lemmox

FB just put shovels in the ground on a datacenter in Alberta. Bringing a new nat gas plant online nearby but it's a little quicker to bring the DC on than the plant.

Many years ago my wife fell down the stairs and suffered, what turned out to be, a broken fragment of bone that poked into a nerve. It took quite awhile to diagnose it because <medicine is hard>. Along that journey to diagnosis she was almost bedridden, and in excruciating pain, most of the time with with physio, chiropractors, muscle treatments, nerve pain meds, etc, all failing to help her function. The opioids she was prescribed made it so she could function. She was high, "dreamy" (for lack of a better word), reported still feeling the pain, but could function. The day after the neurosurgeon performed her spine surgery she voluntarily stopped the opioids and finished her healing from the procedure with ibuprofen. She never looked back and described never feeling an urge to take the meds again. I wish I could say that this experience gave us some great insight into the complexity of opioids but it didn't. In her case they did their job until she didn't need them anymore and that was that. I don't read anything more into it then that.

From the article: In an argument that appeared to flabbergast a small claims adjudicator in British Columbia, the airline attempted to distance itself from its own chatbot's bad advice by claiming the online tool was "a separate legal entity that is responsible for its own actions."

Yes, I recall this as well. It didn't pass the smell test to me - probably wineries making up a reason to justify switching to screw caps.

An interesting trick I found here was to ask ChatGPT to produce tables of concept definitions and include a metaphor for each concept to help understanding. It was quite good at coming up with metaphors and that actually felt kind of magical.

Interesting prompts! IME the quality of the answers the users give to the ChatGPT questions in these prompts will make or break the experience.

I played around with this use case in the spring when my teenage daughter was looking for extra test prep materials. At first the experience was interesting but there was an "AI uncanny valley" shaped problem: the material just didn't seem to fit. It felt wrong.

This uncanny valley was significantly reduced, even eliminated in some instances, by including the entirety of our school district's online material about the course; information about the core competencies (across communication, thinking, personal & societal), the big ideas, the curricular competency & content about the learning standards. Our district has a pretty good website with all of this information laid out for each course and grade level.

Including all of this information in the prompt context resulted in relevant and harmonious content when asking to generate course outlines, student study-prep handouts, and even sample study session pre-tests (although ChatGPT wasn't strong at reliably creating answer sheets for the pre-tests).

Context is key!

> Rust was used primarily because a couple of the founders of the company were Rust experts

It's remarkable how significantly the first few months of a project impact the first few years of a company. How quickly casual & spontaneous decisions, largely based on convenience and familiarity, ossify into legacy drag.

With nuance, I guess? Depends on the conversation.

I'd say that there's more agreement than disagreement. Here 70% of the overdoses are related to drugs cut with unpredictable levels of fentanyl. I'd say I've seen consensus on that being a "bad thing(tm)", and consensus that the criminals cutting street drugs with fentanyl are "bad guys(tm)".

I guess to your point about junkies making bad decisions, my peer group skews older, we've already said everything we can say about junkies making bad decisions 30+ years ago. It's true, but it makes for stale conversation and seems kind of orthogonal to, or at least adjacent to, the current fentanyl situation.

One example provided, which I've not fact-checked and might be limited to BC, was that disclosing addiction during a hospital visit opens up a file with child welfare services.

My wife actually asked this of some professionals working in this space yesterday at a local awareness event (she works in community services, unrelated to drug policy). The answer (paraphrased) was it doesn't directly stop the poisonings. It's in the spirit of harm reduction to open up more support channels. One example given was about going to the hospital and honesty with health professionals. Currently there's disincentives for an addict to disclose their addiction due to legal actions that get triggered upon disclosure. Removing those legal triggers could (would?) allow for more supervision without legal consequence to hopefully reduce the overall harm. It gave us something to chat about.

FWIW since I started referring to this as poisonings (instead of overdoses) I've found folks to be much more amenable to conversation. Nobody likes the idea of the wicked poisoning the vulnerable.

It depends! FWIW the spots I've used event patterns most often are high write loads that benefit either from fan-in micro-batching before hitting the data stores or polyglot systems where a client/system event need to fan-out to multiple systems and we want to insulate the producer from downstream latency/outages.

Oh wow, Pandas and sklearn were there - I was able to train a random forest classifier on the iris dataset. That's pretty slick!