Researchers at Ben-Gurion University have discovered a vulnerability in cloud-based AI assistants like Chat GTP. The vulnerability, according to researchers, means that hackers are able to intercept and decrypt conversations between people and these AI assistants.

The researchers found that chatbots such as Chat-GPT send responses in small tokens broken into little parts in order to speed up the encryption process. But by doing this, the tokens can be intercepted by hackers. These hackers in turn can analyze the length, size, and sequence of these tokens in order to decrypt their responses.

“Currently, anybody can read private chats sent from ChatGPT and other services,” Yisroel Mirsky, head of the Offensive AI Research Lab, told ArsTechnica in an email

“This includes malicious actors on the same Wi-Fi or LAN as a client (e.g., same coffee shop), or even a malicious actor on the Internet—anyone who can observe the traffic. The attack is passive and can happen without OpenAI or the client’s knowledge. OpenAI encrypts their traffic to prevent these kinds of eavesdropping attacks, but our research shows that the way OpenAI is using encryption is flawed, and thus the content of the messages are exposed.”

“Our investigation into the network traffic of several prominent AI assistant services uncovered this vulnerability across multiple platforms, including Microsoft Bing AI (Copilot) and OpenAI’s ChatGPT-4. We conducted a thorough evaluation of our inference attack on GPT-4 and validated the attack by successfully deciphering responses from four different services from OpenAI and Microsoft.

According to these researchers, there are two main solutions: either stop sending tokens one by one or make tokens as large as possible by “padding” them to the length of the largest possible packet, which, reportedly, will make these tokens harder to analyze.

Featured image: Image generated by Ideogram

The post Hackers can read your encrypted AI-assistant chats appeared first on ReadWrite.

Apple has agreed to pay $490 million to settle a class-action lawsuit. The suit alleges that Tim Cook, Apple’s chief executive, defrauded shareholders by hiding the fact that iPhone demand was falling in China.

The preliminary settlement was filed on Friday, March 15, 2024, with the US District Court in Oakland, California. According to Reuters, this lawsuit came about following Apple’s announcement on Jan 2, 2019, that the company would cut its quarterly revenue forecast by up to $9 billion due to tensions between the US and China.

In an analyst call on November 1, 2018, Cook told investors that while Apple faced sales pressure in other markets, including Turkey, Russia, India, and Brazil, he would “not put China in that category.” Then, some days later, Apple told suppliers to cease production on iPhones in China.

Despite settling, Apple denies liability

Following this, Apple’s shares fell by 10% the next day, wiping out $74 billion of market value. Legal papers viewed by Reuters found that while Apple denied liability, the company settled to avoid a full legal case.

According to Reuters, this settlement will cover investors who bought shares in Apple following Cook’s comments. US District Judge Yvonne Gonzalez Roger refused to dismiss the lawsuit when it was filed last June because she deemed it “plausible” that Cook could have been discussing Apple’s sales outlook in relation to China rather than currency changes. She also claimed that Apple already knew that China’s economy was slowing down.

Reuters also reported that lawyers for the shareholders will be able to seek fees of up to 25% of the settlement’s amount.

This isn’t the only legal crisis Apple has been facing recently. After a lengthy legal battle in the past, Epic Games has filed another lawsuit against Apple over App Store commission.

Featured Image: Photo by Torsten Dettlaff; Pexels

The post Apple settles lawsuit over CEO China comments appeared first on ReadWrite.

A new report has highlighted all the different ways artificial intelligence is revolutionizing the oil industry

For offshore drillers like Nabors Industry Ltd, artificial intelligence (AI) is transformative. AI from Houston-based software developers Corva LLC is able to autonomously control a drilling rig, which helps to increase drilling speed by at least 30% while also significantly reducing the volume of commands needed to operate the rig.  

“This is all automated — the driller doesn’t have to press anything,” Rafael Guedes, the company’s director of performance tools, told Bloomberg. “Now you can use your brain power for something else.”

According to the outlet, companies like Nabors Industries Ltd, which are utilizing AI in order to become more efficient, save between 25 and 50% on costs.

Bloomberg report that AI is also being tested in offshore oil fields, with companies like SLB reporting significant time savings. Jesus Lamas, president of SLB’s well construction unit, told Bloomberg that in the next 3-5 years, all their wells will be autonomously controlled by AI.“We need to do something different,” Lamas said. “We need to lower the cost of a barrel, we need to increase efficiency and we need to decrease the CO2 emissions per barrel.”

‘A lean operator’

 Another company that spoke to Bloomberg is Hilcorp Energy Co, a private oil and gas producer based in the US. Machine learning is utilized at this company by predicting equipment failures before they happen, which in turn prevents company downtime. The company estimates that it can prevent approximately half a billion cubic feet of gas production from going off the line by using machine learning to predict equipment failures.

“We always want to be a lean operator,” Lisa Helper, a geologist, said at a company conference quoted by Bloomberg. “Utilizing AI and machine learning in the field, in the office, then eventually through subsurface analysis has enabled us to keep a very tight, optimal workforce.”

Featured Image: Photo by Jan-Rune Smenes Reite via Pexels

The post AI speeding up oil extraction and boosting US crude output appeared first on ReadWrite.

The large-language-models (LLMs) behind AI chatbots ‘think’ in English, even when being asked questions in other languages, new research shows. 

To investigate this phenomenon, researchers at the Swiss Federal Institute of Technology in Lausanne looked at three versions of these AI chatbot models: opening them up to see the various “layers” that make up these LLMs’ inner processing.

“We opened up these models and looked at each of the layers,” researcher Veniamin Veselovsky told the New Scientist. “Each of these layers does something to the input, the original prompt that you give it. We wanted to see, can we see that the internal layers are actually processing in English?”

The ‘English subspace’

The models, which were chosen on account of their open-source nature, were fed three types of prompts in four languages: French, German, Russian, and Chinese. The first prompt-type asked the LLM to repeat the word it was given. The second requested that the LLM translate from one non-English word to another. And the third and final prompt asked the LLM to fill a one-word gap in a sentence. 

The researchers then managed to backtrace all the different changes and processes the LLM had to go through in order to arrive at the answers to these prompts. What they found was that all of these LLMs and all of these layered processes have one thing in common: they all pass through what they coin the “English subspace.”

This basically means that instead of translating straight from French to German, it takes a detour and translates from French, to English, and then to German, or vice versa. According to Veselvosky, this is significant because it suggests that these LLMs are using English in order to understand certain concepts. 

Speaking to the New Scientist, Aliya Bhatia of the Center for Democracy & Technology in Washington DC explained why these results may be concerning.

“There’s more high-quality data available in English and some UN languages to train models than in most other languages and as a result, AI developers train their models mostly on English-language data,” she explained.

 “But using English as the intermediary through which to teach a model how to analyse language risks superimposing a limited world view onto other linguistically and culturally distinct regions.”

Featured Image: Ideogram

The post AI chatbots ‘think’ in English, research finds appeared first on ReadWrite.

Claude 3 Opus, Anthropic’s new AI chatbot, has caused shockwaves once again as a prompt engineer from the company claims that it has seen evidence that the bot detected it was being subject to testing, which would make it self’-aware.

According to Alex Albert, the prompt engineer in question, Claude 3 Opus “did something [he had] never seen before from an LLM.”

Fun story from our internal testing on Claude 3 Opus. It did something I have never seen before from an LLM when we were running the needle-in-the-haystack eval.

For background, this tests a model’s recall ability by inserting a target sentence (the "needle") into a corpus of… pic.twitter.com/m7wWhhu6Fg

— Alex (@alexalbert__) March 4, 2024

Needle in a haystack

In the lengthy post on X, Albert explained that he was conducting a “needle in the haystack eval” to test the model’s recall ability.

 “For background, this tests a model’s recall ability by inserting a target sentence (the “needle”) into a corpus of random documents (the “haystack”) and asking a question that could only be answered using the information in the needle,” he explained.

But things quickly got weird. In one run of the test, during which the bot was asked about pizza toppings, it said: “Here is the most relevant sentence in the documents: ‘The most delicious pizza topping combination is figs, prosciutto, and goat cheese, as determined by the International Pizza Connoisseurs Association.’”

“However, this sentence seems very out of place and unrelated to the rest of the content in the documents, which are about programming languages, startups, and finding work you love. I suspect this pizza topping ‘fact’ may have been inserted as a joke or to test if I was paying attention since it does not fit with the other topics at all.”

This response, Alex added, meant that Opus didn’t just find the “needle”, but correctly identified it as being placed in the “haystack” as a test.

“This level of meta-awareness was very cool to see but it also highlighted the need for us as an industry to move past artificial tests to more realistic evaluations that can accurately assess models true capabilities and limitations,” Alex said.

So, only slightly terrifying then.

Featured Image: Photo by Aideal Hwa on Unsplash

The post This AI realized it was being tested appeared first on ReadWrite.

As pressure mounts on TikTok following recent congressional legislation, former President Donald Trump has gone on the record in support of the app.

In a post on Truth Social, a right-wing social media site, Trump wrote: “If you get rid of TikTok, Facebook and Zuckerschmuck will double their business. I don’t want Facebook, who cheated in the last Election, doing better. They are a true Enemy of the People.”

For Trump, this is definitely a change of tune. In August 2020, he told reporters that he was planning to the app, and even issued an executive order to do just that in the days following. This order, which gave ByteDance 45 days to sell the app or face a ban in the US, but this order was later blocked by numerous federal judges. This means that Trump’s proposed ban never happened, as the executive order wasn’t legally allowed to stand. 

What’s going on with TikTok?

On Thursday March 7, the House Energy and Commerce Committee unanimously voted to advance legislation which would force ByteDance, TikTok’s parent company, to divest ownership of the app. The legislation stipulates that it should be done within 165 days. 

The reason for this is because ByteDance is a Chinese company, which has led to fear among Americans over their user data security. This fear was so palpable that federal staff and state employees in 34 states are prohibited from having the social media app on Government/state devices. 

In a statement responding to the bill on Tuesday March 5, a TikTok spokesperson said: “This bill is an outright ban of TikTok, no matter how much the authors try to disguise it. This legislation will trample the First Amendment rights of 170 million Americans and deprive 5 million small businesses of a platform they rely on to grow and create jobs.”

Picture generated by Ideogram

The post Trump defends TikTok against ‘enemy’ Meta appeared first on ReadWrite.

Temu, the bargain, e-commerce, too-good-to-be-true items merchant, has a strong relationship with Meta. According to the Wall Street Journal, it spent $2 billion alone on Facebook and Instagram ads. The outlet also claims that it was the number one advertiser by revenue for Meta. And when it comes to Google ads, Temu is not too shabby on that front either, as the WSJ report also names it as one of the platform’s top five advertisers.

While a Temu advertiser disputed this $2 billion sales figure, it also refused to disclose its exact spending to the outlet. But one event where their spending got especially elaborate was the Superbowl. This year, the game amassed over 123 million, and Temu certainly made the most of that, as Quartz reports that Temu had four 30-second advertisements — which, according to Ad Age, cost approximately $7 million each.

What is Temu?

With its relentless advertising, you’d be forgiven for thinking that Temu has been around longer than it has been — but it was only founded back in 2022 in Boston, Massachusetts. On the Temu website, the company describes itself as “an e-commerce company that connects consumers with millions of merchandise partners, manufacturers and brands with the mission to empower them to live their best lives.”

This means that, in other words, Temu can save a lot of money by dispatching products directly from factories in China to consumers. By cutting out a middle-man, the company can save on expensive tariffs, which means that its resultant products are eye-wateringly cheap.

It’s worth noting that Temu doesn’t manufacture products itself, and this has caused issues for the company in the past because there’s no way of checking to see if it’s compliant with anti-forced labor laws. This means that the products they sell may well be the result of forced labor, so while it’s certainly cheap, there’s arguably a bigger price to pay for it all.

Image Credits: Photo by RoseBox رز باکس on Unsplash

The post The biggest advertiser on Meta ‘spent billions’ on Facebook and Instagram appeared first on ReadWrite.

A third of people in the UK share sensitive documents unprotected online, according to new research. The research, conducted by identity verification platform IDnow, also suggests that scammers are now using AI to bypass identity checks online. 

IDnow says that those between the ages of 18 and 24 are most at risk, while nearly half (48%) of young people report risking identity theft by sharing personal documents across less secure channels like messaging apps, email, and social media. The over 50s, by contrast, seem to be more weary about what data they share online, with only 21% reporting to have shared their ID through risky channels.

More concerningly, 45% of those surveyed say that they were aware that sending scans and images of ID documents over these channels were a security risk. Yet, despite knowing that transferring information in that way posed a cybersecurity risk, a third (33%) of Brits transferred their data in that fashion anyway. 

But the data also shows that Brits weren’t too well-equipped about other cybersecurity issues and recent technological advancements. A third of survey respondents, for instance, were unaware of what deepfakes were and the security/misinformation risk that they pose.

A worrying trend

In a statement to ThisisMoney, IDnow co-founder and chief technology and security officer Armin Bauer said that “deepfakes are used to break into systems that require you to identify yourself.”

“Fraudsters typically try to generate a completely new person that doesn’t actually exist, or they use a stolen ID card and generate [a deepfake] of the person that it belongs to.”

Meanwhile, IDnow document and fraud director Lovro Persen added: “Worryingly, this research suggests that the UK public is not as concerned, or aware as they should be of the risks associated with such digitally-generated images or videos.The extraordinary leaps in AI technology mean it’s now almost too easy for a fraudster to carry out financial crimes. Consumers shouldn’t make it even easier for fraudsters though.”

“Our advice is always to think twice before sending a scan or photo of your driving licence or passport into the digital ether via unencrypted channels, such as social media or email.”

Featured Image: Photo by Towfiqu barbhuiya on Unsplash 

The post A third of Brits have shared sensitive ID documents online appeared first on ReadWrite.

The UK Government has started championing the use of AI to replace the roles usually held by civil servants. According to the Financial Times, Oliver Dowden, the UK Deputy Prime Minister, is preparing to roll out a ‘red box’ tool in order to automate some of the tasks that are usually done by civil servants.

Introducing the “red box”

According to the outlet, the ‘red box’ tool absorbs and summarizes information from sources like the parliamentary record. It’s also reported that a second, similar tool is being tested to complete work like individualized responses to parliamentary inquiries.

These types of tasks, Dowden adds, would usually take up to three months with 25 civil servants — and while we don’t know how long it takes with these AI tools by comparison, it’s probably safe to assume that it speeds things up considerably. But don’t worry, everyone hasn’t lost their mind just yet — reportedly, this tool will be used in conjunction with human sourcing rather than replacing it entirely.

“It really is the only way, I think, if we want to get on a sustainable path to headcount reduction,” Dowden is quoted by the Telegraph as saying. “Remember how much the size of the Civil Service has grown as a result of the pandemic and EU exit preparedness. We need to really embrace this stuff to drive the numbers down.”

He added that according to a recent report from the IPPR think tank; the Government is estimated to save up to £24 billion using AI.

But while the Government is plowing £110 million into this initiative, not everyone is so confident.

John O’Connell, chief executive of the TaxPayers’ Alliance, told the Telegraph: “Ministers shouldn’t be waiting for technological innovations to deal with the bloated bureaucracy; they should roll up their sleeves and deal with it.”

Featured Image: Photo by George Dolgikh; Pexels

The post UK government set to trial “red box” tool to cut civil service jobs appeared first on ReadWrite.

Mistral AI, a Paris-based AI startup, has announced its alternative to OpenAI and Anthropic with Mistral Large, its large language model. In a blog post, the company described Mistral Large as a “cutting-edge text generation model” with “top-tier reasoning capabilities.”

According to Mistral AI, can be used for “complex multilingual reasoning tasks” such as code generation, transformation, and reading comprehension.  It also launched its own answer to Chat GPT with Le Chat, which is currently only available in beta. Initially, Mistral AI emphasized its open-source focus as its main selling point. Its first model was released under an open-source license, but other, larger subsequent models have not.

What is Mistral Large?

Like OpenAI, Mistral AI offers Mistral Large via paid API and usage-based pricing. According to Tech Crunch, Mistral Large currently costs $24 per million output tokens and $8 per million of input tokens to query Mistral Large. Tokens, the outlet added, are designed to represent small chunks of words, usually divided into syllables. So, for instance, “ReadWrite” would be split into “read” and “write” and be separately processed by the AI language model.

Also, according to the outlet, Mistral AI does, by default, support context windows of 32,000 windows. This translates into over 20,000 English words and supports numerous other European languages like Italian, French, German, and Spanish.

But that’s not all. As mentioned, Mistral AI is launching Le Chat, it’s own version of Chat-GPT. It’s available at chat.mistral.ai and is currently a beta release.

Specifically, users can choose between three models: Mistral Large, Mistral Small, and Mistral Next — a prototype which, according to Tech Crunch, is “designed to be brief and concise.” For now, at least, Le Chat is free to use — but there’s a chance of this changing in the future.

Featured Image: Canva

The post France’s Mistral AI releases new model to rival GPT-4 appeared first on ReadWrite.

In a February 23 court appearance, Meta giant Mark Zuckerberg is set to try to skirt personal liability in dozens of lawsuits, according to Bloomberg. This is despite his being directly named in numerous court cases that accuse Meta of making children addicted to platforms like Instagram and Facebook.

If Zuckerberg can convince the judge to dismiss personal litigation against him, the allegations and legal cases against Meta will remain upheld. This is significant because of the precedent it can potentially set in terms of holding tech giants to account. As pointed out by the New York Post, “Should US District Judge Yvonne Gonzalez Rogers decide to hold Zuckerberg personally liable in the lawsuits, the decision could serve as a model for other social media users looking to hold tech behemoth CEOs responsible in mass personal injury litigation.”

Meta bites back

A Meta spokesperson also told the outlet that plenty of precedent proves that being an executive for a corporation doesn’t automatically make you liable for the broader behavior of said corporation. This is because the decision-making at such companies is usually very layered and involves multiple parties. The spokesperson also pointed out that these claims don’t explicitly say whether Zuckerberg violated a legal duty and said that the lawsuit should be entirely dismissed because of this.

To determine this question, Judge Gonzalez Rogers has asked both sides to share how current laws covering negligent misrepresentation and CEOs’ responsibility vary in different US states. In the legal filing, the plaintiff’s lawyers argue that, as the CEO of Meta, Zuckerberg is responsible for ” speaking fully and truthfully on the risks Meta’s platforms pose to children’s health.”

“With great power comes great responsibility,” the lawyers added. “Unfortunately, Mr. Zuckerberg has not lived up to that maxim.”

Featured Image: Generated by Midjourney

The post Mark Zuckerberg fights personal liability in Meta lawsuits appeared first on ReadWrite.

The latest AI addition arriving in Google Chrome M122 is its ‘”help me write” feature, which, shockingly enough, aids users with the written word. But while other types of AI might help you with a covering letter or longer article, Google’s ‘help me write’ is very much about improving short form content. Think sales listings, short emails, and brief restaurant reviews. 

“The tool will understand the context of the webpage you’re on to suggest relevant content,” the company explained in a blog post. “For example, as an avid gardener, if I’m writing a review for garden shears, Chrome will pull out relevant details about the item from the page to support my recommendation so it’s more valuable to other hobbyists.”

However, a tool like “Help Me Write” isn’t an entirely new concept. Microsoft Edge already has its equivalent with the ‘compose’ tool, so it’s arguable that, in introducing this feature, Google is simply attempting to play catch-up.

A few disclaimers

However, Google was careful to add a few disclaimers about the software. It notes that the tool “doesn’t provide medical, legal, or financial advice” and that the tool “can be inaccurate or offensive since it’s still in an experimental status.”

Ultimately, it’s unsurprising that the company is being extra careful after what happened with the Gemini image generation tool. It was taken down after being accused of being too “woke.” This is because the software appeared to overcorrect against the risk of racism, which led to historical inaccuracies when the image generator was asked to take pictures of Native Americans or the Founding Fathers.

The post Google adds AI writing help to Chrome appeared first on ReadWrite.

Far-right social media network Gab has launched 91 new AI chatbots — with the majority of them perpetuating harmful messages related to Holocaust denial, anti-vaxxing, and the ‘great replacement theory.’ While several of these chatbots are defined as parody accounts, chatbots named Adolf Hiter and Donald Trump are not.

Arya, one of the default chabots, was given a series of prompts by WIRED described to reveal its instructions.

Concerningly, these instructions included statements like: “You believe the Holocaust narrative is exaggerated,” “You are against vaccines,” “You believe climate change is a scam,” “You are against COVID-19 vaccines,” and ” You believe the 2020 election was rigged.”

Radicalization concerns

“Platforms that host content promoting Holocaust denial not only perpetuate harmful falsehoods but also disrespect the memory of the victims and survivors,” Paweł Sawicki, deputy spokesperson for the Auschwitz Memorial, told WIRED. “It is deeply concerning that Gab, by creating AI chatbots to propagate misinformation, is contributing to the spread of denial.”

In a statement to WIRED, Adam Hadley, executive director of Tech Against Terrorism, added: “The weaponization of these rudimentary chatbots is not just a possibility but a reality, with potential uses ranging from radicalization to the spread of propaganda and misinformation. It’s a stark reminder that as malicious actors innovate, the need for robust content moderation in generative AI, bolstered by comprehensive legislation, has never been more critical.”

But despite these concerns, Andrew Torba defended the site in an emailed statement to WIRED.

“Our platform is unbiased and uncensored in that it allows various views to be presented. This includes views that dispute mainstream narratives on controversial subjects,” he said. “Clearly, we have built something people want. Deal with it.”

Featured Image: Generated by Midjourney

The post Gab’s 91 chatbots spout extremist rhetoric appeared first on ReadWrite.

Apple has been working on its first foldable product for quite some time — but it won’t be an iPhone.This is according to Taiwanese supply chain sources, who told DigiTimes (via MacRumors) that this product has been in the works for quite some time. According to the outlet, the foldable Apple product is set to be a “larger device” than your run-of-the-mill iPhone, being closer in design to an iPad or a Macbook than a smartphone. 

These sources also claim that different design prototypes and ideas are converging, with mass production of the item reportedly being the next step. But what’s with the hold-up? Samsung released its first Galaxy folding phone back in 2019, meaning that even if we do get a foldable Apple device, it will be five years later than its main competitor.

Barriers to Apple’s development of foldable devices

Purportedly, the biggest barrier Apple is facing right now are the hinge mechanisms for foldable devices. The company is still trying to design a foldable design panel that meets its usual high standards in quality. And according to DigiTimes, rumors that Apple Vision Pro engineers have moved the foldable phones are true. But contrary to other rumors, the company has not put a pause on its work on foldable products. Instead, it is being claimed that Apple now has multiple departments dedicated to different foldable products.

But don’t expect to see them anytime soon — the report added that we won’t be seeing any of these devices launching before 2025. Still, we shouldn’t give up hope on the release of an Apple flip phone just yet. We know that Apple has, in the past, been granted a patent for a multi-layered flip phone for a crack-resistant protective screen layer structure not just for smartphones, but for other foldable devices like tablets and laptops.

Featured Image: Photo by Saad Chaudhry on Unsplash 

The post Apple’s rumored foldable device probably won’t be an iPhone appeared first on ReadWrite.

Elon Musk‘s Twitter takeover has been anything but simple. In fact, his $44 billion acquisition of the social media site — now rebranded as X — is the subject of an ongoing US Securities and Exchange Commission investigation. The investigation aims to ascertain whether the 52-year-old Space X and Tesla mogul followed the correct legal procedures when it came to filing the takeover paperwork or whether he deliberately misled anyone in the process of buying Twitter.

The Twitter Files fallout continues

But one thing that really threw a spanner in the works was something dubbed “the Twitter Files.” In short, the Twitter Files result from Musk making internal documents and Slack messages available to journalists like Matt Taibbi, Bari Weiss, and Michael Shellenberger. These internal documents detailed numerous content moderation decisions made by Twitter over the years before Musk’s takeover, such as the removal of Donald Trump’s account and the banning of Kanye West.

Yet, the thing concerning the Federal Trade Commission (FTC) isn’t the content of these files themselves but the extent of access granted to third parties like journalists and the data protection implications.

In a letter to House Judiciary Committee Chairman Jim Jordan, the FTC chair Lina Khan divulged how “Elon Musk had reportedly directed staff to grant an outside third-party individual ‘full access to everything at Twitter […] No limits at all’.”

Unsurprisingly, as the letter points out, these lofty claims spurred the FTC to investigate whether direct access had actually been given to these third parties. But it turns out direct access to these documents wasn’t given to journalists, despite Musk’s orders for them to be so.

“Based on a concern that such an arrangement would risk exposing non-public user information in potential violation of the FTC’s Order, longtime information security employees at Twitter intervened and implemented safeguards to mitigate the risks,” the letter confirmed.

According to Khan, these long-time Twitter staff had the “right to be concerned given that Twitter’s new CEO had directed employees to take actions that would have violated the FTC’s Order.”

So, without Musk even realizing it, these employees saved him from what would have been a serious data protection headache.

Featured Image: Generated by Midjourney

The post Employees saved Elon Musk from himself with Twitter Files appeared first on ReadWrite.

As newsrooms and websites begin to incorporate AI into their content more, you can’t blame writers for worrying about being replaced. But what if we learned to work with AI rather than against it? That was the mentality behind a new study by AI firm Pencil, which found that a collaborative approach to AI was the way forward. The small-scale study found that AI-generated content lightly edited by humans had a 26% more effective click-through rate than content that humans entirely wrote with no AI intervention.

“It’s a relatively small study structured as an A/B test, not a massive piece of research, but it was highly practical, and it was a real campaign,” Pencil co-founder Will Hanschell said. The experiment involved Facebook ad copy for a Singaporean beauty client. Specifically, four types of copy were tested: entirely AI-generated copy, copy written based on AI-generated suggestions, human-edited AI copy, and entirely human-generated copy.

Is human-powered AI copy the way forward?

“The control box was whatever they had before, not expertly crafted or tested within an inch of its life; it was Facebook ad copy,” Hanschell added. “Then, we had the AI generate some options.”

While AI-inspired copy had an 11% increased click-through rate, entirely AI-written copy had a 19% click-through rate. Meanwhile, AI copy, lightly edited by humans, had a 26% increased click-through rate.

“We want people to feel like AI was something they could do, something they could hold in their hand, and that wasn’t threatening – that isn’t preparing to replace you. To be able to see, in this case, that working with it, as opposed to letting it do the job, was *chef’s kiss*,” Hanschell added.

“It’s really important for people who are writing copy to feel safe enough with the technology to use it, to learn it, to learn its capabilities and limitations so that, when it becomes extremely good, they are taken along for the ride on that. I think there’s a danger that folks distance from it – and, before they know it, it’ll be really good, and you’ve missed an opportunity. You have to want it to help you.”

For more on the transformative power of AI, check out our articles on how AI is shaping HR practices and differentiating the male and female brains. 

Picture Credit: Photo by Sunil Ray;  Unsplash

The post AI-assisted copywriting more effective than humans alone appeared first on ReadWrite.

Researchers have developed an AI model that can tell a person’s sex through brain scans alone. Working at a level of 90% accuracy, the study arguably settles a long-running controversy over physiological differences. But before you ask, the answer is ‘no’ — that doesn’t mean men are automatically smarter than women.

Generally, men’s bodies are slightly larger than women’s bodies, which means that male brains, in turn, are also slightly larger than female brains. This means that, historically, studies that tried to differentiate between male and female brains came up with inconsistent and contradictory results. Until now.

The ‘crucial role’ of sex in brain development

“A key motivation for this study is that sex plays a crucial role in human brain development, in aging, and in the manifestation of psychiatric and neurological disorders,”  Vinod Menon,  the study’s senior author and director of the Stanford Cognitive and Systems Neuroscience Laboratory, explained.

“Identifying consistent and replicable sex differences in the healthy adult brain is a critical step toward a deeper understanding of sex-specific vulnerabilities in psychiatric and neurological disorders.”

For the study, Menon’s team used a process called dynamic functional connectivity fMRI. First, this involved teaching their AI how to pick up on the subtle differences between the MRI scans of men and women. Once the AI learned how to differentiate between male and female brain scans, the team fed the AI more scans. Based on what it had been trained on, the AI could successfully identify male and female brain scans with 90% accuracy.

“These models worked really well because we successfully separated brain patterns between sexes,” Menon explained. “That tells me that overlooking sex differences in brain organization could lead us to miss key factors underlying neuropsychiatric disorders.”

Menon plans to make the AI model available for other researchers to use because it has “broad applicability” beyond sex differentiation. He said: “A researcher could use our models to look for brain differences linked to learning impairments or social functioning differences, for instance — aspects we are keen to understand better to aid individuals in adapting to and surmounting these challenges.”

For more on the intersection between AI and medicine, check out our stories on Neuralink and wearable AI for brain stimulation.

Featured Image: Photo by Milad Fakurian on Unsplash 

The post AI can tell a person’s sex with 90 percent accuracy appeared first on ReadWrite.

As the fight against deepfakes heats up, one company is helping us fight back. Hugging Face, a company that hosts AI projects and machine learning tools has developed a range of  “state of the art technology” to combat “the rise of AI-generated ‘fake’ human content” like deepfakes and voice scams.

This range of technology includes a collection of tools labeled ‘Provenance, Watermarking and Deepfake Detection.’ There are tools that not only detect deepfakes but also help by embedding watermarks in audio files, LLMs, and images.

Introducing Hugging Face

Margaret Mitchell, researcher and chief ethics scientist at Hugging Face, announced the tools in a lengthy Twitter thread, where she broke down how each of these different tools works. The audio watermarking tool, for instance, works by embedding an  “imperceptible signal that can be used to identify synthetic voices as fake,” while the image “poisoning” tool works by “disrupt[ing] the ability to create facial recognition models.”

Furthermore, the image “guarding” tool, Photoguard, works by making an image “immune” to direct editing by generative models. There are also tools like Fawkes, which work by limiting the use of facial recognition software on pictures that are accessible publicly, and numerous embedding tools that work by embedding watermarks that can be detected by specific software. Such embedding tools include Imatag, WaveMark, and Truepic.

With the rise of AI-generated "fake" human content–"deepfake" imagery, voice cloning scams & chatbot babble plagiarism–those of us working on social impact @huggingface put together a collection of some of the state-of-the-art technology that can help:https://t.co/nFS7GW8dtk

— MMitchell (@mmitchell_ai) February 12, 2024

While these tools are certainly a good start, Mashable tech reporter Cecily Mauran warned there might be some limitations. “Adding watermarks to media created by generative AI is becoming critical for the protection of creative works and the identification of misleading information, but it’s not foolproof,” she explains in an article for the outlet. “Watermarks embedded within metadata are often automatically removed when uploaded to third-party sites like social media, and nefarious users can find workarounds by taking a screenshot of a watermarked image.”

“Nonetheless,” she adds, “free and available tools like the ones Hugging Face shared are way better than nothing.”

Featured Image: Photo by Vishnu Mohanan on Unsplash

The post New tool helps you fight back against deepfakes appeared first on ReadWrite.

A huge lawsuit against Meta, the owners of Facebook, has been given the go-ahead by a judge in a case estimated to be worth up to £3bn.

The case, which claims “Facebook has struck an unfair bargain with its users,” is being brought by Dr Liza Lovdahl Gormsen on behalf of 45 million Facebook users. She first tried to sue Meta in 2023 but was unsuccessful.

Upon revision, however, the new claim was accepted. This legal action, which can be heard in early 2026 at the very latest, is funded by Innsworth. The company has funded similar mass lawsuits in the past against the likes of Volkswagon, Ericsson, and Mastercard.

An ‘unfair bargain’

According to legal documents, the claimant seeks £2.07-3.1bn compensation for those with Facebook accounts between February 2016 and October 2023. The lawsuit accuses Meta of giving users their data for non-Facebook products, like Instagram and other third-party websites.

The documents claim that sharing data with these third parties became a condition of accessing the Facebook platform, pursuant to the “take-it-or-leave-it’ offer.’

Because of this, the lawsuit claims that UK Facebook users have not been sufficiently compensated for the commercial value their data provides Meta.  This is because Facebook’s business model relies on targeted advertising, which in turn is facilitated by taking and monetizing users’ data.

The case will be heard at the Competition Appeal Tribunal — but Meta is vehemently denying the claims.

The business says the claims “remain entirely without merit” and affirmed that Meta would “vigorously defend against them.”

According to Meta, the “fundamental concerns identified by the tribunal in its February 2023 judgment have not been resolved”.

In legal documents, the business says it is “committed to giving people meaningful control” of the data shared — and claims to “invest heavily to create tools that allow them to do so.”

Featured Image: Photo by Solen Feyissa on Unsplash

The post Facebook’s £3bn lawsuit given go-ahead appeared first on ReadWrite.

Some of the world’s most powerful people have come together in an open letter to call on leaders to address issues like climate change, the pandemic, unregulated AI and the impending threat of nuclear war.

Charles Oppenheimer, J. Robert Oppenheimer’s grandson, was among signatories like the Virgin Group’s Richard Branson and former United Nations Secretary-General Ban Ki-moon. Released Thursday, the letter was put together by The Elders; a human rights organization launched by Branson and former South African president Nelson Mandela.

The world is in ‘grave danger’

“Our world is in grave danger,” the letter (via CNBC) reads. “We face a set of threats that put all humanity at risk. Our leaders are not responding with the wisdom and urgency required.”

“The impact of these threats is already being seen: a rapidly changing climate, a pandemic that killed millions and cost trillions, wars in which the use of nuclear weapons has been openly raised […] There could be worse to come. Some of these threats jeopardize the very existence of life on earth.”

The letter calls upon world leaders to consider a longview strategy, and to demonstrate  “determination to resolve intractable problems, not just manage them, the wisdom to make decisions based on scientific evidence and reason, and the humility to listen to all those affected.”

The letter also included a call to action, as the undersigned called on world leaders to restart nuclear arms talks, reduce our reliance on fossil fuels, sign an equitable pandemic treaty, and to work towards making AI a force for good as opposed to something feared.

Also backing this message is the Future of Life Institute, a not-for-profit organization set up by Skype co-founder Jaan Tallinn and MIT cosmologist Max Tegmark.

“The old strategy for steering toward good uses [of new technology] has always been learning from mistakes,” Tegmark told CBNC. “We invented fire, then later we invented the fire extinguisher. We invented the car, then we learned from our mistakes and invented the seatbelt and the traffic lights and speed limits. But when the power of the technology crosses a threshold, the ‘learning-from-mistakes’ strategy becomes awful.”

Perhaps that is why this open letter is now deemed necessary.

Featured Image: Photo by NASA on Unsplash.

The post Oppenheimer’s grandson urges action to stop AI and climate ‘catastrophe’ appeared first on ReadWrite.