seamless for the consumer, you order on the app, walk in, wait for your name to get called, pick up your beverage from the counter and leave. Anecdotally customers that walked into the cafe to pickup mobile orders, seemed to use it as a substitute for the Drive thru which was already long.

One downside to this is that its now impossible to know how long the wait time is if you intend to walk into the cafe to place an order. A room packed with people waiting on mobile orders is the absolute worst case scenario. I imagine this was an unintended consequence of the product.

More importantly, I think the behavior mobile orders incentivise is poor from a #sustainability perspective[1]. AFAIK there's no way to use personal cups via the app, as it prevents the beverage from being prepared ahead of time however it does look like the company is doing some work to address this[2].

It's easy to forget that the products we build have consequences in people's lives and some of those consequences are unintended.

[1]: https://www.cnbc.com/2023/06/03/starbucks-has-a-coffee-cup-climate-issue-as-mobile-drive-thru-booms.html
[2]: https://stories.starbucks.com/press/2023/starbucks-latest-california-borrow-a-cup-test-furthers-companys-shift-toward-reusables/

Replaying old #games is like looking through a looking glass into the past with a new perspective. While there's been a lot of progress since, It's painful to see how poor representation was in gaming.

The anti-remote lobby continues to drop comedic gold. This piece reads like actual satire so I'm half convinced that it was written by AI.

TLDR: you should be scared of remote work because the robots are coming to replace you. Robots can't replace hybrid workers because they can't come into the office yet.

"The main workers who would be potential A.I. victims are those with easily outsource-able jobs, he explained. “So, fully remote, relatively low-level things like call centers, data entry, payroll,” ... On the other side of things, hybrid workers—whose jobs rely, in some part, on in-person work—are almost certainly safe for now. Bloom, who teaches economics courses in classrooms in Palo Alto during the school year, used his own job as an example. “If I have to go in and teach people, the robots [that could do the same] are enormously clunky—like, 2,000 pounds,” he said. “There’s no way any robots will look like a human, even within the next 10 years.”"
https://finance.yahoo.com/news/ai-revolution-could-spell-doom-192604450.html