Introducing Aurora: X’s New AI Image Generator Model
A More Photorealistic and Less Restrictive Model
X has released a new AI image generator model called "Aurora" that is capable of creating far more photorealistic imagery than its previous model, Grok. According to TechCrunch, Aurora has few apparent restrictions on what it will produce, allowing users to generate a wide range of images, including copyrighted characters and public figures.
Aurora’s Capabilities
Aurora is available as a "Grok 2 + Aurora beta" option in the Grok model selector, but users will only get a few queries before hitting the X Premium subscription paywall. The model is willing to create copyrighted characters, such as Mickey Mouse, and public figures, including "a bloodied Donald Trump," but it stops short of generating nude images.
Realistic but Weird
In another example highlighted by TechCrunch, an X user showed off AI-generated images of Ray Romano and Adam Sandler, which were realistic but with some obvious weirdness when it comes to human anatomy and continuity.
Conclusion
Aurora is a significant step forward in the development of AI image generation technology, offering users a more powerful and less restrictive tool for creating photorealistic images. However, as with any technology, there are concerns about its potential misuse. With great power comes great responsibility, and it’s crucial for users to use Aurora responsibly and ethically.
Frequently Asked Questions
Q: What is Aurora?
A: Aurora is a new AI image generator model developed by X, capable of creating more photorealistic and less restricted images than its previous model, Grok.
Q: How do I access Aurora?
A: Aurora is available as a "Grok 2 + Aurora beta" option in the Grok model selector, but users will only get a few queries before hitting the X Premium subscription paywall.
Q: What kind of images can I generate with Aurora?
A: Aurora is capable of generating a wide range of images, including copyrighted characters and public figures, but it stops short of generating nude images.
Q: Is Aurora available for free?
A: No, users will only get a few queries before hitting the X Premium subscription paywall.
It’s tempting to think of explanations as a layer of polish on top of ideas. We believe that the best explanations are often something much deeper: they are interfaces to ideas, a way of thinking and interacting with a concept.
One of the articles that best exemplifies this type of contribution is Gabriel Goh’s Why Momentum Really Works. Gabe, and other optimization researchers, have a perspective on this problem that may be unfamiliar to practitioners. It involves a mathematical formalism of the spectrum of eigenvalues of the optimization problem, as well as a more informal way of interpreting and thinking about them.
In contrast, this diagram, taken from the article, not only conveys the formalism but also shares some of the author’s intuition. Bolstered by the interactivity, it invites readers to step into a way of thinking. Perhaps most interestingly, by reifying a mental model into a computationally-driven interface, the author discovered places where their thinking was incomplete — specifically, introducing momentum flattens the spectrum of eigenvalues in surprising ways.
One thing we’ve found particularly exciting is how articles can make engaging deeply with ideas an easier and smoother process.
Normally, there’s a huge jump from reading a paper to testing and building on it. But we’re starting to see papers where engagement is a continuous spectrum:
Software Engineering Best Practices for Scientific Publishing
Over the past year, we’ve also seen several advantages to using software engineering best practices to operate a scientific journal.
Every Distill article is housed within a GitHub repository, and peer review is conducted through the issue tracker.
Supporting Authors
Over the last year, we’ve put a lot of energy into mentoring individuals on writing Distill articles.
In the next year, we plan to focus more on scalable ways of helping people by:
Continuing our work on the Distill Template, which provides many of the basic tools needed for writing beautiful web-first academic papers.
Writing a Distill Style Guide describing the best practices we’ve discovered.
Sharing our Distill Reviewer Worksheet so that authors can use it to self-evaluate their article and look for areas to improve.
Starting a Distill Community Slack workspace where people can seek advice, mentorship, and co-authors.
Growing Distill’s Team
We believe that growing Distill’s editorial team is one of the most important ingredients for its long-term success.
As important as it is to expand our editors, it’s equally important to make sure we pick the right editors. This means building up a team deeply aligned with Distill’s unusual values and mission.
Growing Distill’s Scope
In the long-run, we believe Distill should be open to expanding to other disciplines, with new editors taking on different topic portfolios.
We had previously believed that, in exploring a new kind of publishing, Distill would be best served by focusing on a single “vertical” (machine learning) where it had editorial expertise.
Conclusion
Distill is a young journal exploring a new style of scientific communication. We have learned a lot of valuable lessons in our first year, but we still have a lot of room to grow.
We hope that you will join us in pushing the boundaries of what a scientific paper can be!
Distill is grateful to all the members of the research community who have supported it to date — our authors, reviewers, editors, members of the steering committee, every one providing feedback on GitHub, and, of course, our readers. We’re glad to have you with us!
Frequently Asked Questions
Q: What is the Distill journal?
A: Distill is an open-access scientific journal that uses a new style of communication, focusing on clarity, reproducibility, and interactivity.
Q: What are the goals of Distill?
A: Our goals are to create a new kind of scientific journal that makes it easier for readers to engage with complex ideas, and to build a community of authors, reviewers, and editors who share our values.
Q: How does Distill operate?
A: Distill operates through a GitHub repository, where articles are hosted, and peer review is conducted through the issue tracker.
Q: What kind of articles does Distill publish?
A: Distill publishes articles that are concise, clear, and focused on a specific idea or concept. We encourage authors to use interactive diagrams, in-browser notebooks, and other innovative formats to help readers engage with their research.
Q: How do I submit an article to Distill?
A: We encourage you to submit an article to Distill! Please visit our GitHub repository and follow the submission guidelines.
Q: Can I become an editor for Distill?
A: Yes, we are always looking for new editors who share our values and are passionate about creating a new kind of scientific journal. Please get in touch with us through our GitHub repository or via email.
Cybercriminals Brought to Justice: Russian Nationals Sentenced for Involvement in Dark Web Market and Ransomware Gangs
Available over the Tor network, Hydra was a bazaar that brokered not just drugs but also fake documents, cryptocurrency laundering services, and other illicit goods and services. Nine months after Hydra was taken down, authorities came for Bitzlato, a cryptocurrency exchange that laundered “a substantial portion of the cryptocurrency that Hydra received.” In all, authorities said, Bitzlato processed roughly $4.58 billion worth of cryptocurrency transactions. Anatoly Legkodymov, a then 40-year-old Russian national residing in China, was arrested by US authorities in the 2023 takedown.
Hydra Takedown and Bitzlato Shutdown
The takedown of Hydra and Bitzlato marked a significant blow to the dark web and cryptocurrency-related crime. However, the individuals involved in these illegal activities continued to operate, and their actions had far-reaching consequences.
Cybercriminals Sentenced for Involvement in Dark Web Market
The sentencing of Russian nationals for their involvement in the dark web market Hydra is a significant development in the fight against cybercrime. The defendants, including Moiseyev, Alexander Chirkov, Andrei Trunov, Evgeny Andreyev, Ivan Koryakin, Vadim Krasninsky, Georgy Georgobiani, Artur Kolesnikov, Nikolay Bilyk, Alexander Khramov, Kirill Gusev, Anton Gaikin, Alexey Gukalin, Mikhail Dombrovsky, Alexander Aminov, and Sergey Chekh, were handed down stiff sentences.
Russian National Linked to Ransomware Gangs Arrested
Mikhail Matveyev, a Russian national, was arrested by Russian authorities for his involvement in ransomware groups, including Babuk, Conti, DarkSide, Hive, and LockBit. US authorities have charged him with the creation or distribution of software intended to hack information systems.
US Sanctions and Bounty Offered
In May 2023, the US Treasury Department announced sanctions against Matveyev, and the US Department of State offered a bounty of up to $10 million for information leading to his arrest.
Conclusion
The stiff sentence of Moiseyev and the arrest of Matveyev are significant milestones in the fight against cybercrime. However, these developments are a rarity in Russia, where the government has historically turned a blind eye to the activities of Russian nationals involved in cybercrime, as long as they target individuals and organizations outside of Russia and its allied nations.
FAQs
Q: What was Hydra?
A: Hydra was a dark web market that brokered illegal goods and services, including drugs, fake documents, and cryptocurrency laundering services.
Q: What was Bitzlato?
A: Bitzlato was a cryptocurrency exchange that laundered a substantial portion of the cryptocurrency received by Hydra. It processed roughly $4.58 billion worth of cryptocurrency transactions.
Q: Who was arrested in connection with Bitzlato?
A: Anatoly Legkodymov, a then 40-year-old Russian national residing in China, was arrested by US authorities in the 2023 takedown.
Q: Who was Mikhail Matveyev?
A: Mikhail Matveyev is a Russian national linked to ransomware groups, including Babuk, Conti, DarkSide, Hive, and LockBit. He was arrested by Russian authorities and faces charges for the creation or distribution of software intended to hack information systems.
Q: What was the US Sanctions and Bounty Offered?
A: In May 2023, the US Treasury Department announced sanctions against Matveyev, and the US Department of State offered a bounty of up to $10 million for information leading to his arrest.
The Ultimate Thrifting Experience: Introducing Encore
If you’re a thrift shopper, you know the thrill of finding the perfect piece, especially if you weren’t looking for that specific item in the first place. Unfortunately, it’s not as easy online. Your searches are almost always limited to tracking down a specific item or items from a certain brand, and with at least half a dozen major players in the clothing resale market, it can be fairly time-consuming.
The Solution: Encore
That’s where Encore comes in. It’s an AI-powered clothing search engine that searches dozens of platforms and shows you exactly what you’re looking for on one screen. That is, if you know what you’re looking for — but if you don’t know what you want, it can help with that too.
The Experience
I decided to give Encore a try. My first few searches were fairly straightforward, ones that I thought would have worked as traditional Google searches. Asking Google to find "vintage Chicago Cubs jackets" showed mostly eBay results with a single Poshmark entry thrown in. Asking on Encore, though, showed a much wider variety of results from Mercari, Etsy, Depop, eBay, ThredUp, RealReal, Craigslist, and more.
Beyond Basic Searches
I tried again with brands, items of clothing, and general styles, and got incredibly specific results each time. But it’s not where Encore shines. Since the site uses a large language model (GPT-4o Mini), your query doesn’t have to be for a single item, or even for a specific kind of item. When I asked Encore to show me "vintage clothing pieces I could wear as a tour guide" (I’m a haunted history tour guide on the weekends), it returned with a list that ranged from an authentic 1890s wool coat priced at nearly $800 and Victorian-style jackets to period-appropriate replica hats and vests.
More Features
Encore offers options to "Shop from the show" and find "outfit inspo for." With the former, you can track down specific outfits worn on your screen. A prompt like "In Emily in Paris S4 E3, what was the dress she wore?" pulls up houndstooth pattern dresses similar to what the character wore. Asking Encore to "Shop from the show Suits, what were the suits Harvey wore in the first season?" showcases classic navy pinstripe suits.
Outfit inspo lets you give a scenario as a search prompt, like, "outfit inspo for a guy going to the US open," or "outfit inspo for a coffee date, minimalistic, neutral," or "outfit inspo for a night out in NYC, give me higher end brands, styles, and colors that go well."
Conclusion
Encore is a game-changer for thrift shoppers and fashion enthusiasts alike. With its AI-powered search engine and ability to find specific outfits, it’s the ultimate resource for finding unique and stylish pieces online.
FAQs
Q: What platforms does Encore search?
A: Encore searches dozens of platforms, including eBay, Etsy, Depop, Mercari, ThredUp, RealReal, and more.
Q: Can I use Encore for free?
A: Yes, the free version includes 2,000 daily product recommendations, a basic fashion model, and the ability to favorite items and view past searches.
Q: What are the benefits of upgrading to Encore Pro?
A: Upgrading to Encore Pro gets you unlimited recommendations, twice as many recommendations per search, the most advanced fashion model, the ability to find items by uploading images, and daily hidden gems.
Q: Does Encore make a commission off each purchase?
A: Yes, Encore makes a small commission off each purchase, but it doesn’t allow for sponsored ads, so search results are genuine.
GenCast: A New AI Model for Accurate Weather Forecasting
A Breakthrough in Weather Prediction
Google DeepMind has developed a new AI model called GenCast, which has been found to be accurate enough to compete with traditional weather forecasting methods. In a recent study, GenCast outperformed a leading forecast model, the European Centre for Medium-Range Weather Forecasts (ECMWF) model, in predicting the path of tropical cyclones and extreme weather events.
How GenCast Works
GenCast is a machine learning weather prediction model trained on weather data from 1979 to 2018. The model learns to recognize patterns in the data and uses that to make predictions about what might happen in the future. Unlike traditional models, which rely on supercomputers to solve complex equations, GenCast bypasses these equations and uses a more efficient approach.
Advantages of GenCast
GenCast has several advantages over traditional models. It can produce one 15-day forecast in just eight minutes, compared to several hours for traditional models. Additionally, GenCast operates at a lower resolution than traditional models, making it more computationally efficient.
Comparison to Traditional Models
GenCast was tested against the ECMWF model, which is one of the world’s top-tier models for forecasting. GenCast outperformed the ECMWF model 97.2% of the time, according to the study. However, the ECMWF model has since been upgraded to a higher resolution, making it difficult to compare the two models directly.
Sustainability
One concern about GenCast is its environmental impact. The model requires significant computational power, which can contribute to greenhouse gas emissions. However, GenCast’s efficiency could help reduce this impact.
Future Developments
While GenCast has shown promise, there are still improvements to be made. The model can be scaled up to a higher resolution, and it can produce predictions at 12-hour intervals, which could be more useful in real-world applications.
Expert Opinion
Stephen Mullens, an assistant instructional professor of meteorology at the University of Florida, expressed some skepticism about GenCast’s potential impact. "People are looking at it. I don’t think that the meteorological community as a whole is bought and sold on it," he said. "We are trained scientists who think in terms of physics… and because AI fundamentally isn’t that, then there’s still an element where we’re kind of wrapping our heads around, is this good? And why?"
Conclusion
GenCast is a significant breakthrough in weather prediction, and its potential to improve forecasts is vast. While there are still improvements to be made, GenCast’s efficiency and accuracy make it an exciting development in the field of meteorology.
FAQs
Q: What is GenCast?
A: GenCast is a machine learning weather prediction model developed by Google DeepMind.
Q: How does GenCast work?
A: GenCast is trained on weather data from 1979 to 2018 and uses a machine learning approach to make predictions about future weather patterns.
Q: Is GenCast more accurate than traditional models?
A: Yes, GenCast outperformed a leading forecast model, the ECMWF model, in predicting the path of tropical cyclones and extreme weather events.
Q: What are the advantages of GenCast?
A: GenCast is more computationally efficient than traditional models and can produce forecasts in just eight minutes.
Q: Is GenCast sustainable?
A: The environmental impact of GenCast is a concern, but its efficiency could help reduce this impact.
Q: What is the future of GenCast?
A: GenCast can be scaled up to a higher resolution and can produce predictions at 12-hour intervals, which could be more useful in real-world applications.
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Meta Develops AI Tool to Boost Productivity
Meta is pushing forward with plans to develop its artificial intelligence offering to businesses, as the social media platform rolls out a new AI tool internally that aims to boost productivity. The AI software, called Metamate, is built on the tech giant’s large language model, Llama, and is used for coding, conducting research, as well as drafting communications internally and externally.
Metamate’s Capabilities
The AI software is primarily a text-based interface and does not currently support video or images. It also lacks the more autonomous, agent-like features of its rivals, such as joining and summarizing meetings, scheduling items in calendars or taking actions on behalf of the user. However, Meta employees, including executives, said they used the tool regularly for different tasks, including coding, finding internal information from the company intranet, and preparing for client calls and follow-ups.
Impact on Productivity
Ratanchandani, Meta’s VP of engineering who leads Metamate’s development, said it was too early to measure its impact on productivity. However, Meta employees who use the tool reported that it has helped them to complete tasks more efficiently and effectively.
Market for AI Agents
The market for so-called AI agents, which can reason and complete complex tasks for people, is projected to grow from $5.1bn to $47bn by 2030, according to data from market research firm MarketsandMarkets.
Conclusion
Meta’s development of Metamate is a significant step towards its goal of becoming a dominant player in the AI market. While the tool is still in its early stages, it has the potential to revolutionize the way businesses operate and could have a significant impact on productivity.
FAQs
Q: What is Metamate?
A: Metamate is an AI software developed by Meta that is used for coding, conducting research, as well as drafting communications internally and externally.
Q: Is Metamate available for external use?
A: No, Metamate is currently only available for internal use at Meta.
Q: What is the market size for AI agents?
A: The market for so-called AI agents, which can reason and complete complex tasks for people, is projected to grow from $5.1bn to $47bn by 2030, according to data from market research firm MarketsandMarkets.
Q: When will Metamate be available for external use?
A: There is no current plan to release Metamate for external use, but the company is considering ways to improve the product, including making it more like an AI assistant.
Automated Audio Captioning: A Multi-Agent Approach to Enhance Performance
Introduction
The Automated Audio Captioning (AAC) task centers around generating natural language descriptions from audio inputs. Given the distinct modalities between the input (audio) and the output (text), AAC systems typically rely on an audio encoder to extract relevant information from the sound, represented as feature vectors, which a decoder then uses to generate text descriptions.
The Approach
Our approach employs multiple audio encoders, specifically BEATs and ConvNeXt, to generate complementary audio representations. This fusion enables the decoder to attend to a wider pool of feature sets, leading to more accurate and detailed captions.
Multi-Encoder Fusion
We employed two pre-trained audio encoders (BEATs and ConvNeXt) to generate complementary audio representations. This fusion enables the decoder to attend to a wider pool of feature sets, leading to more accurate and detailed captions.
Multi-Layer Aggregation
Different layers of the encoders capture varying aspects of the input audio, and by aggregating outputs across all layers, we further enriched the information fed into the decoder.
Generative Caption Modeling
To optimize the generation of natural language descriptions, we applied a large language model (LLM)-based summarization process, similar to techniques used in RobustGER. This step consolidates multiple candidate captions into a single, fluent output, using LLMs to ensure both grammatical coherence and a human-like feel to the descriptions.
Multi-Agent Collaboration
Our approach also involves a new multi-agent collaboration inference pipeline, inspired by recent research showing the benefits of nucleus sampling in AAC tasks. This pipeline consists of three stages:
CLAP-based Caption Filtering
We generate multiple candidate captions and filter out less relevant ones using a Contrastive Language-Audio Pretraining (CLAP) model, reducing the number of candidates by half.
Hybrid Reranking
The remaining captions are then ranked using our hybrid reranking method to select the top k-best captions.
LLM Summarization
Finally, we use a task-activated LLM to summarize the k-best captions into a single, coherent caption, ensuring the final output captures all critical aspects of the audio.
Impact and Performance
Our multi-encoder system achieved a Fluency Enhanced Sentence-BERT Evaluation (FENSE) score of 0.5442, outperforming the baseline score of 0.5040. By incorporating multi-agent systems, we have opened new avenues for further improving AAC tasks.
Future Work
We will explore integrating more advanced fusion techniques and examining how further collaboration between specialized agents can enhance both the granularity and quality of the generated captions.
Conclusion
Our contributions demonstrate the potential of multi-agent AI systems in advancing general-purpose understanding. We hope that our work inspires continued exploration in this area and encourages other teams to adopt similar strategies for fusing diverse models to handle complex multimodal tasks like AAC.
FAQs
Q: What is the main goal of Automated Audio Captioning (AAC)?
A: The main goal of AAC is to generate natural language descriptions from audio inputs.
Q: What are the key innovations in your approach?
A: Our approach employs multiple audio encoders, multi-layer aggregation, and generative caption modeling, as well as a new multi-agent collaboration inference pipeline.
Q: What is the significance of the multi-agent collaboration in your approach?
A: The multi-agent collaboration enables the system to leverage the strengths of each agent, resulting in more accurate and detailed captions.
Q: What is the potential impact of your work?
A: Our work has the potential to advance general-purpose understanding and inspire further exploration in multi-agent AI systems for complex multimodal tasks like AAC.
Q: How did you use NVIDIA technology in your work?
A: We used advanced NVIDIA computer technology, including the Taipei-1 supercomputer cluster and the NVIDIA DGX and OVX platforms, to accelerate our research and development.
01. Download the font from your chosen font provider
When browsing for a font, Adobe Fonts, DaFont, Creative Market, and MyFonts offer the best range of typefaces for beginners and professionals. Adobe Fonts is included with Adobe subscriptions and offers a large variety of professionally crafted fonts, all available for professional use. MyFonts and Creative Market are both font marketplaces where you can buy any font you need to add to your collection, and all are highly endorsed typefaces. DaFont offers over 40,000 free fonts with varied usages, where you can find famous, professional fonts as well as outrageous unnecessary typefaces that can still be quite fun. Once you have bought your font or identified the font you want to download, there will be a Download button clearly identifiable on the page.
02. Open the download
The font will appear in the Downloads icon in the top right-hand side of your browser tab. If it has not already popped up, open the Downloads icon, where you will see a folder titled after your font with a .zip extension. Double click on this folder to open it.
03. Extract the zip
The folder will now be opened in File Explorer. Click on it once to select it if it is not already selected. In the top bar of File Explorer in the same window, there will be a button called Extract All. Click on this to expand and open the zip file. This will open up a file containing one or more .otf files and a .txt file. The .otf files contain the font, and if there are several this means that the font has different weights and variations to install. The .txt file contains licensing information about the font. This is important to read to make sure you use the font according to the correct permissions.
04. Open Font Settings
In your computer’s Windows search bar, search for Font Settings to open the Font Settings app. This is the app on your computer that stores all your fonts, as well as where you can search for an already existing font or install a new one. Make sure this app is open.
05. Drag and drop
Hold down ctrl and click on each .otf file to select all of them if there are multiple. If there is only one, you do not need to hold down ctrl; instead, you just need to click on the file as normal. Drag your .otf files(/s) across into Font Settings where there is a section labelled Drag and Drop to Install. This should install your font and all its variations into your computer’s fonts app.
06. Check the font is installed
Now we need to check that the font is installed. In Font Settings, search for the name of the font and check that it comes up. If it doesn’t, then try the process again.
07. Check the license
Now is the time to read that .txt file we mentioned earlier on. It’s important not to overlook it so that you can ensure the font is licensed to be used in the places you intend to use it. Double click on the .txt file to open it and double check the license.
07. Make the most of your new font freedom!
If your new font appears in Font Settings, it has been successfully installed and is fully ready for use across your whole computer. Congratulations! Now you can enjoy the luxury of your own fully customisable font library.
Conclusion
By following these steps, you can successfully install a font on your Windows computer. Remember to check the license and ensure you have the correct permissions to use the font. With this new font, you can take your projects to the next level and add a personal touch to your work.
FAQs
Q: What are some good font marketplaces to find unique fonts?
A: Adobe Fonts, DaFont, Creative Market, and MyFonts are some of the best font marketplaces to find unique fonts.
Q: How do I know if the font is installed correctly?
A: Check the font in Font Settings to ensure it is installed correctly.
Q: What should I do if the font doesn’t appear in Font Settings?
A: Try the process again and ensure you have followed all the steps correctly.
Q: What is the purpose of the .txt file?
A: The .txt file contains licensing information about the font and is important to read to ensure you use the font according to the correct permissions.
Generative AI Bots and Your Data: What You Need to Know
ChatGPT
To disable AI training on ChatGPT, follow these steps:
On the web: Click your profile picture (top right), then choose Settings > Data control and turn off the Improve the model for everyone toggle switch.
Using the mobile app: Tap the menu button (top left), then the three dots next to your account name to find the Data controls screen and the Improve the model for everyone toggle switch.
Copilot
To disable AI training on Copilot, follow these steps:
On the web: Click your account picture (top right), then click your name and Privacy. You get two toggle switches you can turn off: Model training on text and Model training on voice.
Using the mobile app: These toggle switches are in an almost identical place in the Copilot mobile app. Tap your account picture (top right), then Account and Privacy.
Gemini
To disable Gemini AI training, you need to turn off your chat history. Follow these steps:
On the web: Click your profile picture at the top of the page, then Settings & Privacy > Data privacy to find the Data for generative AI improvement toggle switch.
Using the mobile app: To find the switch in the mobile app, tap your profile picture (top left), then Settings & Privacy > Data privacy > Data for generative AI improvement.
Meta AI
If you’re not in Europe or the UK, your options for avoiding having your Facebook data used for training are minimal. However, if you’re in Europe or the UK, you can object to this collection of data by submitting this well-hidden form, which is labeled as a "right to object" form. In the US, all you have is an alternative form "to submit requests related to your personal information from third parties being used to develop and improve AI at Meta." First, check off "I have a concern about my personal information from third parties that’s related to a response I received from an AI at Meta model, feature or experience." Then, on the form that appears, you need to explicitly explain (and provide screenshots) how your personal data was used.
Other apps
For other AI-using apps that you use, it’s worth digging into the settings and privacy policies to see exactly how your data is being processed. Policies can differ widely. Earlier this year, Adobe updated its privacy policy to confirm that it wouldn’t train its AI on user images. On the other hand, Reddit has signed a deal with OpenAI to train AI on user posts — and there’s nothing you can do about it, except not use Reddit.
Conclusion
As you can see, disabling AI training on various apps and services is not always straightforward. Some apps, like ChatGPT and Copilot, make it easy to turn off AI training, while others, like Gemini and Meta AI, are more obtuse. It’s essential to be aware of how your data is being used and to take control of your online presence. Remember to always read the fine print and to be cautious about what you share online.
FAQs
Q: What is AI training, and why is it a concern?
A: AI training is the process of using your data to improve the accuracy and effectiveness of AI models. This can include training AI to recognize and generate text, images, or other types of content.
Q: Why should I care about AI training?
A: You should care about AI training because it can have significant implications for your privacy and online presence. AI models can use your data to make inferences about you, which can be used for advertising, surveillance, or other purposes.
Q: How can I disable AI training on various apps and services?
A: To disable AI training, you need to check the settings and privacy policies of each app or service. Some apps, like ChatGPT and Copilot, make it easy to turn off AI training, while others may require you to submit a form or explain how your data was used.
Q: What can I do if I’m not happy with the way my data is being used?
A: If you’re not happy with the way your data is being used, you can try contacting the app or service directly and asking them to change their policies. You can also consider deleting your account or using alternative apps and services that are more transparent about their data practices.
Hackers Pocket $155,000 by Injecting Backdoor into Solana Code Library
Supply-Chain Attack
Hackers have made off with as much as $155,000 by sneaking a backdoor into a code library used by developers of smart contract apps that work with the cryptocurrency known as Solana.
Target: Solana-web3.js
The supply-chain attack targeted solana-web3.js, a collection of JavaScript code used by developers of decentralized apps (dapps) for interacting with the Solana blockchain. These dapps allow people to sign smart contracts that operate autonomously in executing currency trades among two or more parties when certain agreed-upon conditions are met.
Backdoored Code
The backdoor came in the form of code that collected private keys and wallet addresses when apps that directly handled private keys incorporated solana-web3.js versions 1.95.6 and 1.95.7. These backdoored versions were available for download during a five-hour window between 3:20 pm UTC and 8:25 pm UTC on Tuesday.
Assume Full Compromise
"This allowed an attacker to publish unauthorized and malicious packages that were modified, allowing them to steal private key material and drain funds from dapps, like bots, that handle private keys directly," stated a message posted to GitHub by Anza, the firm that develops the code library. "This issue should not affect non-custodial wallets, as they generally do not expose private keys during transactions."
Recommendations
Anza urged all Solana app developers to upgrade to version 1.95.8, which at the time this post went live on Ars, was the latest available. The company further encouraged developers who suspect they might have been compromised in the attack to rotate any suspect authority keys, including multisigs, program authorities, and server keypairs.
Solana Labs Statement
The same message was posted to social media by Solana Labs, a developer that has forked its original client.
Conclusion
The attack highlights the importance of vigilance in software development and the potential consequences of supply-chain attacks. Developers must ensure that they keep their software up to date and monitor their code for any suspicious activity to prevent such attacks in the future.
FAQs
Q: What is a supply-chain attack?
A: A supply-chain attack occurs when an attacker injects malicious code into a software library or framework used by multiple applications, allowing them to compromise multiple systems at once.
Q: What is solana-web3.js?
A: solana-web3.js is a collection of JavaScript code used by developers of decentralized apps for interacting with the Solana blockchain.
Q: How much money was stolen in the attack?
A: Hackers made off with as much as $155,000.
Q: What should I do if I suspect I have been compromised in the attack?
A: Rotate any suspect authority keys, including multisigs, program authorities, and server keypairs, and upgrade to the latest version of solana-web3.js.
/cdn.vox-cdn.com/uploads/chorus_asset/file/25535556/STK160_X_TWITTER__C.jpg "X gives Grok a new photorealistic AI image generator")
/cdn.vox-cdn.com/uploads/chorus_asset/file/25777241/1175363196.jpg "Google’s AI weather prediction model is pretty darn good")
/cdn.vox-cdn.com/uploads/chorus_asset/file/25461500/HT054_AI_general_3.jpg "Prevent AI from Training with Your Data")
