Microsoft's Recall should be celebrated as the savior of SMEs and scourge of CEOs A year and a half into the explosion of AI fueled by ChatGPT, the hype and fear of missing out has begun to thin just enough to make out the shape of two starkly different visions for AI: one that imagines using it to replace people and the other that wants AI to enhance people. I spend a lot of my time working with the latter …
Celebrated indeed ...
W . T . H . F ?
More likely is that MS ...
Likely?
Be sure of it, albeit not in the exact sense you mean.
MS Recall is yet another metadata harvesting machine, the product of which (given its mind boggling size and spread) is now possible to analyse/process faster and more efficiently than ever before via AI and the incredible hardware available for it.
What really baffles me is to see how so many people eg: here at ElReg of all possible places, cannot see past their nose and really consider that MS' Recall is something useful and convenient.
But then, such are the times we live in.
.
"What really baffles me is to see how so many people eg: here at ElReg of all possible places, cannot see past their nose and really consider that MS' Recall is something useful and convenient.
Really?! Exactly how is MS Recall 'something useful and convenient'?? Think: how often do you use your own browser's History function to recall your own browsing patterns? And you've had that ability for DECADES.
I like the answers from other posters: AI's recording of your activities only shows the activities and not the reason why you are doing them with the choices made. An AI automation of an activity only automates the activity for the *exact* patterns of necessity shown. Power users have had macro functions available to play back activities (also for decades) but ask how much actually have been automated because they could, and any SME user worth their productivity salt already has shortcuts and macros for what can be accomplished best in that manner.
At this point it is a complete and utter DREAM to think that, by recording your activities, somehow your computer will know what and how you need to accomplish the various tasks you handle every day in the SME environment and be able to take them over by creating automation procedures. These machines don't have that capability yet, regardless of the hype - hell, they can't even stop "hallucinating", read MAKE MISTAKES that they have renamed to whitewash the failure through the power of linguistically-tuned propaganda. They can't even get established requirements right, what makes you think that creating yet another requirement for them to fill will be any more successful??
I read that (from the anon OP) as surprise that some at el Reg would consider it useful and convenient, when it's far, far more likely to be a data harvesting tool. Another MS land-grab.
Given the rest of the post, that seems to be the intention, rather than suggesting it really is useful and convenient when it really isn't.
With Google, OpenAI and now even Apple doing it (harvesting for / from AI), how can we expect MS to not want their slice of the [money] pie? Everything in our society has now turned to be about MONEY, so I guess I'm not surprised any more.
I'm more surprised on how weak the masses are in buying into a system where they've been promised the world, if only they would think about profit and money...yet, after 40+ years, they are still at the razors-edge of distress, still in the financial dumpster. Still hearing the promises yet still grabbing for the gold ring that is, quite intentionally, kept just beyond their reach. How dumb can you be? Apparently, you can't lose by underestimating the gullibility of the average person :-(
I read that (from the anon OP) ...
... rather than suggesting it really is useful and convenient when it really isn't.
Indeed ...
What you read is exactly what I meant to say:
MS' Recall is nothing but a data harvesting tool.
Alas, a few of my fellow commentards didn't get the gist of my contribution.
Hopefully it has now been cleared up.
My sincere apologies for the confusion.
.
how often do you use your own browser's History function to recall your own browsing patterns?
And that's not even a particularly apt comparison; the browser history function is not only much less of a risk than Recall, it's much less harmful to the user.
Chronicles — chronological lists of past events and activities — are complementary cognitive technologies. They improve memory but require human reading and interpretation. An opaque automatic analysis system, as Recall purports to be, is a competitive cognitive technology. It impedes human cognitive processing by offering conclusions without performing the intermediate steps.
As with most applications of generative AI, Recall is another royal road to learned helplessness, to blindly following "computer says so" and unlearning whatever modicum of critical thinking the user might have previously developed. It's intellectual laziness. It robs users of mental paideia and serendipitous discovery. And that actually gets worse the more accurate and capable it becomes, so there's an adverse tradeoff between the penalty of error and the penalty of stultification.
Many people — intelligent, thoughtful, educated people who normally think carefully about consequences — are praising the supposed short-term productivity benefits of gen-AI. Very few, even those concerned about other risks (existential, political, economic, social, whatever) of an AI explosion, seem to be considering the deleterious effect these tools will have on thinking.
micros~1's Recall only records WHAT the user did on their computer. It cannot capture WHY they did it.
Therefore, the LLM's analysis of that log will not reveal any picture which captures why the user is doing what they are doing and will miss the most important part of the process. You can't automate things where you don't understand the reasoning behind them.
Sadly, your business partner is merely engaging in the same blue sky thinking which probably motivated the creation of Recall in the first place.
.. and not catching casual office conversation, phone calls, spoken word on Teams / Zoom or whatever conferencing tool used.
A fair chunk of work I do on a given day is decided by management (as usually lots of things of "equal" priority so management apply their take on which is first (second, etc.)amongst equals) this is done via voice calls, lots of chats with colleagues on Teams, a few typed things but anything vaguely complex and revert to spoken word as faster & clearer.
So a (security / GDPR nightmare) of recall logs is lots of data but with little hint as to why various things were prioritised, how something was solved (the snapshots may have the code to fix a longstanding virtually never reproducible bug, but not the discussions between 2 engineers that led to a lightbulb moment about what the underlying issue was)
Also it's only capturing what's on the screen every 5 seconds, and it's not recording inputs - so that alone won't be sufficient to automate processes.
No. But collected in bulk, it's enough to build a low-resolution but still highly complex model of the activity on end-user devices across an enterprise. And that, in turn, can be used to identify what needs to be inspected more closely to gather the necessary data for a more-detailed model.
Poor maps still speed further exploration. And as we say in the security biz, attacks only get worse.
the businessman in the example is only asking the AI to document what is being done, one inference is that they know the why.
Eg if the computer is that of an office worker who processes orders, then they get the traffic analysis of moving between software as they entering the order details, consulting stock records, emailing the customer and so forth.
Without knowing the "why" it would be difficult to correctly automate the process. Take, for instance, a situation where the business is down to the last few items of some product. Orders come in from two customers. Whoever's doing the stock allocation chooses one customer rather than the other.
Why? Is it random? Is it based on knowledge that one might have a more urgent need? Is it that one is a faster payer than the other? Does the business have more sales to one rather than the other. Is one customer a good mate?
Without knowing why the system has gained no information to enable it to make a similar trade-off in the future in the way the business might approve.
A more useful approach would be software that analyses demand and lead times to avoid that stock level issue entirely but that would be ordinary statistical analysis that needs no fancy clothes.
Depends on the nature of the sales.
For example, for click and collect with collection within one hour, I really want my order to be satisfied by what is actually in stock in the store, ie. When I pay (online) for the last gismo, the store will prevent the checkouts from selling the last gismo to a walk-in customer.
However, if delivery is next week, my order may be fulfilled from a consignment currently in transit. With Dell, for example, I can order a system they will only assemble on receipt of order.
Ah, the Amazon model. Like the insect repellent they had listed in stock and I purchased, only to not receive it within a couple days as expected - then a week, two weeks... Turns out they didn't have any in stock and didn't know when they'd get more. I think after a month we got the order cancelled.
Even if delivery is supposed to be next week, a biz faces serious reputational damage for selling something they don't have on hand.
a biz faces serious reputational damage for selling something they don't have on hand
Yeah, that'll have Amazon quaking in their boots. They're big enough to not give a rat's ass about their reputation. Amazon could randomly ship bombs to 1% of their customers and most of the remaining 99% would continue to take the gamble.
The beauty is that it doesn't matter because current generation AI doesn't know anything or grasp why anything happens. When asked to reproduce an office workflow it will just move the mouse about in a way that's a bit like what it has recorded and produce some text that sounds a bit like the kind of text it recorded and call it good.
The sheer depth of uselessness of this generation of AI for these types of task is very hard to express and it's painfully clear that a most commentators are not close to grasping it.
When those random AI workflows have an acceptable failure rate (cost of AI fuck-ups < cost of employing people to do a decent job) then a lot of companies won't care. Look how much everyone hates chatbots and "AI" support. Look how much use they are in fixing problems. Look how much the companies that use them care; they don't, cos it's cheaper than paying people and customers don't take their custom somewhere else and soon there'll be nowhere else to go cos the "AI" will be everywhere..
Someone with a memorable name needs to come up with a curve that goes up as software tech improved people's lives then starts to go down as it makes them worse until it goes below where it started. I think we're at the peak of the <memorable name> curve and it's going to go down faster than it went up.
"and customers don't take their custom somewhere else"
This is something that the deployers of this crap almost certainly don't measure.
"and soon there'll be nowhere else to go"
This is worrying. However sooner or later somebody might decide that winning the race to the bottom hasn't brought competitive advantage and start a race to the top. Imagine, for instance, how quickly a bank might conquer the banking marke by starting to open branches in towns and villages, staffed by helpful and empowered staff.
Perhaps if you'd like to define "know" and "grasp the why" then someone could take a stab at it.
As always, far too many Reg commentators confine their criticism of LLMs and other gen-AI technology to "I don't know anything about it, but here's what's wrong with it". Your arguments might be useful if the lot of you could show you'd done a little research.
There are many good reasons to oppose gen-AI technology. There are many valid critiques of deep (mostly-)convolutional ANN stacks, transformers, diffusion models, annealed gradient descent, etc. There are many insights that can be gained by following interpretation research, such as the results from various experiments with linear probes, SAEs, feature enhancement and suppression, and so on.
But "buh it doesn't know stuff!", without even the barest attempt to define terms, is not such a critique. It's either a childish refusal to think about the problem, or an appeal to some sort of mystical dualism. Neither is compelling, or even worth entertaining.
SotA LLMs most definitely have a world model; that's been conclusively established by any number of experiments. That world model may be entirely Saussurian — established by the relationships among semantic tokens, with no Piercian "indexical" aspect (because the underlying technology has no pre-linguistic access to the world) — but that does not in itself prove that it's inadequate for something that could be legitimately called "knowing". For that, you first need a definition of "knowing". John Searle could get away with the phenomenological objection ("I'm not sure what thinking is, but I'm pretty sure it isn't that"), but he was talking about something (symbolic evaluation) which has much more of a prima facie distinction from human cognition, and he was John Searle. You're not John Searle, and neither is anyone else mounting the "they don't think" argument here.
Now, if someone wants to mount a real argument that using a very-highly-dimensional vector to select a point on a very-highly-dimensional manifold and then do stochastic gradient descent, update the vector, and repeat the process is not "thinking" in any useful sense, I'd love to see it. I'm far from convinced myself that it's what I'd call thinking, if I were to spend several years coming up with a suitably refined definition of that term. But assuming it's not a priori is anti-intellectual laziness, pure and simple.
(And JFTR, Searle believed AI was possible, because he believed the human CNS was mechanical. He just didn't buy the symbolic-evaluation route.)
I've said it before and I'll say it again: deploy that crapware on my business system that has SECURE information on it, and you're GONE. I've used Ubuntu for years, and ever since a proper version of Chrome was released for it, there has not been one single thing I need to do that can't be done with Ubuntu. In fact, almost everything I do is on Ubuntu VMs save for playing games.
It's high time somebody shot Microsquishy's marketing department and management responsible for this abortion.
It's funny; with most controversial / polarizing ideas, you always get a faction of people saying "now, hang on; if you look at it from the angle of the ( doctors / unemployed / politicians / shareholders / capitalist elite ), I can see that this proposal make sense, because...". With Recall, however, this is the first such post I've seen in several weeks (excluding comments from Microsoft, of course) - and this story is completely built on a guess. Seems almost poetic, doesn't it.
What is it they say - "If the nicest thing you can say about an idea is that it's not illegal, it's probably not a good idea."? And I'm not even sure you can say THAT about Recall.
I've been critical of a number of Pesce's pieces (say that five times fast) in the past, but in this case my sense of his tone is that he's not praising Recall, just noting how it could serve an automation goal. And he doesn't seem to be claiming that's necessarily good.
So while engaged in a task at random time open excel and add two random numbers and convert the sum into a date, then proceed until another random time, then open word on the org's procedures manual or code of conduct and spell/grammar check a random page, then proceed...
Get the idea? Use your imagination - spell checking invoices probably good too.
With these diversions Recall's AI should be producing some LSD standard hallucinations.
Throwing clever clogs into the looms the AI revolution. :)
Didn't one of the recent Reg articles on Microsoft's latest terrible idea mention the anti-phreaking scheme from Stevenson's Cryptonomicon, where the protagonist writes a program that just keeps popping windows up randomly on his screen and obscuring his actual activity? That'd work a treat for Recall, if all it's doing is screen grabs and not actually logging keystrokes. (See the novel for the full description; the screen activity is designed to mask the actual typing and so on.)
Yes, it'd be annoying to work in that sort of environment at first, but I expect most users could get used to it quickly. Particularly those who can touch-type.
"Recall provides the perfect tool to capture all interactions with the computer"
I challenge the assumption of 'process capture' by recording interactions with the computer. For example, I once had to sort out PCI-DSS compliance for a client, and found that they were completing their self-certification without referring to what was actually being done -- a staffer was merely filling in the periodic computerised compliance report by tweaking the previous one, which had been completed by tweaking the previous one, and so on. The interactions with the computer were all perfectly reasonable. The problem was that the information entered bore no relation to reality. So a review via "Recall" would have yielded a positive report on a process that actually yielded utter garbage.
Birmingham City Council/Edinburgh University train-wreck Oracle Fusion implementations need some Recall to fix them.
Maybe useful for Fujitsu for Horizon too.
Might be useful to put on Paula ‘can’t remember’ Vennels (should she ever get annother job) and other CEO/Government Minister’s laptops - for future UK Public Inquiries!!!
The problem with this is no different to the problem with business improvement using traditional tools - the egos of the directors. My experience is that none of them like you turning up and saying stuff like "we've collected data, done the analysis and it turns out that need to change xxx to yyy to improve zzz by 28% with a confidence limit of 2%." The concept that you can improve the business with some observation and data crunching is seen as threat by directors who assume that any business success is wholly due to their genuis (and any failures are someone else's fault, of course). It's bad enough when highly-paid consultants do it, but a "free" app running in the background.....?
In a previous life, when we were split off from the main company, manglement decided they needed costings for the work that was previously uncosted
For stores, they decided 2mins to book in each item... whether a large device containing multiple cards, serial numbers and licenses or a box of 200 ethernet cables (@3hr 40 per box)
Quite. The idea that knowing what's going on in your business "costs more than most SMEs can afford" seems bizarre to me. Particulary in an SME it probably means someone in a senior position spending a total of a few days' time spread across the year actually down on the shop floor. If you can't be bothered to do that, you probably chose the wrong career.
And business process development is not simply be about snapshotting current practice and codifying it:
Plus you need to check; there bike might be a comedy bike and you really have to turn the handlebars left to steer right…
A customer workflow system I analysed in the 1990s had impressive user productivity levels, but little actual completion of work. I discovered the users had discovered work could be passed hot potato like round and round the system…
This post has been deleted by its author
It wouldn't be a comedy bike - the comedy motorbike would be the one where you turn left to go left. To make a motorbike turn left you turn the bars to the right. Many people, some who have been riding motorbikes for years, don't believe it when you tell them and then are gobsmacked when they next get on the bike. All that time they'd never understood what they were doing, but still they managed to get where they are going.
Sorry was thinking of bicycle couriers, not motorbikes:
The Backwards Bicycle: What it takes to Unlearn Old and Learn New Habits
Not ridden one myself, but have driven a reverse steer Land Rover (with and without blindfold) as part of a training course…
I really noticed the different handling when I went from a smallish motorbike to a very heavy Kawasaki Z1300, 6 cylinder monster. It was noticeable I had to nudge the bar to the right when turning left, to drop the bike towards the left then straighten up again.
I mostly rode a fairly light bike -- a KZ200 -- at relatively low speeds, and the reason I don't believe I had to turn the handle bars the "wrong" way was because I didn't.
Handling depends on wheel size and weight -- and on steering geometry and speed. It's more complex than any simplistic description like "you have to turn the handle bars left to turn right"
Think about it for a second. You're in a turn to the left so you need the 'bike to lean over to the left in order to keep the forces equal. If the bend tightens and you need to go more to the left then you need the bike to lean more. However, turning the bars to the left at that point would cause the front wheel to track inside the rear one and so lift the bike back towards the vertical. To go more left, you need the bike to lean more...so you steer right, thus dropping the bike further over. Obviously, this logic only applies when forward motion is involved - but it's rarely a problem generating sufficient forward momentum on a decent bike: it's what they're rather good at :)
Of course, you can always perch your upper thigh on the saddle so your whole body is on the 'low' side of the bike and thus bring the CofG a fair bit further down - but that's generally considered bad form on highways. If, on the other hand, you're giving it the full beans on a track....well. It'll delay your having to get off until maybe the next fast curve ;)
Because of steering geometry, when stationary, simply turning the wheel drops or lifts the bike.
You don't need to turn the bike left to throw you right so you can turn right: gravity does that because turning the handlebars drops you to the side.
Since this is a gravity effect, the height of the COG wrt to other dynamic and static centers is critical.
Bike steering effect is a complex action that has been studied. It's not simply a matter of "turn left to turn right".
There is so much wrong with this entire argument. It seems that the argument is that the company wants to snoop on employees to try to understand how they do their job (what protocol they follow). Well, shouldn't the company have provided that at the start? The employee ought to be doing what the company told them to do, not just making up random stuff from day to day (that's the job of middle management).
Besides, wouldn't it be more productive to pick a representative employee and have somebody haunt them for a day taking notes? Better than looking at screenshots and guessing, they can ask questions immediately rather than expecting the employee to remember inconsequential moments from days ago.
Lolz.
I'd they had the time or inclination... They would have already done so without recall features. Don't need recall. This is just data for Microsoft to hoover-up to train AI. And you can't stop them.
There is virtually no business value here and a ton of potential liability and problems for employees and employers. I would say never win11, but I will have to eventually... For example, in the states, can't even run tax software in Linux. I think the enterprise will see this initially for what it is, but they can only hold out so long before upgrading anyway--and MS knows that too.
In most cases the provided protocol is a little bit wrong.
Usually it's simply that steps are missed and irrelevant items required, but sometimes it's simply plain wrong.
The people actually doing the job generally know this and just get on with it. They're often really surprised when someone shows them the "official" process, and will happily point out the gaps given half a chance.
This is why ISO 900X exists, the theory is supposed to be that each procedure is continually updated to match the reality of what the employees really do.
In practice, that never happens, and "Recall" won't change that - every five seconds assumes the user is very slow, and LLM summaries hallucinate by design.
Except that it doesn't record any keypresses or mouse clicks - it just OCRs text from screenshots several times a minute. The article assumes that literally everything is recorded and that's not the case. In which case you can't automate workflows from it unless you're making lots of guesses at what the user is actually doing - which are likely to be horribly wrong because there's nowhere near enough information there to do so accurately.
Lots of weird and bad understandings would be made by doing this. Imagine if an error message came up on screen whilst doing a task and that got captured by Recall. Your attempt to understand the workflow would then assume that that message was part of the work. And that's a simple problem - users task switch between applications constantly, often for unpredictable reasons : how do you construct a reproducible workflow from that?
Jesus, has anyone here ever been employed in business?
If someone employs you to do something, be it create a workflow from an existing process or whatever, nobody is interested in that goal being accomplished. The only thing that matters is that it looks like you are doing it and you have some result which looks like what was expected.
If an LLM hallucinates its way to a plausible outcome, that is mission accomplished. Perception is reality.
If all this is doing is snapshotting what it sees, without understanding why the thing that's on the screen is on the screen, then I don't see how the data could be of any real value. Certainly, if they tried this with one of my colleagues, the analysis would likely show that completion of most business processes routinely include steps which involve playing Solitaire.
Open Excel, add some numbers to a column
Switch to ERP, fiddle with some stuff
*take phone call from customer*
Write email to senior colleague about the call, *get a phone call back from them*
Start a thread on IM with peers on the matter
Thread drifts into unrepeatable nonsense
*phone customer back explaining it'll be a chargeable feature*
Go back to the Excel spreadsheet, continue adding numbers
Write an email to your mum
Browse the Register
Browse the news
Start researching the latest VM offerings of cloud provider
...
I could go on (and no, this isn't like my day) but you get the picture - a lot of noise and missing context
Don’t forget about the random notifications and “news and interest” teasers…
The only real value of Recall is to see what applications you are running and websites you are visiting. So that means opportunities to identify the use of non-Microsoft products and the viewing of NSFW materials.
The description of SME staff and value reminds me of presentations I used to give on "knowledge management" - the focus being on corporate knowledge and minimising the disruption when key staff move on. I used the title "Ignorance Management" because it's ignorance that actually causes the problems.
I centred a large part of my talk on a modified Johari window:
Top left: What you know you know - explicit knowledge, the stuff you could write down in a handover note without prompts
Top right: What you don't know you know - implicit knowledge, what you wouldn't put in a handover note but could describe if specifically asked.
Bottom left: What you know you don't know - explicit ignorance, where you know your limits, and when to refer to others or look up in a book or online.
Bottom right: What you don't know you don't know - implicit ignorance, where you don't know your limits and where you generate material for Monday's "Who Me".
Training adds to the first category, although much of it moves into the second with familiarity and routine.
Education shifts a lot out of the fourth category but, contrary to what many assume, the real value doesn't come from populating the first one. Education (especially formally structured academic education) converts implicit ignorance into explicit ignorance and teaches limits. An expert is someone who knows their limits.
That doesn't add anything to this thread other than supporting the view that Microsoft's development won't really help anyone except Microsoft...
You've missed:
"Stuff we are SUPPOSED to do this way but actually do another way because we can't be arsed".
and
"Stuff we are SUPPOSED to do this way but actually do another way because the way we are supposed to do it was invented by some knob head who has no effing clue"
I've been known to say that every company has five management systems:
The one documented in the Quality Manual;
The one documented in the Health and Safety Manual;
The one documented in the Environmental Protection Manual;
The one documented in the Finance and Governance Manual;
What people actually do - which may, often unintentionally, include bits of the other four.
The meatsacks reading this article have generally outdone so called AI.
1) Watching the screen of a user willl *not* capture a process ... too many unknows regarding why things are done a particular way !!!
2) Sounds like an attempt to 'find' a use case for 'Recall' ... meatsacks have seen this ... AI would not !!!
3) MS must be putting pressure on the people to promote the 'wide & varied (!!!???)' uses for 'Recall' [Cough .. spit]
4) Looks like 'Recall' is coming whether you like it or not .... typical MS 'WE KNOW BEST !!!'
5) 'Recall' will *NOT* be aiding the replacement of jobs by AI any time soon !!!
6) I did not think the hallucinations from AI would actually pass to the 'Meatsack world', and you thought COVID was bad !!!
:)
Given Microsoft's track record of reliability and confusing software, it will be hard to configure and have exceptions and problems all over the place. In using this recall feature for the smallest SMEs (1-20 employees), we are going to end up with huge privacy issues within the company.
People use each other's computers; they browse for shoes while working; staff leave computers logged in overnight. They watch naughty videos at lunchtime. SMEs are, compared to big corporates, much less disciplined in their approach to security. All this will be recorded in such as way as to make it easy to identify processes with AI. I can't see this being anything but a distraction.
Just 'a distraction'!
If you send your health records to your doctor or a CV to a potential employer, do you really want your doctor or potential employer to have a relatively insecure copy of your data that companies like, um, take one at random, Microsoft getting hold of it?
What are they going to do with it? This is Microsoft remember, who changed the 'close window' click to 'close window and accept what it asked' in the not too distant past.
Did they ask MY permission to take a copy of MY data and did they ask MY permission to use it and, again, give me a valid reason for doing it?
A series of pictures will not give any hint to why the things on said pictures actually are there. It might give information on some sequences but i really doubt that those will be even complete sequences, as many others already wrote, people tend to "hop" between tasks.
So i will try to reduce it on a really cumbersome ERP/logistics abomination a company i know keeps using.
(I really would have liked to complete that sentence with "against better knowledge" but... sadly...)
Now imagine all this on the same full screen mode browser window:
You begin the day by entering your number and clicking on the "i´m here" button.
You begin any task by entering your number, the task number (at least, that one is from a scanned barcode) and clicking on the "beginning task now" button.
You end or interrupt a task the same way, only with the respective other buttons.
While doing your work (in this case, someone assembling electrical cabinets) you switch to the material/logistics tab and click yourself to RSI because every little cable gland, contact strip, switch, relay..... needs to be traced. Normally, there should be automatic import for the parts list as soon as the engineering section closes their design phase but there are always hiccups in the plugin and much too often additional requests, which trigger another task number.
All a series of screenshots will show is a picture of the respective menus. Since the real work is done without the computer, recall will quite likely show lots of the same image of the last processed things while being totally unable to show the insane loss of time due to the many shortcomings of that ERP/logistic abomination. It will also be unable to optimize any workflow, as it cannot see/capture said work.
After all, everything is within the software, all the menus and submenus are used in the correct order. How would any "AI" determine that the main issue isnt with the correctness or speed of the "process" but that this "process" exists ?
More office oriented, how will any "AI" optimise your email or text writing ? Its not like the interruptions from real life interferences (meetings, calls, other stuff) will ever be shown on screen.
All in all, its a nice try to agressively find a made up reason for Recall.
And it failed.
"...How long until Microsoft admits the real purpose of Recall is to give businesses the tools they need to improve staff productivity by automating their everyday chores?..."
About the same time as it takes to realize that when at the receiving end of a pink slip, one cares very little whether they have been replaced by an Ai, or by their "Ai-enhanced" colleague, on account of the end result being the same.
They've postponed its introduction to "utilize the insights of the user community to review quality and security" of Recall.
In other words: they've blindly come up with some use-case that will showcase AI so Microsoft's stock valuation can go up without any thought of the consequences.
Like most other management tools ClippyAI will be a) restricted to senior management and b) used to further exercise control over the workforce. For one such example, see where management attached a device under your desk to detect when and how often you were away from your desk. Purely in the interests of empowering the workers /s
. . . Register readers miss the forest for the trees by focusing only on what a single component can do right now. Once Recall is deployed, getting it to log mouse actions and key presses should be relatively trivial. I can also envision the screen capture process being sped up when the user is interacting with an input device to ensure that what's on the screen matches up more closely with inputs. Even if not, the AI inference engine should be able to fill in the gaps to reproduce a reasonably continuous log of total user activity. Combine this with mandatory always-on webcams and microphones, and the computers in an office will be able to, in aggregate, create a complete picture of all activity within that office. For people working from home, the company will be able to get a more complete sense of the ambient workplace environment where the remote workers are.
Furthermore, a wealth of information about how the employees are slacking off will become available. It will, of course, be immediately obvious when employee # 104512313576 is browsing Amazon instead of filling out form 27B/6, but, again, correlated with activity from the webcams, the company will know when an employee closes their eyes for a moment, perhaps to pray for the sweet release of death, thus reducing their efficiency. This will make it easier to submit employee for disciplinary procedures or send them to mandatory efficiency enhancement training with Patricia the taser-equipped HR droid.
As to objections that the AI will not know "why" the employee is doing something, it can certainly be programmed to detect a broad spectrum of activities which are not obviously work related (such as reading The Register on company time) and alert TPTB or perhaps autonomously disable the offending activity.
The possibilities are pretty limitless once the basic functionality is in place, and the lack of imagination expressed by the commentards here is pretty discouraging.
I think there is probably some cognitive dissonance involved. People have jumped on the consensus that Recall is, and will always be, worthless to everyone but Microsoft, and they seem to be unwilling to consider any ideas that contradict this. The correct argument to be making, in my opinion, is that whatever benefits there are seem to be outweighed by the privacy concerns at the moment.
"... whatever benefits there are seem to be outweighed by the privacy concerns at the moment."
Please list some *Real* benefits ... not made up nonsense about replacing jobs *but* real happening now benefits !!!
There are *no* real world benefits to 'Recall', even MS is making up stuff to sell it !!!
The primary purpose of 'Recall' is to include on every PC functionality that has *other* uses in mind further down the road , so to speak !!!
This functionality will *not* be under your control and when the all the hackers etc work out their methods *they* will have a field day !!!
:)
As I said, cognitive dissonance, and/or the lack of imagination that Throatwarbler mentioned above. Thanks for illustrating.
"real happening now benefits" do not exist that I am aware of. Nobody is using it yet. The point is that as the tech matures, people and organizations will undoubtedly find plenty of productive uses for it. It's a privacy and security nightmare, to be sure, but it's being developed, and there isn't much we can do about that. All we can do is try to make people aware of how easily this data can be accessed and stolen by malicious actors.
Yankee Doodle Doofus,
FYI: "Cognitive dissonance is the mental discomfort that results from holding two conflicting beliefs, values, or attitudes."
To quote your response, "As I said, cognitive dissonance, and/or the lack of imagination that Throatwarbler mentioned above. Thanks for illustrating."
I am confused, I have *no* conflicting beliefs regarding 'Recall' and it is through having some imagination that I am able to forsee future less than excellent uses for it.
Please clarify what exactly was illustrated !!!???
"The point is that as the tech matures, people and organizations will undoubtedly find plenty of productive uses for it."
This is based on nothing more than 'hope', as you stated 'Nobody is using it yet.'
There is no god-given rule that states that 'Tech' matures, or that maturity is towards something 'Good/better/life affirming'.
Your whole argument can be summed up as 'Things will get better because that is what *I* want and believe'
Fine, state that *but* don't attempt to argue that your view is true because my view is flawed !!!
We are both extrapolating 'little information' into the future, I am including my experience to date; you are simply hoping for the best !!!
I personally prefer experience to hope *but* best of luck .... I will save my 'I told you so' for later !!!
:)
Thank you ... brave down voters [1 I can almost identify !!! :)]
I do like well reasoned and erudite argument and your downvotes are *not* this !!!
Too afraid to put your mouth where your 'mouse click is' ???
Of course not ... silly me !!!
Once again ... best of luck in the land of 'hope'...
I think I may need quite a few 'I told you so' tokens ...
Starting to 'stamp then out' now, to get ahead of future demand
[As you can see, Experience over hope works like that and can save you much time or disappointed customers !!!]
:)
> "Cognitive dissonance is the mental discomfort that results from holding two conflicting beliefs, values, or attitudes."
Indeed it is, and one of the most common (and least healthy) ways to avoid that mental discomfort is to ignore or deny any info that conflicts with a strongly held belief, which is what you seem to be doing in this thread.
"and/or the lack of imagination"
A long time ago, back in the days when S/W was written without the assistance of CoPilot or ever StackOverflow we used to do this thing called error handling. Very old-fashioned, I know, but it involved thinking "what errors might be thrown up by this piece of code hitting a problem". It required using a degree of imagination and then adding code to handle the errors we could imagine happening. Those of us old enough to have done that are actually using that imagination and what we can see are a huge stack of likely bad consequences arising from nobody responsible (responsible? Ha!) for this idea having stopped to do that old-fashioned thing.
Those consequences include all manner of complications relating to legislation such as GDPR and also the neat packaging of all sorts of information for exfiltration by the bad guys.
To put it simply, this has FAIL written all over it.
Dear Ghod, I hope you simply forgot to add the sarcasm tag! But your last sentence makes that a difficult thing to believe:
> The possibilities are pretty limitless once the basic functionality is in place, and the lack of imagination expressed by the commentards here is pretty discouraging
So far, the consensus appears to be "we, the commentards do not want this imposed upon us" and, in response to the ideas given in TFA, that does not change: what is suggested is unappetising.
You bemoan the lack of imagination from everyone else, not being able to come up with any palatable reason for wanting Recall active - and yet your brilliant idea is for an even more appalling scenario; the only thing you forgot to add is the cameras in the lavatories, to ensure that loo breaks are really needed!
The scenario they present is indeed appalling, to me, to you, and I assume to Throatwarbler also. This doesn't mean it's not something that some organizations would want to do. I think Throatwarbler was just pushing back against what many in these comments seem to be saying, which is that there is no possible way that this technology, once ironed out, could be used by organizations to learn how to automate parts of their workflows. It does indeed seem to be a possible use-case, and people with more imagination than I will surely be able to come up with plenty of others.
"You bemoan the lack of imagination from everyone else, not being able to come up with any palatable reason for wanting Recall active - and yet your brilliant idea is for an even more appalling scenario; the only thing you forgot to add is the cameras in the lavatories, to ensure that loo breaks are really needed!"
You misunderstood my point. Everyone else in this thread, for the most part, is commenting on how useless Recall in itself is for the *legitimate* purposes they imagine. My point is that Recall can be combined with either current or easily-added capabilities to make it an essential component of a grim corporate Panopticon. I guarantee you that if I've thought of it, so has someone else; I'm just trying to sound the alarm that the commentard community is underestimating the scope of the threat.
> Once Recall is deployed, getting it to log mouse actions and key presses should be relatively trivial.
Good! As that will also enable independent analysis showing just how much time is wasted navigating the bloated and inefficient GUI that MS foist on us in Windows and Office.
Windows has had the capability to spy on the device (not the individual user, the whole device) since Win7. It was available to corporations that use the Windows server shit-ware.
I've had one site where it was requested through corporate, by some investigative agency (?) to install and enabled it to log all activity and flush it to some hard coded address.
So none of this is new, it's just been made more public and people don't like what they are starting to finally see. The capabilities of the windows operating system to literally spy on anyone at any time has always been there.
Been out for a while. Hasn't been the end of the world or the new millennium.
At its core, Rewind builds upon Scribe, a meeting-recording bot with a user base of 30,000. It's not just a handy app; it’s a surveillance system for your memories, cataloguing every word and visual you encounter.
Is Rewind part of macos, is it on by default, and can the user easily disable it? The answers to these questions might form an important distinction between the Apple product and Recall. Rewind presumably doesn't require special hardware, either, so it must be doing something different. Finally, and critically, Apple users trust Apple way more than most sensible people trust Microsoft; whether that trust is justified is certainly up for debate.
What a crock of shite.
Recall is a solution in search of a problem. The author is trying to find a problem. The more obvious one would be, "How do I automate the process of stealing every bit of IP my company has?"
It isn't secure. It can't be made secure. If it could be, the number of data breaches would be declining. They are accelerating at an accelerating pace.
The only way to prevent data being stolen is to make sure it's never collected in the first place. The author needs to be collected and put into detox, until they have all of the Kool-Aid out of their system.
Recall is yet another perfect example of corporate america and their extremism in their never ending quest for more.
Sad thing is most of the time these examples of extremism actually lose money both directly or indirectly thru bad will and all the side effects they may cause.
The dream scenario of many AI / ML companies including Microsoft is NOT to allow SME to better analyse and understand their business process. That functionality, if effective, would offer some additional incentive for SME to pay the monthly subscription for Microsoft Copilot. If it would offer SME like 1% additional incentive to pay the monthly subscription for copilot, then it would to Microsoft be worth 1% of monthly Copilot rates minus the costs and overhead to implement it.
The dream scenario of many AI / ML companies including Microsoft is to be able to use ML and generative AI to first capture the low hanging fruit that could be automated and then create tools for it to sell (better said rent) to SME. Then apply the Apple Store / Google App Store / Steam Store business model and charge a 30% rate / tax for it. That 30% rate then means to charge 30% of the net cost savings the SME (and other businesses) renting the automation could have. If an SME for example can automate away 2 people, then the 30% tax equals to 30% of the cost to employ those 2 people. This includes their before tax wages, contributions to health insurance, pension fund, office space and equipment... In many industrialized countries that would be very roughly 5000 euros / dollars a month per removed employee. 2 removed employees then would net the AI / ML company or Microsoft 2 x 5000 x 30% / 100% = 3000 euros / dollars A MONTH forever.
=> What do you think Microsoft would prefer: giving SME a small extra incentive to use Copilot by offering the SME the ability to better analyse and improve their business process or offer the same SME a 10000 - 3000 Microsoft tax savings each month and collecting itself 3000 a month forever (while that inherently would require the SME to pay forever for Copilot for all personnel anyway)?
Think about it: with recall plus telemetry Microsoft has access to two things the SME don't have:
* keystrokes and mouse movements and clicks linked to these copilot screenshots
* access to Copilot plus "classic Windows telemetry / spyware data" in thousands up to millions of similar SME and companies worldwide
LLM / AI / ML training gets better and better (or less bad if you prefer that) by getting more and more data. Together with not having access to "classic Windows telemetry / spyware data", SME therefore are at a huge disadvantage to use their (and their workers!) data to create generative AI / ML solutions to automate part of their business process when not aided by Microsoft or other AI / ML companies.
=> Whether Microsoft will be able to pull it off or not is to be seen, but Recall is IMO part of Microsoft's gamble that they might in the future. They might not have a clue how to do it now, but in the meantime Recall plus "classic Windows telemetry / spyware data" stores as much as possible accessible data (at the cost of the customers / SME). If Microsoft finds a way in 5 five years to pull this off, the data will still be there for them to abuse. The SME still will be largely unable to put it to any meaningful use.
=> IF Microsoft (or other generative AI / ML companies) are able to pull that gamble off, they'll (divided over the dominant generative AI / ML companies) be able to grab around low double digit percentages *of the entire GDP of all industrialized nations*. THAT IMO is what this AI / ML gold rush is all about: the hope and dream to capture and monopolise double digit percentage OF THE ENTIRE global GDP. It's a bold (vampire corporate) dream and risky endeavour for which they are willing to invest heavily, risk losses, completely ignore risks, ethics, human values, privacy, national and international laws, the environment, the possibility it will create a grim dark world for most of the worlds population if successful and it's all worth it to them to take that gamble because so much seems to be up for the grabbing.
Will they pull it off? That's the big question. Could there be bubbles that burst and burst badly? Likely. Will that refrain them from trying again and again until they might succeed? Unlikely so long they see remaining chances to get there. Every burst bubble will create disappointments and reluctance to investors, but also new pieces of technology and insights to what worked or failed trying.
the number of employees a company amasses is directly linked to its economic power not necessarily its need for human labour. Once we have left the fields to work in an office we have become less of a production factor but more a political factor. Right at this moment billions of office worker are pondering on how to climb the hierarchy rather than to provide more value to the company. Microsoft products are perfect to let people mimic productive work. Endless hours are spent operating MS products without any actual value. Microsoft's recall will learn to perfectly mimic these useless interactions and will provide ample opportunities for companies to burn money even better as computer can burn money faster and 24/7 :D
I don't think that is what will happen.
First off, I belive Recall is such a thoroughly bad idea, with such enormous risks and liabilities baked right into it, that it could only have come from ono person at the very top of Microsoft*. A person that nobody dares go against.
That CEO-like person gets this idea that: "Maybe if we have AI capture all of the data going through a PC and also all the users responsed to that data, we can create a robot that can at least simulate users well enough to perform their daily tasks. Great!".
Then they set about getting The Organisation to buy this Fantastic, World Beating, Idea.
This is when they will encounter two things that are working well at Microsoft. Their smart developers, their talented product managers, and their legal team all piles in and explains why this is likely to end badly, what the possible risks are, and what the consequences could be. The other thing they encounter, is the balkanisation. There is no way, ever, that any department will build a working API that allows "The Competition" to leverage "Their Product" for their nefarious workings. If one makes them do it, they will totally half-ass it, and we get stuff like Microsoft Projectflow. This affects everyone at once so the CEO with the Bright Idea cannot on-board anyone to commit to this.
The CEO-like Solution becomes: "Fuck all of you guys! I'll have my very own team dump the screen buffer into a database and have the AI figure out stuff from the picures!!!. That's easy enough!"
Now, The "Microsoft Viva Experience"-team are probably the slime comitted to this idea: For them, it is just another sales feature to be able to better track the employees activities and to get more KPI's for the managers, also some illegal ones like sexual orientation, one suspects.
So, it will be sold as "Helping with automation". In reality it is Surveillance for Everyone.
*) Or a subversive consultant group, intent on destroying capitalism, that deliberaty only provides bad ideas and sell them to CEO's.
Reading your article, Initially I was not sure whether you are sarcastic, as the general perception about Microsoft & Windows is that, while for Apple is admiration as in the new Calc App in 2024. That aside, I can relate to what you are able to see with Recall for RPA use-cases. I have always been saying, that as Gen Alpha (and even Millenials) come in numbers to workforce, they will refuse to work with the current set of dumb enterprises applications which have no memory or personalization about their usage - this is a generation that is used to their Netflix, Spotify, TikTok and Instagram surfacing content even before they knew they want to see it, where as current enterprise applications force the users to repeat the menu navigation every single time, even though that is the job description for the role.
I hope Microsoft sees the potential for RPA usecases, especially with the power of LLMs and their magical 'agent' calling mechanisms.
Microsoft claims that "StepsRecorder" has been depreciated. Yet it was just installed on my OS.
https://www.google.com/search?q=steps+recorder+windows+11&sca_esv=4845ad2973331627&source=hp&ei=1sRrZq7pMsSe0PEP_Nan2AU&iflsig=AL9hbdgAAAAAZmvS5vtvLTe7bbEBwVcBK6ufIDwvd0Gb&oq=StepsRecorder&gs_lp=Egdnd3Mtd2l6Ig1TdGVwc1JlY29yZGVyKgIIAjINEAAYgAQYsQMYgwEYCjIHEAAYgAQYCjIHEAAYgAQYCjIHEAAYgAQYCjIHEAAYgAQYCjIHEAAYgAQYCjIHEAAYgAQYCjIHEAAYgAQYCjIHEAAYgAQYCjIHEAAYgAQYCkiOFFAAWABwAHgAkAEAmAFFoAFFqgEBMbgBAcgBAPgBAvgBAZgCAaACU5gDAJIHATGgB-kI&sclient=gws-wiz
What do you think!
Business leaders (especially those of the mindless MBA beancounting sect) care nothing about increasing productivity unless that allows for a reduction in headcount. Growth is unlikely to to show in one or two quarters wereas laying off 10% of your employees will. The reduction in quality won't show for a while either and even then all you do is redefine the minimum viable product and if your customers have nowhere else to go they'll suck it up anyway.
It all seems quite dystopian right now.
The reason why PC's became ubiquitous was that they allowed small business to get away from the mainframe model which was both highly constraining and comparatively expensive. The problem is that buying something generates just one unit of revenue which just doesn't work -- from the very beginning of automated data processing Herman Hollerith figured out that the only way to truly generate revenue was by leasing.
So the big providers are doing just that. But they've figured out how to lease something abstract so you don't even get a nicely polished wooden box (or painted steel cabinet) for your money. Obviously the ground will first get seeded with low cost corn to lure the marks in but once in the net closes and they're on that treadmill for life.
Another use-case, for individuals, would be to document something done recently (or maybe not so recently). Around 6 months ago, I set up a Jellyfin server at home as an LXC container in my proxmox cluster, and I passed the graphics of the AMD APU of the node through to the container to enable hardware transcoding. I had to research online as to how to do this. It took me most of an evening to get it working, and I told myself I would go back the next day to document my process. I never did this, and if I had to perform the passthrough again, I would have to research it all over again. Recall could conceivably document this for me, after the fact. Of course I was using Linux, not Windows, and I would never have Recall enabled on purpose if I was using Windows, so it's a mute question in my case...
Dear Microsoft
You do realise you could be shooting yourself in the foot here?
If the beancounters were able to see that users spend far too much time recovering from operating system gotchas, such as updates, then they might (should) ask questions, such as "Are there systems out there that require less maintenance?"
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
