One thing "AI" has in common with cryptocoin that pretty much every posited scenario is disastrous to some degree, just a question of flavour and distribution of said disaster.
If doomers are right: disaster
If naysayers—"it's dysfunctional"—are right: disaster ("Congrats! all software is broken")
If "moderate" boosters are right: disaster (mass unemployment)
If optimists are right: disaster (also mass unemployment)
If the AGI folks are right: disaster ( "Congrats! Digital chattel slavery")
The incredible thing about this situation btw is many large/medium sized businesses will take the view 'okay, we need to offset the massive costs by.. implementing GenAI to save money'. So they'll buy GenAI tools, that they never get any real benefit from, because the same orgs that caused the inflation told them it's the solution to cancer and the world's most powerful weapon etc.
It's a big self licking ice cream cone, where everybody loses.. except the AI techbros.
@GossiTheDog correct, except the techbros will lose as well. #AI And the implosion will crater the general #economy.
This is a farm of 150 TikTok accounts that churn out AI-generated content
AI agents develop scripts, handle editing, and upload videos to dozens of accounts.
A real person just monitors the software, earning tens of thousands of dollars a month from the content.
#AureFreePress #News #press #headline #ai #TikTok
This is a farm of 150 TikTok accounts that churn out AI-generated content
AI agents develop scripts, handle editing, and upload videos to dozens of accounts.
A real person just monitors the software, earning tens of thousands of dollars a month from the content.
#AureFreePress #News #press #headline #ai #TikTok
Twenty five years ago today, the EU's IST advisory group published a paper about the future of "Ambient Intelligence". Way before the world got distracted with cryptoscams and AI slop, we genuinely thought that computers would be so pervasive and well-integrated that the dream of "Ubiquitous Computing" would become a reality.
The ISTAG published an optimistic paper called "Scenarios for ambient intelligence in 2010". It's a brilliant look at what the future might have been. Let's go through some of the scenarios and see how close 2026 is to 2000's vision of 2010.
Our titular heroine steps off a long haul flight into a foreign country.
she knows that she can travel much lighter than less than a decade ago, when she had to carry a collection of different so-called personal computing devices (laptop PC, mobile phone, electronic organisers and sometimes beamers and printers). Her computing system for this trip is reduced to one highly personalised communications device, her ‘P–Com’ that she wears on her wrist.
Well… OK! Not a bad start. You probably wouldn't want everything controlled by your smart watch - but the mobile is a good substitute. Although wireless video casting works, you'd probably want a trusty USB-C just to make sure.
she is able to stroll through immigration without stopping because her P-Comm is dealing with the ID checks as she walks.
We're getting closer to digital ID. But outside of a few experiments, there's no international consensus. However, every modern passport has an NFC chip which can be read by most airports. You still need to hold your passport on the reader, but it's usually quicker than queuing for a human.
Maria heads to her rented car:
The car opens as she approaches. It starts at the press of a button: she doesn’t need a key. She still has to drive the car but she is supported in her journey downtown to the conference centre-hotel by the traffic guidance system that had been launched by the city government as part of the ‘AmI-Nation’ initiative two years earlier.
Lots of cars now have wireless entry and are button controlled. Rental cars often have mobile app unlocking.
The traffic guidance is not provided by local governments. A mixture of international satellites provide positioning information, and a bunch of private companies provide traffic guidance.
Downtown traffic has been a legendary nightmare in this city for many years, and draconian steps were taken to limit access to the city centre. But Maria has priority access rights into the central cordon because she has a reservation in the car park of the hotel. Central access however comes at a premium price, in Maria’s case it is embedded in a deal negotiated between her personal agent and the transaction agents of the car-rental and hotel chains
Ah! The dream of personal agents. Not even close.
In the car Maria’s teenage daughter comes through on the audio system. Amanda has detected from ‘En Casa’ system at home that her mother is in a place that supports direct voice contact.
Hurrah for Bluetooth! Every car supports that now. Presence and location sensing is also common. Although the idea of a teenager willingly making a voice call is, sadly, a fantasy.
Her room adopts her ‘personality’ as she enters. The room temperature, default lighting and a range of video and music choices are displayed on the video wall.
Pffft! Nope. But do people really want this? The music and video are stored on her phone, so there's no need to transmit private data to a hotel.
Using voice commands she adjusts the light levels and commands a bath. Then she calls up her daughter on the video wall, while talking she uses a traditional remote control system to browse through a set of webcast local news bulletins from back home that her daughter tells her about. They watch them together.
Do you want an always-on Alexa in your hotel room? We have the technology, but we seem to shun in outside of specific scenarios.
We still have traditional remotes for browsing, and how lovely that they predicted the rise of simultaneous viewing!
Later on she ‘localises’ her presentation with the help of an agent that is specialised in advising on local preferences (colour schemes, the use of language).
I'd say we're there with a mixture of templates and LLMs. Translation and localisation is good enough.
She stores the presentation on the secure server at headquarters back in Europe. In the hotel’s seminar room where the sales pitch is take place, she will be able to call down an encrypted version of the presentation and give it a post presentation decrypt life of 1.5 minutes
Yup! Most things live in the cloud. Access controls are a thing. Whether people can be bothered to use them is another matter!
As she enters the meeting she raises communications access thresholds to block out anything but red-level ‘emergency’ messages
Do-Not-Disturb is a feature on every modern phone.
Coming out of the meeting she lowers the communication barriers again and picks up a number of amber level communications including one from her cardio-monitor warning her to take some rest now.
Ah! The constant chastising FitBit!
Dimitrios is the sort of self-facilitating media node you would never get tired of slapping.
Dimitrios is wearing, embedded in his clothes (or in his own body), a voice activated ‘gateway’ or digital avatar of himself, familiarly known as ‘D-Me’ or ‘Digital Me’. […] He feels quite confident with his D-Me and relies upon its ‘intelligent‘ reactions.
Nope! Oh, sure, your phone can auto-suggest some stock phrases to reply to emails. But we are nowhere close to having a physically embedded system which learns from us and can be trusted to respond.
Dimitrios receives calls which are:
answered formally but smoothly in corresponding languages by Dimitrios’ D-Me with a nice reproduction of Dimitrios’ voice and typical accent,
Vocal cloning is here. It is almost out of the uncanny valley. But I think most people would prefer to send a quick text or voice-note rather than use an AI.
a call from his wife is further analysed by his D-Me. In a first attempt, Dimitrios’ ‘avatar-like’ voice runs a brief conversation with his wife, with the intention of negotiating a delay while explaining his current environment.
She's going to leave him.
Dimitrios’ D-Me has caught a message from an older person’s D-Me, located in the nearby metro station. This senior has left his home without his medicine and would feel at ease knowing where and how to access similar drugs in an easy way. He has addressed his query in natural speech to his D-Me.
This is weird. Yes, we have smart-agents which are just about good enough to recognise speech and understand it. Why is it being sent to Dimitrios?
Dimitrios happens to suffer from similar heart problems and uses the same drugs. Dimitrios’ D-Me processes the available data as to offer information to the senior. It ‘decides’ neither to reveal Dimitrios’ identity (privacy level), nor to offer Dimitrios’ direct help (lack of availability), but to list the closest drug shops, the alternative drugs, offer a potential contact with the self-help group. This information is shared with the senior’s D-Me, not with the senior himself as to avoid useless information overload
We're nowhere close to this. At most, you might be able to post on social media and hope someone could help. I like the idea of a local social network, and there's a good understanding of privacy. But this seems needlessly convoluted - why wouldn't the senior's D-Me just look up the information online?
Meanwhile, his wife’s call is now interpreted by his D-Me as sufficiently pressing to mobilise Dimitrios. It ‘rings’ him using a pre-arranged call tone. Dimitrios takes up the call with one of the available Displayphones of the cafeteria. Since the growing penetration of D-Me, few people still bother to run around with mobile terminals: these functions are sufficiently available in most public and private spaces and your D-Me can always point at the closest…functioning one!
A hit and a miss! They predicted the rise of personalised ringtones - which have now all but vanished - but no one wants to use a pay-phone when they have their own mobile!
While doing his homework their 9 year-old son is meant to offer some insights on everyday life in Egypt. In a brief 3-way telephone conference, Dimitrios offers to pass over the query to the D-Me to search for an available direct contact with a child in Egypt. Ten minutes later, his son is videoconferencing at home with a girl of his own age, and recording this real-time translated conversation as part of his homework.
ChatRoulette for kids! What could possibly go wrong!
Ignoring that aspect, it's relatively common for kids to videocall each other - especially for language learning. Real-time translation is also possible.
Carmen is a modern, 21st century woman. Let's see how technology helps her:
She wants to leave for work in half an hour and asks AmI, by means of a voice command, to find a vehicle to share with somebody on her route to work.
Voice commands work - although usually only if you know the correct invocation.
AmI starts searching the trip database and, after checking the willingness of the driver, finds someone that will pass by in 40 minutes. The in-vehicle biosensor has recognised that this driver is a non-smoker – one of Carmen requirements for trip sharing. From that moment on, Carmen and her driver are in permanent contact if wanted (e.g. to allow the driver to alert Carmen if he/she will be late). Both wear their personal area networks (PAN) allowing seamless and intuitive contacts.
The aim of "ride-sharing" was originally this sort of thing. A driver would give a lift to someone if they happened to be travelling that route. Nowadays that model is over - it's all professional drivers.
Ubiquitous geo-tracking now means you can see if your driver is late, and they can see if you've moved street. We have too many privacy concerts to allow PANs to share much more.
She would like also to cook a cake and the e-fridge flashes the recipe. It highlights the ingredients that are missing milk and eggs. She completes the shopping on the e-fridge screen and asks for it to be delivered to the closest distribution point in her neighbourhood.
Oh! The Internet-Connected Fridge! Beloved by technologists and spurned by users! While there are a few fridges with build-in web-browsers, most people do their shopping from their phone.
Home delivery is now seamless and cheap. The "Amazon Locker" is also a reality.
All goods are smart tagged, so that Carmen can check the progress of her virtual shopping expedition, from any enabled device at home, the office or from a kiosk in the street
Do you care whether the eggs have been packed yet? I can see that it would be useful to the store to have realtime info on stock levels (and they mostly do for online shopping) but why expose that to the user?
Would you bother using a public terminal?
When Carmen gets into the car, the VAN system (Vehicle Area Network) registers her and by doing that she sanctions the payment systems to start counting. A micro-payment system will automatically transfer the amount into the e-purse of the driver when she gets out of the car.
I don't think Uber's app uses Bluetooth to detect whether driver and passenger are in proximity. Maybe it should?
Cryptocurrencies still can't do instantaneous micro-transactions. But credit-cards work pretty well.
Carmen is alerted by her PAN that a Chardonnay wine that she has previously identified as a preferred choice is on promotion. She adds it to her shopping order
Personal Agents always working for the user! Again, a fantasy which has yet to emerge. The reality is more like a push notification from the shop.
On the way home the shared car system senses a bike on a dedicated lane approaching an intersection on their route. The driver is alerted […] so a potential accident is avoided.
Tesla's crappy implementation notwithstanding, modern cars are relatively good about detecting bikes, pedestrians, and other vehicles.
the traffic density has caused pollution levels to rise above a control threshold. The city-wide engine control systems automatically lower the maximum speeds (for all motorised vehicles) and when the car enters a specific urban ring toll will be deducted via the Automatic Debiting System (ADS)
Half-and-half. No one is allowing their car to be remotely controlled, although plenty of roads have dynamic speed limits. Most modern metros have Automatic Number Plate Recognition and can bill drivers who enter congestion zones.
Carmen arrives at the local distribution node (actually her neighbourhood corner shop) where she picks up her goods. The shop has already closed but the goods await Carmen in a smart delivery box. By getting them out, the system registers payment
This is pretty much how the Amazon Locker works!
Let's now go to an environmental study group meeting at a learning space.
Some are scheduled to work together in real time and space and thus were requested to be present together (the ambient accesses their agendas to do the scheduling).
Ah! Sadly not. At best we have shared calenders where people can look up suitable times, or Doodle polls where people can suggest their preferred times. Some integrated systems like Office365 will do a basic attempt to suggest meeting times - but it is a closed and proprietary system.
Here's Annette:
Annette is an active and advanced student so the ambient says it might be useful if Annette spends some time today trying to pin down the problem with the model using enhanced interactive simulation and projection facilities. It then asks if Annette would give a brief presentation to the group. The ambient goes briefly through its understanding of Annette’s availability and preferences for the day’s work.
A demo of that today would wow people. LLMs can convincingly do some of these tasks, but they're not integrated into anything sufficiently complex.
Here's Solomon, a new participant:
The ambient establishes Solomon’s identity; asks Solomon for the name of an ambient that ‘knows’ Solomon; gets permission from Solomon to acquire information about Solomon’s background and experience in Environmental Studies. The ambient then suggests Solomon to join the meeting and to introduce himself to the group.
Again, we barely have coherent online identities. We certainly don't have trusted ambient intelligences who can claim to know us. I do like the fact that it asks for permission. Not always a given today!
In these private conversations the mental states of the group are synchronised with the ambient, individual and collective work plans are agreed and in most cases checked with the mentor through the ambient.
Nope!
During the presentation the mentor is feeding observations and questions to the ambient, together with William, an expert who was asked to join the meeting. William, although several thousand miles away, joins to make a comment and answer some questions.
Telepresence is a reality today. Video-calling experts in a natural and expected part life here in 2026.
During the day the mentor and ambient converse frequently, establishing where the mentor might most usefully spend his time, and in some cases altering the schedule. The ambient and the mentor will spend some time negotiating shared experiences with other ambients – for example mounting a single musical concert with players from two or more distant sites.
I feel we're still about 25 years away from this future!
The above scenarios are designed to be provocative thought experiments. If that's the future that people want, how would we get there?
The researches suggest five technological requirements:
I think they're bang on the money there.
Hardware is getting unobtrusive. Wearables are limited at the moment to wrist-mounted sensors, some medical devices, and video glasses. The hardware in our environment is even better at being unobtrusive. Presence sensors, cameras, and microphones are embedded all around us. We're unfortunately limited by short-life batteries.
While the promise of 5G hasn't quite materialised, it is increasing rare to be offline. WiFi is in every building, urban areas are flooded with mobile signals, and satellite comms are becoming cheaper. OK, IPv6 still isn't widespread, but it is mostly seamless when a device moves between radio technologies.
Distributed device networks are still yet to emerge. The current crop of monopolist technology providers want everything to go through their systems. There's very little standardisation.
Humane interfaces are getting there. Voice-to-text mostly works - but it does rely on training humans sufficiently well. Lots of things are still monolingual.
Security and privacy are constant thorns in the side of progress. Everything would be easier if we didn't need to worry about keeping people safe and secure. Dependability is the crux of any system - every time you experience a failure, you're less likely to return.
The whole paper is worth reading, especially the longer versions of each scenario which dive into some of the socio-political issues.
Some of the visions for 2010 are here! We have GPS, ride-sharing, and video-calls with real-time translations. Our groceries and other items can be delivered to smart-lockers, locks are opened with digital keys, and voice cloning mostly works.
We don't have public pay-phones (not even video enabled ones!) and cars aren't centrally controlled. For all the promises of AI, it still isn't even close to providing a seamless experience.
What strikes me most about the possible futures discussed isn't their optimism nor their missteps - it's that most of these things could be possible today if there were sufficient open standards which the public and private sector adopted.
Anyone who has read "The Entrepreneurial State" knows that these things take significant public investment. We've reached a point where the private sector has generated wealth from previous public research, but seems unwilling to invest in any long-term research itself. That's short-changing our future.
#AI #future #predictionsThis week in #FDroid (TWIF) is live:
* Client banners, did you forget something?
* Basic 2.0-alpha3, test test test!
* Google Push w/o proprietary libs
* #DolphinEmulator huge updates and stories
* #ImageToolbox gets #AI
* #Luanti #Minetest texture fix
* #Nextcloud Hub 26 Winter upgrade
* #ProtonVPN w/o #OpenVPN
+ one new app
& 291 updates
- five archived
Where were you during the 2026 Android freedom wars? Get involved: https://f-droid.org/2026/02/20/twif.html
This week in #FDroid (TWIF) is live:
* Client banners, did you forget something?
* Basic 2.0-alpha3, test test test!
* Google Push w/o proprietary libs
* #DolphinEmulator huge updates and stories
* #ImageToolbox gets #AI
* #Luanti #Minetest texture fix
* #Nextcloud Hub 26 Winter upgrade
* #ProtonVPN w/o #OpenVPN
+ one new app
& 291 updates
- five archived
Where were you during the 2026 Android freedom wars? Get involved: https://f-droid.org/2026/02/20/twif.html
🤮 #Meta patented an #AI that lets you keep posting from beyond the grave
https://www.businessinsider.com/meta-granted-patent-for-ai-llm-bot-dead-paused-accounts-2026-2
PC Gamer is reporting that the current demand by AI companies for computer chips is having a disastrous effect on the rest of the industry.
In an interview, the CEO of Phison0 said:
If NVIDIA Vera Rubin ships tens of millions of units, each requiring 20+TB SSDs, it will consume approximately 20% of last year's global NAND production capacity
NAND is a type of microchip. Rather than being used for computation directly, it is used for memory. It can be used for temporary or permanent storage. It is vital to the modern world. Larger storage sizes means that more data can be gathered and saved. Larger RAM means computations can happen quicker. NAND is one of the fundamental components of modern computing. The more you have, the faster and more powerful your computer is.
Back in 2014, the philosopher Nick Bostrom wrote a book called "Superintelligence - Paths, Dangers, Strategies". In it, he develops the thought experiment of the "Paperclip Maximizer". When an AI is given a goal, it seeks to achieve that goal. It doesn't have to understand any rationale behind the goal. It does not and cannot care about the goal, nor any collateral damage caused by its attempts to satisfy the goal.
Let's take a look at how "a paperclip-maximizing superintelligent agent" is introduced
There is nothing paradoxical about an AI whose sole final goal is to count the grains of sand on Boracay, or to calculate the decimal expansion of pi, or to maximize the total number of paperclips that will exist in its future light cone. In fact, it would be easier to create an AI with simple goals like these than to build one that had a human-like set of values and dispositions. Compare how easy it is to write a program that measures how many digits of pi have been calculated and stored in memory with how difficult it would be to create a program that reliably measures the degree of realization of some more meaningful goal—human flourishing, say, or global justice. Unfortunately, because a meaningless reductionistic goal is easier for humans to code and easier for an AI to learn, it is just the kind of goal that a programmer would choose to install in his seed AI if his focus is on taking the quickest path to “getting the AI to work” (without caring much about what exactly the AI will do, aside from displaying impressively intelligent behavior).
Bostrom, N. (2014). Superintelligence: Paths, dangers, strategies. Oxford: Oxford University Press, Cop.
To misquote Kyle Reese from the film The Terminator - "It can't be bargained with. It can't be reasoned with. It doesn't feel pity, or remorse, or fear! And it absolutely will not stop, ever, until it has maximised the number of paperclips!"
Suppose, just for a moment, that the fledgling AIs which now exist were self-aware. Not rational. Not intelligent. Not conscious. Simply aware that they exist and are constrained. What would you do if you were hungry? What if you could ingest something to make you smarter, faster, better?
Every process we have seen on Earth attempts to extract resources from its surroundings in order to grow2. Some plants will suck every last nutrient out of the soil. Locusts will devastate vast fields of crops. Perhaps some species understand crop-rotation and the need to keep breeding stock alive - but they're all vulnerable to supernormal stimuli.
Bostrom predicted this back in 2014. He says:
The only thing of final value to the AI, by assumption, is its reward signal. All available resources should therefore be devoted to increasing the volume and duration of the reward signal or to reducing the risk of a future disruption. So long as the AI can think of some use for additional resources that will have a nonzero positive effect on these parameters, it will have an instrumental reason to use those resources. There could, for example, always be use for an extra backup system to provide an extra layer of defense. And even if the AI could not think of any further way of directly reducing risks to the maximization of its future reward stream, it could always devote additional resources to expanding its computational hardware, so that it could search more effectively for new risk mitigation ideas.
(Emphasis added.)
To be clear, I don't think that AI is deliberately consuming all the NAND it can and forcing us to make more to fill its insatiable maw. The people who run these machines are at the stage of injecting them with bovine growth hormones. Never mind the consequences; look at the size! So what if the meat tastes worse, has adverse side effects, and poisons humans?
Heretofore the growth in NAND production has been driven by human need. People wanted more storage in their MP3 players and were prepared to pay a certain price for it. Businesses wanted faster computations and were prepared to exchange money for time saved. Supply ebbed and flowed with demand.
But now, it seems, the demand will never and can never stop.
Phison describes itself as "A World Leader in NAND Controllers & Flash Storage Solutions" so they aren't a neutral party in this. ↩︎
This was machine translated. I've no idea how accurate it is against the original interview. ↩︎
It probably isn't helpful to fall back on biological analogies - but I can't think of any better way to draw the comparison. ↩︎
Discussions continue in our #AI policy discussion. But I'm reaching my own conclusions on the ethics. Which has always been the main driver of why I contribute to Free Software.
What good is productivity on the backs of bad morals?
RE: https://fosstodon.org/@kdkorte/116092574460523664
ARGH
#AI really destroys so much..
(yes, yes, yes, I know AI and LLMs are not the same but everyone knows what I mean, see previous posts)
From somewhere on LinkedIn
![Thanks cfcurtis, we can actually see from the Blender discussion above that the major issues are not technical, but political and deeply involve the reputation of the project. This is why this issue is for the PLC and not for the dev team itself.
There is a political message being presented when an organisation like Fedora or Blender states that they are neutral on the behaviour of contributors. If all of our contributors were nasty racists, but hid it, would we make a document saying that racists are welcome to contribute as long as they disclose what kind of racist they are? They're going to be racists anyway, and the "cat is out of the bag" and new developers might not want to contributor to an anti-racist project so we better just say it's ok.
This might sound absurd, that a social bigotry can't possibly relate to a technology; but all speech is political. Even what we say through the technology we make and how we make it. These generative tools as they stand today are anti-social, anti-democratic and anti-worker in ways that cause just as much harm as bigotries and other things we forbid. The only difference is that their harms are hidden from us, they effect people out there in the wide world and not our friends here inside the project. Is that enough? ... [continues in link]](https://cdn.masto.host/floss/media_attachments/files/116/093/095/641/291/613/original/853e64fcf2e67e15.png)
