I write about artificial intelligence and machine learning when I am not trying to become a better guitar player. These posts reflect my personal views, not my employer's or affiliated organizations.
My #fedi22 interests are: #artificialintelligence #machinelearning #linux #programming #python #webdev #infosec #privacy #climatechange #fitness #cycling #hiking #music #bassguitar #writing #blogging #craftbeer
On vibe coding:
I used to take notes with pen & paper in school even though I could just use my phone and record the whole lecture. Writing it myself with pen & paper increased my understanding of the topic.
Similarly - I can "vibe" code and I can make sure I read all the code the vibe generates... but I won't understand it nearly as well. And maybe that's okay - if your management is fine with relying on cursor to resolve incidents.
“#Microsoft has simply given us no other option,” #Signal says as it blocks #Windows Recall - Ars Technica
Recall, the #AI tool rolling out in #Windows11 that will screenshot, index, and store almost everything a user does every three seconds.
Signal is invoking an API for protecting copyrighted material to add an extra layer of #privacy.
I use the muted words and tags feature to reduce what I deem to be toxic posts in my feed, both on Bluesky and Mastodon.
These filters help make reading the sites' posts more enjoyable.
I agree! The ability to curate your feed to focus on items of interest, and filter out noise boosts one's mood. Mastodon's filters are a rich implementation.
>I have no doubt that agent loops will get incorporated into more day to day automation tedium that's historically been too specific for general purpose tools and too esoteric and unstable to automate traditionally.
This is my observation, too.
>We'll be seeing more custom, ad hoc, throw-away #LLM #agent loops in our bin/ directories.
I agree!
>Grab your favorite bearer token and give it a shot.
👍👊👍👊
#RCS daily message volume in the USA has surged to over 1 billion. Yet legacy #SMS daily volume is 6 billion.
RCS adoption has room to grow. The big three #USA carriers now support RCS. Many regional, rural and #MVNO carriers, though, need to make infrastructure investments.
https://wccftech.com/googls-says-iphone-adoption-of-rcs-leads-to-a-billion-daily-messages/
This YouTuber deployed #Microsoft’s latest #Phi4 #LLM on a #RaspberryPi 5.
It _only_ takes 10 minutes to respond to a prompt.
An LLM For The Raspberry Pi | Hackaday
>Microsoft’s latest Phi4 LLM has 14 billion parameters that require about 11 GB of storage. Can you run it on a Raspberry Pi?
https://hackaday.com/2025/05/10/an-llm-for-the-raspberry-pi/
Yes, 32 GB. I'll correct that.
Reflecting on this more, humans can evaluate the benefits and risks of hallucinating in the unique human social context (impact to one's credibility for example).
Consider cases of embarrassed and discredited lawyers who submitted legal briefs riddled with factual errors.
AI, in its current state, doesn't value the outcomes in this context -- because to it there is no gain or cost for it to do so.
I'd imagine AGI would need this human like ability.
I think we humans have biases and cognitive considerations that influence our reasoning and the conclusions we make.
If true, #hallucinations cast serious doubt on whether the end goal of #AGI can be achieved with today’s #LLM architectures and training methods.
While ongoing research explores #RAG and hybrid models and inference techniques, no implementation to date has fully eliminated flawed reasoning.
What consumer would trust mission-critical decisions if an AGI is known to confidently state falsehoods?
As much as I love the #Fediverse social media experience, I must compliment #Bluesky for their user-centric, composable and scalable approach to #content moderation and filtering, e.g. empowering the end user to curate their #socialmedia feeds via block lists and labeling subscription services.
The Fediverse takes a more server-centric approach, which on balance, offer admins more empowerment than end users e.g. admins share and manually apply blocklists.
What #Nvidia #GPU architecture do you use with your #Linux PC ?
1. Older - Maxwell, #Pascal, Volta
2. Newer - Turing, Ampere .. #Blackwell
/end
Like many #Linux users, I use that #OS to extend old #PC #hardware lifecycle. My #Dell #Optiplex 9020 includes a Pascal based #GPU.
My issue is I cannot install a more modern GPU because of physical constraints with the chassis itself.
>Maxwell, Pascal, and Volta architectures are now feature-complete with no further enhancements planned. ... Users should plan migration .., as future toolkits will be unable to target [these] GPUs. 1/
Test result.. yes, yes it does..
Unless they invent a novel monetization strategy for a social media site, the main path forward is selling user profile and behavior data to 3rd parties.
