Comment Bitwarden Staff on Reddit (Score 5, Informative) 70
Bitwarden staff on Reddit say it has been restored.
https://www.reddit.com/r/Bitwa...
Submission + - Apple has had over a decade to fix this iOS email copy bug, and it still hasn (nerds.xyz)
BrianFagioli writes: Apple has a maddening bug that has persisted across well over a decade of iOS releases. When you long-press an email address on iPhone or iPad and choose âoeCopy Email Address,â what ends up on your clipboard is not just the email address — it is âoemailto:someone@example.com.â A URI prefix you did not ask for, tacked onto something the menu explicitly told you it was copying correctly. Paste it somewhere that does not strip it automatically and you have a problem.
The astonishing part is not the bug itself — it is that Apple has never fixed it. This is a company that has overhauled its chip architecture, shipped new programming languages, and added satellite connectivity to a phone. But âoeCopy Email Addressâ still does not copy just the email address. The fix is trivially simple: strip the prefix when the user picks that option. Apple knows this is broken. At some point, leaving it unfixed stops being an oversight and starts being a choice.ââââââââââââââââ
The astonishing part is not the bug itself — it is that Apple has never fixed it. This is a company that has overhauled its chip architecture, shipped new programming languages, and added satellite connectivity to a phone. But âoeCopy Email Addressâ still does not copy just the email address. The fix is trivially simple: strip the prefix when the user picks that option. Apple knows this is broken. At some point, leaving it unfixed stops being an oversight and starts being a choice.ââââââââââââââââ
Comment I kinda love this (Score 1) 41
I have purchased a lot of kindle books and audibles. Will it give me a refresher after not reading a series in awhile or even after I dropped reading mid book? Can it elucidate ideas I am unfamiliar with without me needing to pull out a browser or ChatGPT?
Comment Re: No proton for me (Score 1) 30
Why not use simplelogin or another alias system?
Submission + - Proton Mail Suspended Journalist Accounts at Request of Cybersecurity Agency (theintercept.com)
An anonymous reader writes: The company behind the Proton Mail email service, Proton, describes itself as a “neutral and safe haven for your personal data, committed to defending your freedom.” But last month, Proton disabled email accounts belonging to journalists reporting on security breaches of various South Korean government computer systems following a complaint by an unspecified cybersecurity agency. After a public outcry, and multiple weeks, the journalists’ accounts were eventually reinstated — but the reporters and editors involved still want answers on how and why Proton decided to shut down the accounts in the first place.
Martin Shelton, deputy director of digital security at the Freedom of the Press Foundation, highlighted that numerous newsrooms use Proton’s services as alternatives to something like Gmail “specifically to avoid situations like this,” pointing out that “While it’s good to see that Proton is reconsidering account suspensions, journalists are among the users who need these and similar tools most.” Newsrooms like The Intercept, the Boston Globe, and the Tampa Bay Times all rely on Proton Mail for emailed tip submissions. Shelton noted that perhaps Proton should “prioritize responding to journalists about account suspensions privately, rather than when they go viral.” On Reddit, Proton’s official account stated that “Proton did not knowingly block journalists’ email accounts” and that the “situation has unfortunately been blown out of proportion."
The two journalists whose accounts were disabled were working on an article published in the August issue of the long-running hacker zine Phrack. The story described how a sophisticated hacking operation — what’s known in cybersecurity parlance as an APT, or advanced persistent threat — had wormed its way into a number of South Korean computer networks, including those of the Ministry of Foreign Affairs and the military Defense Counterintelligence Command, or DCC. The journalists, who published their story under the names Saber and cyb0rg, describe the hack as being consistent with the work of Kimsuky, a notorious North Korean state-backed APT sanctioned by the U.S. Treasury Department in 2023. As they pieced the story together, emails viewed by The Intercept show that the authors followed cybersecurity best practices and conducted what’s known as responsible disclosure: notifying affected parties that a vulnerability has been discovered in their systems prior to publicizing the incident.
Martin Shelton, deputy director of digital security at the Freedom of the Press Foundation, highlighted that numerous newsrooms use Proton’s services as alternatives to something like Gmail “specifically to avoid situations like this,” pointing out that “While it’s good to see that Proton is reconsidering account suspensions, journalists are among the users who need these and similar tools most.” Newsrooms like The Intercept, the Boston Globe, and the Tampa Bay Times all rely on Proton Mail for emailed tip submissions. Shelton noted that perhaps Proton should “prioritize responding to journalists about account suspensions privately, rather than when they go viral.” On Reddit, Proton’s official account stated that “Proton did not knowingly block journalists’ email accounts” and that the “situation has unfortunately been blown out of proportion."
The two journalists whose accounts were disabled were working on an article published in the August issue of the long-running hacker zine Phrack. The story described how a sophisticated hacking operation — what’s known in cybersecurity parlance as an APT, or advanced persistent threat — had wormed its way into a number of South Korean computer networks, including those of the Ministry of Foreign Affairs and the military Defense Counterintelligence Command, or DCC. The journalists, who published their story under the names Saber and cyb0rg, describe the hack as being consistent with the work of Kimsuky, a notorious North Korean state-backed APT sanctioned by the U.S. Treasury Department in 2023. As they pieced the story together, emails viewed by The Intercept show that the authors followed cybersecurity best practices and conducted what’s known as responsible disclosure: notifying affected parties that a vulnerability has been discovered in their systems prior to publicizing the incident.
Comment 45 million (Score 1) 1
Is Israel spending money to build up Palestine? Iâ(TM)m guessing not or else that would be the narrative being pushed. That would be better propaganda :(
Submission + - Google's $45 Million Contract With Netanyahu's Office to Spread Israeli Propagan (dropsitenews.com) 1
Alain Williams writes: Publicly available government contracts show that Israel’s advertising bureau, which reports to the prime minister’s office, has since embarked on a mass advertising and public messaging effort to conceal the hunger crisis. The push includes the use of American influencers widely reported on last month. It also includes a high-dollar spending spree on paid advertising, yielding tens of millions for Google, YouTube, X, Meta, and other tech platforms.
“There is food in Gaza. Any other claim is a lie,” asserted a propaganda video published by Israel’s foreign ministry to Google’s YouTube video sharing platform in late August and viewed more than 6 million times. Much of the video’s reach results from an ad placed during an ongoing and previously unreported $45 million (NIS 150 million) advertising campaign initiated between Google and Netanyahu’s office in late June. The contract—which is with both YouTube and Google's advertising campaign management platform, Display & Video 360—explicitly characterizes the ad campaign as hasbara, a Hebrew word whose meaning is somewhere between public relations and propaganda.
Records show that the Israeli government similarly spent $3 million (NIS 10 million) for an advertising campaign with X. The French and Israeli advertising platform Outbrain/Teads is also set to receive roughly $2.1 million (NIS 7 million).
The existence of an Israeli Google ads campaign to discredit the UN’s primary aid agency for Palestinian refugees, UNRWA, was similarly reported by WIRED last year. Hadas Maimon, head of public awareness for Israel’s diaspora ministry, stated during the March 2 Knesset hearing that, “For almost a year now, we have been leading a major campaign on the issue of UNRWA.”
“There is food in Gaza. Any other claim is a lie,” asserted a propaganda video published by Israel’s foreign ministry to Google’s YouTube video sharing platform in late August and viewed more than 6 million times. Much of the video’s reach results from an ad placed during an ongoing and previously unreported $45 million (NIS 150 million) advertising campaign initiated between Google and Netanyahu’s office in late June. The contract—which is with both YouTube and Google's advertising campaign management platform, Display & Video 360—explicitly characterizes the ad campaign as hasbara, a Hebrew word whose meaning is somewhere between public relations and propaganda.
Records show that the Israeli government similarly spent $3 million (NIS 10 million) for an advertising campaign with X. The French and Israeli advertising platform Outbrain/Teads is also set to receive roughly $2.1 million (NIS 7 million).
The existence of an Israeli Google ads campaign to discredit the UN’s primary aid agency for Palestinian refugees, UNRWA, was similarly reported by WIRED last year. Hadas Maimon, head of public awareness for Israel’s diaspora ministry, stated during the March 2 Knesset hearing that, “For almost a year now, we have been leading a major campaign on the issue of UNRWA.”
Submission + - New AI Model Turns Photos Into Explorable 3D Worlds, With Caveats (arstechnica.com)
An anonymous reader writes: On Tuesday, Tencent released HunyuanWorld-Voyager, a new open-weights AI model that generates 3D-consistent video sequences from a single image, allowing users to pilot a camera path to "explore" virtual scenes. The model simultaneously generates RGB video and depth information to enable direct 3D reconstruction without the need for traditional modeling techniques. However, it won't be replacing video games anytime soon.
The results aren't true 3D models, but they achieve a similar effect: The AI tool generates 2D video frames that maintain spatial consistency as if a camera were moving through a real 3D space. Each generation produces just 49 frames — roughly two seconds of video — though multiple clips can be chained together for sequences lasting "several minutes," according to Tencent. Objects stay in the same relative positions when the camera moves around them, and the perspective changes correctly as you would expect in a real 3D environment. While the output is video with depth maps rather than true 3D models, this information can be converted into 3D point clouds for reconstruction purposes.
The results aren't true 3D models, but they achieve a similar effect: The AI tool generates 2D video frames that maintain spatial consistency as if a camera were moving through a real 3D space. Each generation produces just 49 frames — roughly two seconds of video — though multiple clips can be chained together for sequences lasting "several minutes," according to Tencent. Objects stay in the same relative positions when the camera moves around them, and the perspective changes correctly as you would expect in a real 3D environment. While the output is video with depth maps rather than true 3D models, this information can be converted into 3D point clouds for reconstruction purposes.
Submission + - AI Generated 'Boring History' Videos Are Flooding YouTube (404media.co)
An anonymous reader writes: As I do most nights, I was listening to YouTube videos to fall asleep the other night. Sometime around 3 a.m., I woke up because the video YouTube was autoplaying started going “FEEEEEEEE.” The video was called “Boring History for Sleep | How Medieval PEASANTS Survived the Coldest Nights and more.” It is two hours long, has 2.3 million views, and, an hour and 15 minutes into the video, the AI-generated voice glitched. “In the end, Anne Boleyn won a kind of immortality. Not through her survival, but through her indelible impact on history. FEEEEEEEEEEEEEEEE,” the narrator says in a fake British accent. “By the early 1770s, the American colonies simmered like a pot left too long over a roaring fire,” it continued. The video was from a channel I hadn’t seen before, called “Sleepless Historian.” I took my headphones out, didn’t think much of it at the time, rolled over, and fell back asleep.
The next night, when I went to pick a new video to fall asleep to, my YouTube homepage was full of videos from Sleepless Historian and several similar-sounding channels like Boring History Bites, History Before Sleep, The Snoozetorian, Historian Sleepy, and Dreamoria. Lots of these videos nominally check the boxes for what I want from something to fall asleep to. Almost all of them are more than three hours long, and they are about things I don’t know much about. Some video titles include “Unusual Medieval Cures for Common Illnesses,” “The Entire History of the American Frontier,” “What It Was Like to Visit a BR0THEL in Pompeii,” and “What GETTING WASTED Was Like in Medieval Times.” One of the channels has even been livestreaming this "history" 24/7 for weeks.
In the daytime, when I was not groggy and half asleep, it quickly became obvious to me that all of these videos are AI generated, and that they are part of a sophisticated and growing AI slop content ecosystem that is flooding YouTube, is drowning out human-made content created by real anthropologists and historians who spend weeks or months researching, fact-checking, scripting, recording, and editing their videos, and are quite literally rewriting history with surface-level, automated drek that the YouTube algorithm delivers to people. YouTube has said it will demonetize or otherwise crack down on “mass produced” videos, but it is not clear whether that has had any sort of impact on the proliferation of AI-generated videos on the platform, and none of the people I spoke to for this article have noticed any change.
The next night, when I went to pick a new video to fall asleep to, my YouTube homepage was full of videos from Sleepless Historian and several similar-sounding channels like Boring History Bites, History Before Sleep, The Snoozetorian, Historian Sleepy, and Dreamoria. Lots of these videos nominally check the boxes for what I want from something to fall asleep to. Almost all of them are more than three hours long, and they are about things I don’t know much about. Some video titles include “Unusual Medieval Cures for Common Illnesses,” “The Entire History of the American Frontier,” “What It Was Like to Visit a BR0THEL in Pompeii,” and “What GETTING WASTED Was Like in Medieval Times.” One of the channels has even been livestreaming this "history" 24/7 for weeks.
In the daytime, when I was not groggy and half asleep, it quickly became obvious to me that all of these videos are AI generated, and that they are part of a sophisticated and growing AI slop content ecosystem that is flooding YouTube, is drowning out human-made content created by real anthropologists and historians who spend weeks or months researching, fact-checking, scripting, recording, and editing their videos, and are quite literally rewriting history with surface-level, automated drek that the YouTube algorithm delivers to people. YouTube has said it will demonetize or otherwise crack down on “mass produced” videos, but it is not clear whether that has had any sort of impact on the proliferation of AI-generated videos on the platform, and none of the people I spoke to for this article have noticed any change.
Submission + - Microbe with bizarrely tiny genome may be evolving into a virus (science.org)
sciencehabit writes: The newly discovered microbe provisionally known as Sukunaarchaeum isn’t a virus. But like viruses, it seemingly has one purpose: to make more of itself.
As far as scientists can tell from its genome—the only evidence of its existence so far—it’s a parasite that provides nothing to the single-celled creature it calls home. Most of Sukunaarchaeum’s mere 189 protein-coding genes are focused on replicating its own genome; it must steal everything else it needs from its host Citharistes regius, a dinoflagellate that lives in ocean waters all over the world. Adding to the mystery of the microbe, some of its sequences identify it as archaeon, a lineage of simple cellular organisms more closely related to complex organisms like us than to bacteria like Escherichia coli.
The discovery of Sukunaarchaeum’s bizarrely viruslike way of living, reported last month in a bioRxiv preprint, “challenges the boundaries between cellular life and viruses,” says Kate Adamala, a synthetic biologist at the University of Minnesota Twin Cities who was not involved in the work. “This organism might be a fascinating living fossil—an evolutionary waypoint that managed to hang on.”
Adamala adds that if Sukunaarchaeum really does represent a microbe on its way to becoming a virus, it could teach scientists about how viruses evolved in the first place. “Most of the greatest transitions in evolution didn’t leave a fossil record, making it very difficult to figure out what were the exact steps,” she says. “We can poke at existing biochemistry to try to reconstitute the ancestral forms—or sometimes we get a gift from nature, in the form of a surviving evolutionary intermediate.”
What’s already clear: Sukunaarchaeum is not alone. When team leader Takuro Nakayama, an evolutionary microbiologist at Tsukuba, and his colleagues sifted through publicly available DNA sequences extracted from seawater all over the world, they found many sequences similar to those of Sukunaarchaeum. “That’s when we realized that we had not just found a single strange organism, but had uncovered the first complete genome of a large, previously unknown archaeal lineage,” Nakayama says.
As far as scientists can tell from its genome—the only evidence of its existence so far—it’s a parasite that provides nothing to the single-celled creature it calls home. Most of Sukunaarchaeum’s mere 189 protein-coding genes are focused on replicating its own genome; it must steal everything else it needs from its host Citharistes regius, a dinoflagellate that lives in ocean waters all over the world. Adding to the mystery of the microbe, some of its sequences identify it as archaeon, a lineage of simple cellular organisms more closely related to complex organisms like us than to bacteria like Escherichia coli.
The discovery of Sukunaarchaeum’s bizarrely viruslike way of living, reported last month in a bioRxiv preprint, “challenges the boundaries between cellular life and viruses,” says Kate Adamala, a synthetic biologist at the University of Minnesota Twin Cities who was not involved in the work. “This organism might be a fascinating living fossil—an evolutionary waypoint that managed to hang on.”
Adamala adds that if Sukunaarchaeum really does represent a microbe on its way to becoming a virus, it could teach scientists about how viruses evolved in the first place. “Most of the greatest transitions in evolution didn’t leave a fossil record, making it very difficult to figure out what were the exact steps,” she says. “We can poke at existing biochemistry to try to reconstitute the ancestral forms—or sometimes we get a gift from nature, in the form of a surviving evolutionary intermediate.”
What’s already clear: Sukunaarchaeum is not alone. When team leader Takuro Nakayama, an evolutionary microbiologist at Tsukuba, and his colleagues sifted through publicly available DNA sequences extracted from seawater all over the world, they found many sequences similar to those of Sukunaarchaeum. “That’s when we realized that we had not just found a single strange organism, but had uncovered the first complete genome of a large, previously unknown archaeal lineage,” Nakayama says.
Submission + - Help wanted to build open source Advanced Data Protection for everyone
WaywardGeek writes: Recall that Apple was ordered to back-door Advanced Data Protection in the UK. We need to take action now to protect users.
I helped build Google's Advanced Data Protection (Google Cloud Key VaultService) in 2018, and Google is way ahead of Apple in this area. I know exactly how to build it an can have it done in spare time in a few weeks, at least server side. The whole world would be able to use it for free, protecting backups, passwords, message history, and more, if we can get existing applications to talk to the new data protection service.
However, I need help. I've got the algorithms and server-side covered. This would be a distributed trust based system, so I need folks willing to run the protection service. I'll run mine on a Raspberry PI. Areas where I need help include:
* Running protection servers. This is a T-of-N scheme, where users will need say 9 of 15 nodes to be available to recover their backups.
* Android client app, and preferably tight integration with the platform as an alternate backup service.
* Same with iOS
* Authentication. Users should register, and login before they can use any of their limited guesses to their phone unlock secret.
The scheme splits a secret among N protection servers, and when it is time to recover the secret, which is basically an encryption key, they must be able to get key shares from T of the original N servers. This uses a distributed oblivious pseudo random function algorithm, which is very simple.
In plain English, it provides nation-state resistance to secret back doors, and eliminates secret mass surveillance, at least when it comes to data backed up to the cloud. iOS and Android systems don't currently do that. The UK and similarly confused governments will need to negotiate with operators in multiple countries to get access to any given users's keys. There are cases where rational folks would agree to hand over that data, and I hope we can end the encryption wars and develop sane policies that protect user data while offering a compromise where lives can be saved.
So, nothing too serious :-)
Are you up for this challenge? Are you ready to plunge into this with me?
I helped build Google's Advanced Data Protection (Google Cloud Key VaultService) in 2018, and Google is way ahead of Apple in this area. I know exactly how to build it an can have it done in spare time in a few weeks, at least server side. The whole world would be able to use it for free, protecting backups, passwords, message history, and more, if we can get existing applications to talk to the new data protection service.
However, I need help. I've got the algorithms and server-side covered. This would be a distributed trust based system, so I need folks willing to run the protection service. I'll run mine on a Raspberry PI. Areas where I need help include:
* Running protection servers. This is a T-of-N scheme, where users will need say 9 of 15 nodes to be available to recover their backups.
* Android client app, and preferably tight integration with the platform as an alternate backup service.
* Same with iOS
* Authentication. Users should register, and login before they can use any of their limited guesses to their phone unlock secret.
The scheme splits a secret among N protection servers, and when it is time to recover the secret, which is basically an encryption key, they must be able to get key shares from T of the original N servers. This uses a distributed oblivious pseudo random function algorithm, which is very simple.
In plain English, it provides nation-state resistance to secret back doors, and eliminates secret mass surveillance, at least when it comes to data backed up to the cloud. iOS and Android systems don't currently do that. The UK and similarly confused governments will need to negotiate with operators in multiple countries to get access to any given users's keys. There are cases where rational folks would agree to hand over that data, and I hope we can end the encryption wars and develop sane policies that protect user data while offering a compromise where lives can be saved.
So, nothing too serious
Are you up for this challenge? Are you ready to plunge into this with me?
Comment iPhones are expensive in China⦠(Score 1) 56
Government taxes foreign phones in China making iPhones more expensive than those in other places, including Hong Kong. iPhones are not allowed to use the AI being sold with the phones elsewhere, plus a host of other restrictions. So basically, an iPhone purchased in China isnâ(TM)t the same as an iPhone purchased elsewhere. Many people buy them from third party distributors to get non-mainland SKUs or go elsewhere to buy iPhones to sidestep restrictions/tariffs/taxes/etc.
Submission + - China Halts Rare Earth Exports to U.S. (thegatewaypundit.com)
AmiMoJo writes: China has halted exports of seven critical rare earth elements to the United States, a move that threatens to disrupt supply chains across key American industries, including automotive, semiconductor, and aerospace sectors. China’s Ministry of Commerce recently added seven rare earth elements—including dysprosium, terbium, and lutetium—to its restricted export list. These elements are essential for manufacturing high-performance magnets used in electric vehicles, advanced weaponry, and consumer electronics.
Additionally: US chipmakers outsourcing manufacturing will escape China's tariffs
U.S. chipmakers that outsource manufacturing will be exempt from China's retaliatory tariffs on U.S. imports, according to a notice by the main Chinese semiconductor association on Friday.
Given the highly specialized and multi-country nature of chip supply chains, there was uncertainty within the industry about how tariffs would be applied to chip imports.
"For all integrated circuits, whether packaged or unpackaged, the declared country of origin for import customs purchases is the location of the wafer fabrication plant," the state-backed China Semiconductor Industry Association (CSIA), which represents the country's largest chip companies, said in an "urgent notice" on its WeChat account.
For U.S. chip designers such as Qualcomm and AMD that outsource manufacturing to Taiwanese chipmaking giant TSMC 2330.TW, Chinese customs authorities will classify these chips' place of origin as Taiwan, according to EETop, an information platform and forum for Chinese chipmakers.
This means China-based companies importing such chips will not be forced to pay China's retaliatory tariffs on U.S. imports, EETop said on its WeChat account.
https://www.reuters.com/techno...
Additionally: US chipmakers outsourcing manufacturing will escape China's tariffs
U.S. chipmakers that outsource manufacturing will be exempt from China's retaliatory tariffs on U.S. imports, according to a notice by the main Chinese semiconductor association on Friday.
Given the highly specialized and multi-country nature of chip supply chains, there was uncertainty within the industry about how tariffs would be applied to chip imports.
"For all integrated circuits, whether packaged or unpackaged, the declared country of origin for import customs purchases is the location of the wafer fabrication plant," the state-backed China Semiconductor Industry Association (CSIA), which represents the country's largest chip companies, said in an "urgent notice" on its WeChat account.
For U.S. chip designers such as Qualcomm and AMD that outsource manufacturing to Taiwanese chipmaking giant TSMC 2330.TW, Chinese customs authorities will classify these chips' place of origin as Taiwan, according to EETop, an information platform and forum for Chinese chipmakers.
This means China-based companies importing such chips will not be forced to pay China's retaliatory tariffs on U.S. imports, EETop said on its WeChat account.
https://www.reuters.com/techno...
Submission + - AI proves that fingerprints are not unique, upending the legal system (earth.com)
fjo3 writes: Fingerprint analysis has been a dependable tool in crime-solving for more than a century. Investigators lean on fingerprint evidence to identify suspects or connect them to specific crime scenes, believing that every print offers a distinctive code.
Yet, a team of researchers has found that prints from different fingers of the same person can sometimes appear more alike.
This insight came from an artificial intelligence model that revealed surprising connections between prints.
Yet, a team of researchers has found that prints from different fingers of the same person can sometimes appear more alike.
This insight came from an artificial intelligence model that revealed surprising connections between prints.
Submission + - DeepMind Details All the Ways AGI Could Wreck the World (arstechnica.com)
An anonymous reader writes: Researchers at DeepMind have ... released a new technical paper (PDF) that explains how to develop AGI safely, which you can download at your convenience. It contains a huge amount of detail, clocking in at 108 pages before references. While some in the AI field believe AGI is a pipe dream, the authors of the DeepMind paper project that it could happen by 2030. With that in mind, they aimed to understand the risks of a human-like synthetic intelligence, which they acknowledge could lead to "severe harm." This work has identified four possible types of AGI risk, along with suggestions on how we might ameliorate said risks. The DeepMind team, led by company co-founder Shane Legg, categorized the negative AGI outcomes as misuse, misalignment, mistakes, and structural risks.
The first possible issue, misuse, is fundamentally similar to current AI risks. However, because AGI will be more powerful by definition, the damage it could do is much greater. A ne'er-do-well with access to AGI could misuse the system to do harm, for example, by asking the system to identify and exploit zero-day vulnerabilities or create a designer virus that could be used as a bioweapon. DeepMind says companies developing AGI will have to conduct extensive testing and create robust post-training safety protocols. Essentially, AI guardrails on steroids. They also suggest devising a method to suppress dangerous capabilities entirely, sometimes called "unlearning," but it's unclear if this is possible without substantially limiting models.
Misalignment is largely not something we have to worry about with generative AI as it currently exists. This type of AGI harm is envisioned as a rogue machine that has shaken off the limits imposed by its designers. Terminators, anyone? More specifically, the AI takes actions it knows the developer did not intend. DeepMind says its standard for misalignment here is more advanced than simple deception or scheming as seen in the current literature. To avoid that, DeepMind suggests developers use techniques like amplified oversight, in which two copies of an AI check each other's output, to create robust systems that aren't likely to go rogue. If that fails, DeepMind suggests intensive stress testing and monitoring to watch for any hint that an AI might be turning against us. Keeping AGIs in virtual sandboxes with strict security and direct human oversight could help mitigate issues arising from misalignment. Basically, make sure there's an "off" switch.
If, on the other hand, an AI didn't know that its output would be harmful and the human operator didn't intend for it to be, that's a mistake. We get plenty of those with current AI systems—remember when Google said to put glue on pizza? The "glue" for AGI could be much stickier, though. DeepMind notes that militaries may deploy AGI due to "competitive pressure," but such systems could make serious mistakes as they will be tasked with much more elaborate functions than today's AI. The paper doesn't have a great solution for mitigating mistakes. It boils down to not letting AGI get too powerful in the first place. DeepMind calls for deploying slowly and limiting AGI authority. The study also suggests passing AGI commands through a "shield" system that ensures they are safe before implementation.
Lastly, there are structural risks, which DeepMind defines as the unintended but real consequences of multi-agent systems contributing to our already complex human existence. For example, AGI could create false information that is so believable that we no longer know who or what to trust. The paper also raises the possibility that AGI could accumulate more and more control over economic and political systems, perhaps by devising heavy-handed tariff schemes. Then one day, we look up and realize the machines are in charge instead of us. This category of risk is also the hardest to guard against because it would depend on how people, infrastructure, and institutions operate in the future.
The first possible issue, misuse, is fundamentally similar to current AI risks. However, because AGI will be more powerful by definition, the damage it could do is much greater. A ne'er-do-well with access to AGI could misuse the system to do harm, for example, by asking the system to identify and exploit zero-day vulnerabilities or create a designer virus that could be used as a bioweapon. DeepMind says companies developing AGI will have to conduct extensive testing and create robust post-training safety protocols. Essentially, AI guardrails on steroids. They also suggest devising a method to suppress dangerous capabilities entirely, sometimes called "unlearning," but it's unclear if this is possible without substantially limiting models.
Misalignment is largely not something we have to worry about with generative AI as it currently exists. This type of AGI harm is envisioned as a rogue machine that has shaken off the limits imposed by its designers. Terminators, anyone? More specifically, the AI takes actions it knows the developer did not intend. DeepMind says its standard for misalignment here is more advanced than simple deception or scheming as seen in the current literature. To avoid that, DeepMind suggests developers use techniques like amplified oversight, in which two copies of an AI check each other's output, to create robust systems that aren't likely to go rogue. If that fails, DeepMind suggests intensive stress testing and monitoring to watch for any hint that an AI might be turning against us. Keeping AGIs in virtual sandboxes with strict security and direct human oversight could help mitigate issues arising from misalignment. Basically, make sure there's an "off" switch.
If, on the other hand, an AI didn't know that its output would be harmful and the human operator didn't intend for it to be, that's a mistake. We get plenty of those with current AI systems—remember when Google said to put glue on pizza? The "glue" for AGI could be much stickier, though. DeepMind notes that militaries may deploy AGI due to "competitive pressure," but such systems could make serious mistakes as they will be tasked with much more elaborate functions than today's AI. The paper doesn't have a great solution for mitigating mistakes. It boils down to not letting AGI get too powerful in the first place. DeepMind calls for deploying slowly and limiting AGI authority. The study also suggests passing AGI commands through a "shield" system that ensures they are safe before implementation.
Lastly, there are structural risks, which DeepMind defines as the unintended but real consequences of multi-agent systems contributing to our already complex human existence. For example, AGI could create false information that is so believable that we no longer know who or what to trust. The paper also raises the possibility that AGI could accumulate more and more control over economic and political systems, perhaps by devising heavy-handed tariff schemes. Then one day, we look up and realize the machines are in charge instead of us. This category of risk is also the hardest to guard against because it would depend on how people, infrastructure, and institutions operate in the future.
Slashdot Top Deals
In seeking the unattainable, simplicity only gets in the way. -- Epigrams in Programming, ACM SIGPLAN Sept. 1982
