Notices by FeralRobots (feralrobots@mastodon.social), page 3

@ChrisWere @atomicpoet
there's this brother duo that does strength training vids; they're also competitive body builders, & they did a vid following them both through a "cut" cycle getting ready for a competition, subsisting on broccoli & chicken for two weeks. They looked like shit, exhausted, washed out, & totally owned it. 'This sucks' being a continual refrain.

@nyrath
[facepalming intensifies /]
guess we're either gonna bring 2 astronauts home in a leaky capsule or find out how much slack there is in ISS' resource budget.

"Sunk cost fallacy" is almost as pernicious a minomer as "slippery-slope fallacy."

Sometimes slopes really are slippery, & sometimes you're not gonna get the resources you'd need in order to use a different solution.

Pisses me off sometimes that the good things Machine Learning CAN do (like radically improving hearing aid tech) are getting swamped by the bullshit flowing from LLMs. (Srsly, this could improve the lives of a lot of people. THIS is the kind of stuff we should talk about, not robotic clones attending meetings.)
https://front-end.social/@stephaniewalter/112568501168054468

Implicit claim: training off captured data is done locally
Likely bullshit, in several ways.
To my knowledge only Apple has baked in the hardware required to support this (their secret, previously unused chips).
Without GPU support, would use lots of CPU time; either way locally-training machines would run hot & have shortened lifecycles. (faster turnover! drives subscription-based licensing!) ...
#Microsoft #Recall

Then we get to the whole MS-OpenAI relationship. It's been widely pointed out that given observable financial realities, MS could end up stealth-owning OpenAI via compute credits (which it will probably report as revenue); but is anything in the GPT family portable? How hyper-local can you get?
I think the bottom line ends up being that #Microsoft #Recall is largely #bullshit, in the sense of not really being a product that will ever exist in anything like it's described form.
...

I think the whole 'everything is local' spin will drive a lot of breathless speculation about MS using open source models, or only-local models, etc., & I think that's just MS enabling people in bullshitting themselves. Truth is we know nothing about this system because it doesn't exist yet; & velocity of dev suggests that when it does exist, it'll be a cobbled-together mess with massive vulnerabilities that's passing huge amounts of tokenized data to a cloud-based LLM.
#Microsoft #Recall

Implicit: local data does not contribute to the cloud model's training
Again, what's the utility of the product if it's not available through the cloud? Microsoft gave up on mobile. Fantasy-assistant #Samantha isn't very useful if she can't be with you everywhere.
Bullshit element here may be that Recall as a useful product is vaporware that's primarily intended to drive share price, not an actual product. But M$ usually still wants revenue.
#Microsoft #Recall

... & would kill battery life on the as-yet-vaporware #CoPilot Notebooks.
So I don't think it'll really (at least not fully) "train locally". The scheme must somehow involve training in the cloud. Which means some kind of data has to pass to the cloud. Where (again) it can be exploited.
#Microsoft #Recall

Final point because I look back & realize I didn't explicitly make it: #Microsoft #Copilot #Recall is basically in no small part a way to mine the personal files & business work product of users for #GenAI training data. Any claims to the contrary don't pass the smell test. In the process they will be incentivizing more-rapid hardware update cadences, which means either new licensing revenue, or more users shifting to subscription model.
All with added energy & resource use from running harder.

But does that really matter? Yes: because what we know they CAN do (take screenshots every 5 seconds, encrypt them badly, do some kind of localized tokenization that sends exploitable data to the cloud using systems with known weaknesses, etc.) is pretty bad.
#Microsoft #Recall

@weirdwriter
Their assumption that algorithmic serendipity is somehow inherently un-tainted by algorithmic bias would be charming if it weren't so insidiously dangerous.

@powersoffour @futurebird @cyberlyra
Thinking this could violate data privacy laws in the EU & in a number of US states. So, not surprised (& in fact conditionally pleased) that the capability is no longer there.

@goatsarah
OK so not just me.

@inthehands
I'm having some trouble with this. My experience, along with counsel I've been given over the years, is that one MUST attend to the clearly stated job requirements. That someone's flagged your application to ensure it passes, they must be addressed if one's to make it through the initial screen.
& that in an annoying number of cases, the requirements literally CAN'T be ignored, because the screen is automated.

@goatsarah
Goodbye, Earl.
(Srsly though here in the US this would more typically end with Earl murdering Wanda & probably Maryanne, too, if she lived in easy driving distance.)
https://youtu.be/bqnrXRuebWg

@ironchamber @anildash @mathowie
What I read in school (early 90s) focused mostly on menu systems for applications; hypertext usability wasnt a big thing yet. But now that I think of it there've been Nielsen-Norman Group articles over the years. They're AI shills now but for a long time there was solid stuff in there, even if one had to adjust some for Jakob's massive ego.

@ironchamber @anildash @mathowie
There was actually a fair body of ergonomics research on this, but the literature I'm familiar with is pre-web. I'd have to look up for specifics (I still have that ergonomics topics reader from school) but my recollection is that they worked OK for a small set of users with very narrow needs, but were a major detriment to power users. IIRC there was speculation that it would also tend to narrow the options that were used through a negative feedback loop.

@ironchamber @anildash @mathowie
OTOMH, the way sharing works in Android, I never know where I'm going to find the target for my share, because Android keeps resorting based on how it conceptualizes 'recency.'
I've also tried out adaptive UIs in several desktop apps over the years - bad-penny pattern there is pruning elements from the menu structures if they're not used within some period of time. (Word did this at one point, IIRC, but I've used others.)

@anildash @mathowie
man do I ever hate, hate, HATE adaptive UIs. I have used many & have yet to see one that didn't make things harder for me on the average, vs making things easier on one task maybe 30% of the time. The added cognitive load of trying to figure out where the feature I wanted has gone just leaves me irritated until I sigh & accept it as unchangeable. But I never like it.