Search the Community

Microsoft has signed deal with CoreWeave for AI computing power, say sources by Paul Hill Microsoft has signed a deal with CoreWeave, a provider of AI computing power, that could be worth billions of dollars over a number of years. The news was disclosed by sources familiar with the matter, to CNBC. While Microsoft and CoreWeave have failed to confirm the information, the sources told CNBC that the deal was made to ensure that OpenAI’s ChatGPT had enough computing power going forward. Through a partnership with Microsoft, OpenAI currently uses Microsoft Azure infrastructure to run ChatGPT, which is resource intensive. It seems that the agreement was made earlier this year. In recent weeks, we’ve seen the price of NVIDIA’s shares rocket up as investors anticipate higher earnings for the company on the back of generative AI services like ChatGPT. CoreWeave offers cloud computing services also powered by NVIDIA hardware. The revelation about this deal comes just one day after CoreWeave announced that it had secured $200 million in a Series B funding extension, bringing the round’s total to $421 million. The $200 million was invested by Magnetar Capital. According to CoreWeave, the funding is helping it fill a gap in the market that legacy cloud computing providers are struggling to fill. “By combining easy access to high-powered GPUs for training AI models with fast and flexible infrastructure and by focusing on a specific type of compute, CoreWeave continues to differentiate itself from other companies in the space,” said Ernie Rogers, Magnetar’s chief operating officer. “Magnetar believes CoreWeave sits in a sweet spot for enabling world-class results across a number of industries. We are proud to have been the lead investor for CoreWeave’s Series B funding round and its extension.” Now that knowledge of this agreement is public, it could add even more fuel to NVIDIA’s stock price as it suggests the company could see even more demand for its products as CoreWeave seeks to provide resources for Microsoft and OpenAI. Source: CNBC

OpenAI develops LLM that uses a chain of thought like humans by Paul Hill OpenAI has released a new paper outlining some advancements it has made in eliminating the common problem of hallucinations where AI just makes stuff up. The paper outlines two models called outcome supervision and process supervision to weed out hallucinations and how they perform. With outcome supervision, OpenAI trains reward models to provide feedback on the final result the AI gives. With process supervision, the reward model provides feedback at every step of the way, creating a human-like chain of thought. In its research paper, OpenAI tested both models on a math dataset and found that the process supervision method led to “significantly better performance”. It’s important to note that the process supervision method has only been tested in the area of mathematics so far and that it will take more work to see how it performs more generally. Explaining the possible outcomes of the process supervision method, OpenAI said: “If these results generalize, we may find that process supervision gives us the best of both worlds – a method that is both more performant and more aligned than outcome supervision.” It’s still too early to say how much this step-by-step verification will help to address hallucinations more generally, but hopefully, it will because hallucinations are probably the number one issue with LLMs right now. Just this week, a lawyer that had used ChatGPT for his work and submitted false information detailing fake cases that the AI had dreamt up. OpenAI has not given a timeline for how long it will take to implement process supervision in ChatGPT which is available to the public. It’s still in the research phase and needs to be tested on general information. While initial results are good, OpenAI does mention that safer methods can incur reduced performance called an alignment tax. The results show so far that process supervision doesn’t incur this tax while working on math problems but we don’t know what will happen on more general information.

A lawyer used ChatGPT for legal research, but later found the chatbot created fake cases by Omer Dursun In a recent court case, a lawyer relied on ChatGPT for legal research, resulting in the submission of false information. The incident sheds light on the potential risks associated with AI in the legal field, including the propagation of misinformation. The case revolved around a man suing an airline over an alleged personal injury. The plaintiff's legal team submitted a brief citing several previous court cases to support their argument, seeking to establish a legal precedent for their claim. However, the airline's lawyers discovered that some of the referenced cases did not exist and promptly alerted the presiding judge. Judge Kevin Castel, presiding over the case, expressed astonishment at the situation, labeling it an "unprecedented circumstance." In an order, the judge demanded an explanation from the plaintiff's legal team. Steven Schwartz, a colleague of the lead attorney, confessed to utilizing ChatGPT to search for similar legal precedents. In a written statement, Schwartz expressed deep regret that he "had never previously used AI for legal research and was unaware that its content could be false." Screenshots attached to the filing showed a conversation between Schwartz and ChatGPT. In the prompt, Schwartz asked if a specific case, Varghese v. China Southern Airlines Co Ltd, was genuine. ChatGPT affirmed its authenticity, indicating that the case could be found in legal reference databases such as LexisNexis and Westlaw. However, subsequent investigations revealed that the case did not exist, leading to further doubts about the other cases provided by ChatGPT. In light of this incident, both lawyers involved in the case, Peter LoDuca and Steven Schwartz from the law firm Levidow, Levidow & Oberman, have been summoned to an upcoming disciplinary hearing on June 8 to explain their actions. This event has prompted discussions within the legal community regarding the appropriate use of AI tools in legal research and the need for comprehensive guidelines to prevent similar occurrences. Source: NYT

Microsoft Bing search is being added to ChatGPT; new Bing Chat plugin partners revealed by John Callaham As part of the Microsoft Build 2023 developers conference, the company has announced even more features for its Bing search and Bing Chat services. The biggest reveal is that Bing search will soon be incorporated directly into OpenAI's ChatGPT chatbot AI. Of course, Microsoft is a major financial and development partner with OpenAI in ChatGPT, and Microsoft has been using OpenAI's tech for its Bing Chat bot since it launched earlier this year. In a blog post, Microsoft stated: ChatGPT will now have a world-class search engine built-in to provide timelier and more up-to-date answers with access from the web. Now, ChatGPT answers can be grounded by search and web data and include citations so you can learn more—all directly from within chat. Subscribers to ChatGPT Plus will get first access to the new Bing search features, but it will be enabled for free users in the near future via a plugin that will add Bing search to the chatbot. Speaking of plugins, Microsoft announced earlier in May that Bing Chat would support third-party plugins. Today at Build 2023, the company confirmed that developers would be able to develop plugins that would work on several different platforms, including Bing Chat, ChatGPT, Dynamics 365 Copilot, Microsoft 365 Copilot, and Windows Copilot. Microsoft also announced new Bing Chat plugin partners. They include: Expedia Instacart Kayak Klarna Redfin TripAdvisor Zillow Bandsintown Bohita Cloudflare Coupert Fareportal FiscalNote Golden Lexi Shopper Likewise Notable One Word Domains PromptPerfect Shopify Skyscanner Spotify Spotnana Trip.com Microsoft said these new plugins will be available in Bing Chat, or in the Bing sidebar on Edge, in the coming weeks on both desktop and mobile platforms. The company also commented that Bing Chat plugins will be incorporated into the newly announced Windows Copilot and into the Edge web browser in the near future. In case you want to read more, you can find the rest of the Build 2023 coverage here.

Google search rival Neeva is shutting down its search engine to work on AI products by John Callaham Neeva, a search engine company that was founded in 2019 by a former executive at Google, announced on Saturday it is shutting down the search engine part of the business. It will instead concentrate its efforts on developing AI and large language models (LLMs). In a blog post, the company stated it discovered that it was difficult to make a new search engine with just 50 team members. It stated: We overcame these obstacles and built a search stack from the ground up, running a crawl that fetched petabytes of information from the web and used that to power an independent search stack. Neeva decided not to use ads for its search engine, unlike Google's business model. This is ironic because the person who founded Neeva, Sridhar Ramaswamy, was previously Google's senior vice president of advertising and commerce. Instead, Neeva used a subscription model to fund development, which cost $5.99 a month or $49.99 for an annual plan. For that price you could use its search browser extension for Chrome, Edge, Firefox, Safari, and others, along with iOS and Android apps. The subscription also threw in a password manager and a VPN with that plan. Neeva officially launched in 2021 in the US and expanded to other countries in 2022. In January 2023, the company launched NeevaAI, which beat Google's Bard and Microsoft's Bing Chat to the punch by offering AI answers to search questions. It expanded to other countries in February just as Bard was announced and Bing Chat launched. However, it would appear that wasn't enough to keep Neeva's search business going. This weekend's blog post stated: Contrary to popular belief, convincing users to pay for a better experience was actually a less difficult problem compared to getting them to try a new search engine in the first place. Neeva will shut down its consumer search engine in the coming weeks, along with the Neeva.com site. All of its user data will be deleted, and people who signed up for Neeva Premium will receive a refund. More information has been posted on an FAQ page. Neeva's blog post states that the company itself will go on. It says it wants to help develop LLMs more effectively and more affordably for enterprise users. It added that it will reveal more "future of our work and our team in the next few weeks."

Apple tells employees not to use AI chatbots like ChatGPT over confidential data leak fears by Paul Hill Apple has told its employees not to use generative AI tools such as ChatGPT and GitHub Copilot according to The Wall Street Journal which got the information from people in the know. Apple is not the only big tech firm to take such actions, Samsung has also banned its employees from using generative AI chatbots. Apple apparently told employees that using these chatbots could cause the accidental release of confidential information. While most people are familiar with ChatGPT and what it does, you may not have used GitHub Copilot. First of all, GitHub is owned by a major Apple competitor, Microsoft. With Copilot, users can automate some of their software development and Apple is concerned that Microsoft could intercept secret Apple code to see what it’s working on or just copy the products. Luckily for Apple employees who want to delegate jobs to AI, Apple is working on its own generative AI product, according to the report. It’s not clear whether Apple employees are able to use this internally yet but as soon as that goes live, there will be no need for them to resort to products like ChatGPT. Apple is due to hold its WWDC developer conference early next month. The company is expected to reveal its mixed-reality headset and it wouldn’t be too much of a surprise if we did see some sort of generative AI at least demoed. It seems like a lot of tech firms have been working on generative AI for a while now and have been quick to launch their own products; Apple could be in the same position. Speaking of WWDC, it was reported a few days ago that Apple could unveil sideloading apps on iOS. This feature has been on Android for a long time, if not from the beginning. It's definitely going to be interesting to see Apple open up iOS a bit more. Source: The Wall Street Journal

Twitter has sent a letter to Microsoft claiming it used data without paying for it by Paul Hill Twitter has sent a letter to Microsoft CEO Satya Nadella complaining that Microsoft has been improperly using Twitter data, according to The New York Times which saw the letter. Twitter has accused Microsoft of not paying to use data, using data more than had been agreed, and sharing the data with government agencies without Twitter’s permission - this could be concerning if authoritarian governments were the recipients. Microsoft spokesman, Frank Shaw, told NYT that Microsoft doesn’t pay Twitter for its data. He said the company will review the letter from Twitter and respond accordingly. Perhaps reassuringly, Shaw said that Microsoft looks forward to continuing its long-term partnership with Twitter. In a recent interview with CNBC, Elon Musk, the owner of Twitter, said that the platform was like a plane that was on fire about to crash, financially speaking. In an effort to stop losing money hand-over-fist, Musk quickly fired about half of the employees at the company. He now admits it was very hasty but that decision had to be taken quickly. He said the company could probably do with getting some of those people back if they’re not too annoyed, he said. With the big splash ChatGPT and other generative AI services have made, platforms like Twitter and Reddit have said they want compensation for their data being used to train the bots. Musk has said that Microsoft essentially runs OpenAI now, or at least has significant power over the company. OpenAI’s ChatGPT powers Microsoft’s Bing Chat. Given Microsoft's involvement and the huge amount of money it has, this could explain why Musk is going after Microsoft for payment for the training data. The letter doesn’t say if Twitter will take legal action or how much financial compensation it has demanded but says that Microsoft uses Twitter data in several of its products including Xbox, Bing, advertising, and cloud computing. It wants Microsoft to submit a report by next month about how much Twitter data it uses and which governments have gained access to the data. It also told Microsoft to abide by the Twitter developer agreement in its handling of the data. Source: The New York Times

OpenAI launches ChatGPT app for iPhone users by Rahul Naskar A couple of months ago, we heard rumors of OpenAI working on the ChatGPT app for mobile users. Luckily, it's no longer a rumor, as OpenAI has recently announced the ChatGPT app for iPhone users. However, only US users can download the app on their iPhones for now. ChatGPT is a free app on iOS, but you will need a Plus subscription to benefit from GPT-4's capabilities. As is the case with ChatGPT on the web, Plus subscribers will also get access to new features earlier than normal users on the iPhones. Plus, the app syncs history across devices. Long story short, iPhone users get the same experience as on the web. As highlighted by OpenAI, the ChatGPT app for iPhone has the following features: Instant answers: Get precise information without sifting through ads or multiple results. Tailored advice: Seek guidance on cooking, travel plans, or crafting thoughtful messages. Creative inspiration: Generate gift ideas, outline presentations, or write the perfect poem. Professional input: Boost productivity with idea feedback, note summarization, and technical topic assistance. Learning opportunities: Explore new languages, modern history, and more at your own pace. OpenAI has promised that it will expand to markets beyond the US in the coming weeks. Besides iPhone users, the ChatGPT app will also be available for Android users "soon." Meanwhile, you can download the ChatGPT app on your iPhone here from App Store. ChatGPT Plus subscription with GPT-4 capabilities costs $20 per month. Microsoft's ChatGPT-powered Bing AI chatbot also incorporated GPT-4, and the app is available on App Store for free. You can download the New Bing app here from App Store. Source: OpenAI

Elon Musk on Sam Altman and ChatGPT: I am the reason OpenAI exists by Ishtiaqe Hanif Elon Musk, one of the three co-founders of OpenAI, questioned the changing business practices of the company during an interview with David Faber at Tesla's Giga Texas factory. He said, "I came up with the name" and "I am the reason that OpenAI exists.". Musk claimed the non-profit startup wouldn't have come to where it is today without his instrumental role. Initially committing to invest $1 billion in backing, Musk invested roughly $50 million in the startup. He resigned from the board of OpenAI after a dispute over his attempt to buy out the startup and a conflict of interest. Since then, it switched to a for-profit entity and accepted a $1 billion infusion from Microsoft to accelerate the development of ChatGPT and image generator DALL-E. Musk questions transitioning from a non-profit business to a $30 billion limited-profit company. I’m still confused as to how a non-profit to which I donated ~$100M somehow became a $30B market cap for-profit. If this is legal, why doesn’t everyone do it? — Elon Musk (@elonmusk) March 15, 2023 He was also skeptical about the governance structure, policies and revenue plans of OpenAI after turning into a closed-source entity. The move conflicts with its name, which suggests it's open-source. In a tweet, he revealed it had access to the Twitter database for training, which he put a stop to. After the launch of ChatGPT, he remarked it was scary good, and it can quickly turn dangerous. Sam Altman, the CEO of OpenAI, replied to his tweet, agreeing about the potential cybersecurity risk and being close to getting real AGI within the next decade. This interview happened on the same day Sam Altman found addressing a US Senate panel on the rise of generative AI and its possible effects on many different industries suggesting AI needs regulation. Source: CNBC

OpenAI CEO Sam Altman tells US Senate panel generative AI needs regulation by John Callaham OpenAI founder and CEO Sam Altman found himself in Washington DC, addressing a US Senate panel on the rise of generative AI and its possible effects on many different industries. Altman told the panel that such AI systems, like his company's own ChatGPT, need to be regulated by the government. Reuters reports that Altman believes the government should require some kind of licensing and testing requirements for AI systems. He also stated that the use of AI to interfere with upcoming elections by users who may create realistic-looking false images or video is a "significant area of concern." That was brought home during the start of the hearing by US Senator Richard Blumenthal. Bloomberg posted a video of Blumenthal who played a recording of his voice talking about the effects of AI. However, he revealed that the recording was actually made by an AI that replicated his voice that was trained via his previous US Senate speeches. Even that recording's remarks were written by an AI program. CNN reports that after that recording was made, Blumenthal said that AI replicated voice could have also created false "endorsement of Ukraine’s surrendering or Vladimir Putin’s leadership." Altman also said at the panel that businesses should have the right to ban their content from being used by AI training models. He also stated that he preferred AI use a subscription model, like OpenAI's own ChatGPT Plus, rather than an ad-based model, which is being used by Microsoft's Bing Chat.

We interview Stardock's Brad Wardell about Galactic Civilizations IV: Supernova and ChatGPT by John Callaham Stardock CEO Brad Wardell started working on the first Galactic Civilizations game in 1993 as the company's first software product for the now defunct OS/2 operating system. Since then the 4x strategy game has gone through several versions on Windows. This year, Stardock announced Galactic Civilizations IV: Supernova, a standalone expansion of Galactic Civilizations IV with tons of new features. However, most people didn't expect one of those features was incorporating OpenAI's ChatGPT technology into the game. Supernova uses ChatGPT to allow players to create their own alien civilizations with just some text prompts. We got in touch with Wardell by email to ask him some more questions about Galactic Civilizations IV: Supernova, and how he thinks ChatGPT will cause Stardock to actually bring in more team members to develop games. The Galactic Civilizations series has been ongoing for 30 years now. What are your thoughts about continuing to make games in this franchise for such a long time? Brad Wardell - I’ve long felt that as long as there were new and interesting things to try out in the 4X genre, I’d keep at it. The impetus for me, 30 years ago, was to try and see if you could make a multithreaded AI. An AI that would calculate in the background. Now, here we are, using a large language model like ChatGPT to give players unique, personalized experiences. Do you believe the series has served as an inspiration for other 4x games over the decades? Brad Wardell - Certainly. I’ve borrowed heavily from other games myself, not the least being the Civilization series. It’s always validating when a concept we came up with shows up in other games. When it came to making the Supernova standalone expansion what were the dev team's main goals? Brad Wardell - I would have preferred to make it a normal expansion. But, as we got in there, we realized we would need two copies of almost every data file to do that. What really exploded the scale of this “expansion” has been the rise of AI. What started out as a fairly normal expansion process became something different when we realized we could use AI to create a real-time, living universe for players. The integration with ChatGPT is one of the first such additions to a major PC game. How did the idea come about to add those features in Supernova? Brad Wardell - Stardock has been working on multi-parameter AI for the past several years in another title we’re developing. So, we’ve been keeping up to date with OpenAI’s work for some time. As soon as they had an API we could use, we realized there was an opportunity to vastly increase the amount of content available for players and create a lot of new, emergent gameplay. How hard or easy was it to incorporate the text prompts for ChatGPT so it could make new civilizations and quests for the game? Brad Wardell - It took quite a bit of time. For one thing, we didn’t put it directly into the game. It’s hosted on our cloud services tech called Tachyon. This way, any of our games can make API calls to Tachyon and then it can use whatever is the best tech to deliver what is needed. On the art side, we also had to make sure that the art was being trained on Stardock art and not, say, art from ArtStation, as we have a lot of artists here who have strong feelings about where the training data comes from. Another challenge is making sure that the GPT results weren’t purely cosmetic. We want to be able to generate in-game quests and have it affect relationships. We want it to affect even what the ships look like, so that took quite a bit of time as well. What are some of the more surprising things you have learned or encountered about adding ChatGPT into Supernova? Brad Wardell - How well it works. We didn’t really expect the writing results to be as good as they’ve been. It helps that ChatGPT is very familiar with GalCiv and its style of writing already so that results tend to fall into our existing, slightly humorous, writing style right away. In a forum post, you stated that adding ChatGPT into Supernova is actually causing Stardock to hire more people, not less. Can you go into specifics on that? Brad Wardell - Indeed. It’s changing the types of people we’d be hiring. But let me walk you through the business reason for this, as I think many people haven’t really thought through the issue to its logical conclusion. In game development, you need to have a critical mass of X in order for it to be worthwhile. People who think AI generated art or AI generated writing is going to eliminate all the jobs haven’t taken the critical mass feature threshold into account. AI will displace jobs, but that’s a different discussion. For example, in the past, we just couldn’t justify many features or classes of content because even if we had 3 artists, they could only produce, say, 3 units of content. Therefore, we wouldn’t hire the artists. But now, those 3 artists can produce 300 units of content via AI assist. That amount of content exceeds the threshold needed for a feature to be worthwhile to have in the game and thus, 3 new art positions are created. The same is true with writing. There’s now enough written content in the game to justify contracting a dedicated editor (and God knows, we’ve needed text editing for years). But, you need a lot of writing before you can make the business case of needing to hire writers and editors. The AI has allowed us to reach that threshold. What other features do you consider to be important in Supernova? Brad Wardell - A big part of it is the accessibility. We want to make GalCiv IV: Supernova have a lot more depth than GalCiv III (or even GalCiv IV base game on Epic). And not to harp on the AI stuff again but as a practical matter, without AI assist tech like GitHub Copilot and our own internal GPT servers, a lot of these features would just be well beyond the scope of our team. So, for example, Supernova has a new combat system, new planet management, cultural progression, minor raced system, star system upgrades, and core world upgrades, all with a UI designed to make it accessible for players new to the genre. How will the game evolve during the Early Access period and when do you think it will be available for full access? Brad Wardell - We are targeting late Summer or early Fall for the full release. We have a lot of features we are still looking to get into it. Our measure of success is if GalCiv IV: Supernova becomes the go-to game for turn-based 4X players. What can you tell us about Stardock's current and upcoming game lineup, and is it possible those games will add ChatGPT features as well? Brad Wardell - We have some unannounced games that make use of our multiparameter AI tech. I can’t say more than that right now. :) Finally, is there anything else you want to say about Galactic Civilizations IV: Supernova? Brad Wardell - Galactic Civilizations IV: Supernova represents a big step forward for the franchise. We can’t wait for people to try it out and for fans to see that we’ve been listening and writing down their ideas for the past several years since we released GalCiv III. The last GalCiv game I got to be this involved with was GalCiv II: Twilight of the Arnor, so on a personal note, this is a big treat for me. We want to thank Brad Wardell for taking time to answer our questions. The game is available in Early Access for Steam and the Epic Games Store for $39.99. If you have already bought the original Galactic Civilizations IV on the Epic Games Store, you can get Supernova for a big discount. Disclaimer: Neowin's relationship to Stardock

OpenAI brings 70+ ChatGPT plugins and web search to all ChatGPT Plus users by Tushar Mehta In their initial leg, generative AI tools like ChatGPT have begun to shoulder much of our grunt work. Despite its impressive (and invariably eerie) conversational abilities, ChatGPT draws information from pre-curated and outdated data instead of fetching it from the internet in real-time. This may soon end, as OpenAI now announces support for web browsing on ChatGPT. In addition, ChatGPT plugins are now more widely available. OpenAI recently updated its release notes for ChatGPT, announcing the chatbot is now getting internet access. Meanwhile, plugins — or tools which allow you to access third-party services through ChatGPT — are now moving from the alpha to beta phase and will now be available for all ChatGPT Plus users, i.e. those who pay to use extra features in the chatbot. Notably, plugins were first made available to insiders back in March 2023 with services including OpenTable, Zapier, Expedia, Instacart, KAYAK, Shopify, and more. OpenAI says ChatGPT Plus users will now have more than 70 plugins to use, and Twitter user @DataChaz shares a list of available applications. Source: OpenAI If you're a ChatGPT Plus subscriber, you can access web browsing and plugins by heading to ChatGPT Settings from the bottom left corner of the chatbot interface. Under Settings, go to Beta features and enable Web browsing and Plugins. After you enable the toggles, you will find Plugins accessible from the top of the chat window, where you can install the desired extension, as demonstrated by Twitter user @DataChaz: 58 new #ChatGPT plugins have now landed in the @OpenAI store! 🔥 Here's a quick overview ⬇ Full list in 🧵 pic.twitter.com/XqvxkbBkz5 — DataChazGPT 🤯 (not a bot) (@DataChaz) May 12, 2023 OpenAI does not explicitly reveal when users in the free tier will have access. If you're a developer and want your application(s) to be accessible through ChatGPT, you can join the waitlist here. The announcement comes roughly a week after Microsoft opened up its AI chatbot for a public preview. The company also announced significant updates to its Bing AI chatbot, including image generation, support for images and videos besides text as prompts, a persistent Edge sidebar for AI functionality, and more. Additionally, Microsoft said plugins are coming to Bing chat soon, and details will be announced at the upcoming Microsoft Build conference scheduled from May 23 to 25.

OpenAI no longer uses API customer data to train its LLMs by Paul Hill Sam Altman, the CEO of OpenAI, has confirmed to CNBC that the company no longer uses API customer data to train its large language models. OpenAI updated its Terms of Service to reflect this at the start of March but didn’t make a song and dance about it. If you use ChatGPT directly, this data will still be used for training, unless you go incognito. In an interview, Sam Altman told CNBC that customers “clearly want us not to train on their data, so we’ve changed our plans: We will not do that.” Unfortunately for those using ChatGPT directly, this is not the case by default. The collection of data is such an issue Samsung has banned employees from using chatbots like ChatGPT over security leaks. As an entirely new category of software, companies like OpenAI as well as wider society are still getting to grips with the best practices. Earlier today, Neowin reported on the fact that the Competition and Markets Authority was going to start investigating how these generative AI products could affect competition and consumers. Another way in which these bots have had to retrospectively be improved upon is in relation to guard rails. Since their launch, they’ve been adapted a little bit to ensure they don’t say offensive things. When users try to get the bots to say something offensive, the bots recall pre-written scripts letting the user know they can’t help with that request. Source: CNBC

UK competition watchdog launches review into generative AI by Paul Hill The UK’s Competition and Markets Authority (CMA), which recently blocked Microsoft’s purchase of Activision Blizzard, is now launching an initial review of “artificial intelligence models” such as those that power Bing Chat and ChatGPT. The CMA launched the review after the government asked regulators to look into how AI will affect consumers, businesses, and the UK economy. Through the initial review, the CMA wants to look at three things in particular. It wants to see how these AI models could evolve. It wants to look into the opportunities and risks for competition and consumer protection. Finally, it wants to set out principles to support competition and protect consumers. Various regulators in the UK will be looking into how AI affects their target area. AI touches on several important issues such as safety, security, copyright, privacy, and human rights. Take for example the replacement of writers by AI, well it’s not so simple a matter of asking the AI to spit out an article for you. In terms of copyright, the output generated by the AI actually belongs to the company that makes the AI. While Google or OpenAI won’t come after you for cheating on your homework, they very well may come after people using outputs for commercial purposes. All of these various issues will be investigated by the UK’s different regulators but the CMA will focus more narrowly on the implication AI has for competition and consumer protection. “AI has burst into the public consciousness over the past few months but has been on our radar for some time. It’s a technology developing at speed and has the potential to transform the way businesses compete as well as drive substantial economic growth,” said Sarah Cardell, Chief Executive of the CMA. “It’s crucial that the potential benefits of this transformative technology are readily accessible to UK businesses and consumers while people remain protected from issues like false or misleading information. Our goal is to help this new, rapidly scaling technology develop in ways that ensure open, competitive markets and effective consumer protection.” The CMA is now seeking evidence from stakeholders (basically, anyone who may be affected) until June 2. After it has collected these insights and done its own analysis, it will publish a report with its findings in September. If you’d like to find out more or track the development of this work, head over to the initial review webpage.

Mozilla acquires fake review identifying startup Fakespot by Ishtiaqe Hanif The developers of Mozilla recently announced it acquired Fakespot, a startup whose browser extension can detect fake product reviews on shopping sites like Amazon. People rely on reviews to make shopping decisions online and loads of positive reviews on one product can sway decisions in its way. Many sellers use paid reviews to boost their product ratings and rankings and Consumer Choice reports companies are not doing enough to fight back against scammers. Integrating the technology in Firefox will go toward helping users to shop with the confidence of knowing fake reviews will be flagged. Since the rise of ChatGPT usage, bad actors have started using it to flood the market with realistic fake reviews to better fool buyers. It's more challenging now for an average consumer to identify a manufactured review due to the (mostly) good wording of the AI chatbot. However, as Techradar points out in its report, AI bots also have a tendency to not hide the fact that an AI wrote them, with one literally writing: “As an AI language model, I don’t have a body” in the text of a review. Mozilla plans to invest more in Fakespot and claims it will enhance the online shopping experience for millions of people. According to Mozilla, the extension will continue working on other platforms and browsers like Chrome on Windows, Android, iPadOS, and iOS. Addinjg that with its goal to put customers first, it will also develop features that will stay exclusive to Firefox browsers to identify authentic products with real reviews and ratings. On the company's site, the founder of Fakespot, Shouid Khalifah writes, “ We are joining a company that develops one of the most popular browsers in the world in Firefox with a lineage that dates back to the origins of the internet.” Mozilla shared in a press release, that they are continuing their work around ethical AI and responsible advertising. Customers have higher satisfaction when using Fakespot according to their data. Fewer product returns also mean less environmental damage through reduced shipping and packaging. Source: Techradar via Mozilla

AI systems from Microsoft, Google, OpenAI, and others to be tested at DEFCON 31 in August by John Callaham The rise of generative AI is causing many people to be concerned that they won't behave in ethical and responsible ways. Indeed, one of Google's highest-ranking AI researchers, Geoffrey Hinton, recently left the company so he could be free to warn the world about the potential dangers of AI. Today, US Vice President Kamala Harris will be meeting with the CEOs of four of the leading companies in AI development: Microsoft's Satya Nadella, Google's Sundar Pichai, OpenAI's Sam Altman, and Anthropic's Dario Amodei. The White House says the participants will discuss the growing use of AI and "the importance of driving responsible, trustworthy, and ethical innovation with safeguards that mitigate risks and potential harms to individuals and our society." The Biden Administration also announced that those four companies, along with Hugging Face, NVIDIA, and Stability AI, have all agreed to have their AI systems tested at the annual DEFCON hacking conference. DEFCON 31 is scheduled to be held in Las Vegas from August 10-13. The White House press release stated: This independent exercise will provide critical information to researchers and the public about the impacts of these models, and will enable AI companies and developers take steps to fix issues found in those models. Testing of AI models independent of government or the companies that have developed them is an important component in their effective evaluation. In addition, the US Office of Management and Budget will create a draft document for a policy on how the US government should use AI, and will offer it for public comment. Also, the National Science Foundation will add $140 million for funding of seven new National AI Research Institutes, which will bring the number of those groups to 25.

Microsoft Bing Chat is in open preview, and will add chat history, visual search and more by John Callaham Over the past couple of weeks, Microsoft has been fairly quiet with its updates on its Bing Chat service. However, there have been hints that the company was preparing some big announcements for its AI chatbot. Today, the company revealed a ton of new and upcoming features for Bing Chat and its Edge web browser. That includes the news that Bing Chat is now in full Open Preview mode. While Microsoft seemed to have eliminated the Bing Chat waitlist a couple of months ago, the company has now officially stated that there is indeed no more waitlist to try out the chatbot. All you need is a Microsoft Account and you are all set to use Bing Chat on mobile or via Edge on the desktop. Microsoft claims that there are now over 100 million daily active users for Bing Chat, with over half a billion total chats since its launch. It also says that daily installs of the Bing mobile app have gone up four times more than normal since Bing Chat was first introduced. One of the most requested features by Bing Chat users is a way to save previous chats. Today, Microsoft said that ability will be put in "shortly", allowing users to stop a chat in mid-session and then return again where the chat was paused. It added: And when you want to dig into something deeper and open a Bing chat result, your chat will move to your Edge sidebar, so you can keep your chat on hand while you browse. Over time, we’re exploring making your chats more personalized by bringing context from a previous chat into new conversations. Microsoft will also be adding ways to quickly export chat conversations and also share them on social media. It stated: For times when you want to easily share your conversation with others in social media or continue iterating on a newly discovered idea, you can export it directly – the format stays the same to make an easy transition to continue in collaborative tools like Microsoft Word. In March, Microsoft launched Bing Image Creator, a way to use text prompts to create art with generative AI. Since then the company claims that over 200 million images have been generated with the AI program. Today, the company said that Bing Image Creator is expanding its reach. Previously it only supported English, but now the AI art maker can now accept text prompts from all of Bing's over 100 supported languages. The company added it is working to add visual searches to Bing Chat, so you can upload images in the chat to search for content related to it in Bing. It will also add visual elements for searches in chat, including things like charts and graphs. Microsoft also revealed some changes coming to its Edge browser in relationship to Bing: Edge mobile will also soon include page context, so you can ask questions in Bing chat related to the mobile page you're viewing. The compose feature in sidebar can also now tailor drafts based on feedback you give like tone, length, phrasing and more. Bing Chat will also offer improvements for summaries of large articles and documents within Edge, and will introduce what Microsoft calls Edge actions: Available in the coming weeks, people will soon be able to lean on AI to complete even more tasks with fewer steps. For example, if you want to watch a particular movie, actions in Edge will find and show you options in chat in the sidebar and then play the movie you want from where it’s available. Microsoft adds that it will continue to monitor the use of Bing Chat to make sure that its safeguards are working for the ethical use of its AI tools. It stated: Together with our partners at OpenAI, we’ve continued to implement safeguards to defend against harmful content based on what we’re learning and seeing in preview. Our teams continue to work to address issues such as misinformation and disinformation, content blocking, data safety and preventing the promotion of harmful or discriminatory content in line with our AI principles. The new info on Bing Chat updates comes on the same day that the CEOs of Microsoft, Google, OpenAI, and Anthropic will be meeting with US Vice President Kamala Harris in Washington DC to discuss plans for keeping the use of AI ethical and responsible.

Meta talks about new malware threats including ones posing as ChatGPT by Ishtiaqe Hanif Meta was pleased to announce their latest efforts in identifying and preventing malware campaigns that targeted business users. The company's security teams used a range of methods to combat malware, including malware analysis, continuous improvement of detection systems, product updates, community support, and education. It has also shared threat information with other companies, and has taken legal action against threat actors. These combined efforts limited the lifespan of any single malware strain and increased the cost for malicious groups, which forced them to invest more resources in constantly adapting their tactics. As it had observed many malware campaigns hosted outside of social media, Meta encouraged people to exercise caution when downloading software or files from the internet. The company also shared additional security tips in their newsroom. Before discussing NodeStealer, a new malware family they recently discovered, Meta shared the latest trends they had observed in the malware threat landscape. Its research showed that many malware campaigns used custom-built tooling to target business users on particular internet services. Malicious groups had become very adaptive to disruptions and were spreading across multiple internet services to ensure a complex, multi-pronged malware campaign could withstand takedowns by any one service. Meta illustrated this point with an example of a long-running malware family called Ducktail, which had been evolving for several years as a result of enforcements by Meta and industry peers. Ducktail targeted multiple platforms across the internet, including LinkedIn, various browsers, and file-hosting services. Ducktail had adapted to the company's continued detection and mitigation efforts by granting business admin permissions to requests for ad-related actions sent by attackers. Similarly, malware operators used popular trends and issues to attract people's attention and trick them into clicking on malicious links or downloading malware. The company investigated and took action against several malware strains that took advantage of people's interest in OpenAI's ChatGPT to trick them into installing malware pretending to provide AI functionality. It also discovered around ten malware families using ChatGPT and similar themes to compromise accounts across the internet. Malware operators used cloaking to circumvent automated ad review systems and leveraged popular platforms such as social media, file-sharing services, and even official web stores to distribute their malware. Meta blocked over 1,000 unique ChatGPT-themed malicious URLs from being shared on its platforms and shared them with industry peers. Public reporting and blocking of these malicious strains forced their operators to rapidly evolve tactics to stay afloat. You can learn more about cybercriminals distributing fake ChatGPT apps to push malware and security reasearchers and white-hat hackers fighting back. Source: Meta

Microsoft might launch private ChatGPT for businesses that need their data safe and isolated by Karthik Mudaliar Microsoft is planning to launch a privacy-centric version of ChatGPT in response to concerns that people's data may be exploited for the training of artificial intelligence models. The move is set to appeal to industries like healthcare, finance, and banking that have refrained from using ChatGPT due to the potential risk of their staff sharing sensitive information with the system. A separate version of the AI tool will run on dedicated cloud servers where data will be kept separate from that of other customers. The data on these dedicated servers will be isolated from the main ChatGPT system to ensure privacy. The private setup could cost as much as 10 times more than what customers currently pay to use ChatGPT. OpenAI also plans to launch a new privacy-focused business subscription that won't feed user data to train models by default. OpenAI has also sold Morgan Stanley a private ChatGPT service using which the bank's wealth management division is able to ask questions and analyze thousands of the bank's market research documents. Since Microsoft already has a multi-year, multi-billion dollar investment into OpenAI, it is allowed to resell its products without breaking any terms. The privacy-centric AI service from Microsoft can be a game changer for companies that tend to deal with important and sensitive data. Samsung banned its employees from using generational AI chatbots at work or on devices they use for work when it found that some of its employees had uploaded company source code. Salespeople from Microsoft are already inquiring organizations about the upcoming product, as many existing customers have contracts with Azure that could prove beneficial in managing the data securely. Microsoft is expected to launch its private AI service later this quarter. Source: The Information

Box announces AI integration into its products in partnership with OpenAI by Karthik Mudaliar Cloud storage company Box has announced that it is introducing new artificial intelligence (AI) features across its products in partnership with OpenAI. The features will include analyzing information across customer contracts, summarizing financial documents, surface insights from surveys, and more. Aaron Levie, co-founder and CEO of Box, said: “We are at the start of a platform shift in enterprise software driven by recent advancements in generative AI, and nowhere is the potential impact greater than in enterprise content. We’ve seen a step function improvement in our ability to analyze and synthesize the massive amounts of data contained within an organization’s unique documents, videos, presentations, spreadsheets, and more. When combined with AI, we will be able to unlock the value of this content and make every person in a company smarter and more productive. Content is an organization’s most important data, and with Box AI we’re just getting started with how we’ll transform the way work gets done.” Although there's not a separate product yet, many companies are now jumping on the AI bandwagon to refine their existing functionalities. Box is also doing the same with what it calls Box AI. The company is currently focusing on a few use cases that we have already seen possible with generative AI. For starters, by clicking on the Box AI button, users can ask questions such as "Summarize this document for me", or "When does the NDA expire?" which could come in handy for complex and lengthy documents or contracts. Here are some more use cases that Box mentioned in their press release: Sales teams will be able to use Box AI to get answers to questions in complex contracts to speed up the sales cycle. Analysts will be able to have Box AI summarize lengthy financial reports to inform their rating recommendations. Legal teams will be able to ask Box AI to identify key clauses, terms, and obligations from a contract to speed up review cycles. Operations teams will be able to tell Box AI to extract key takeaways from a budget to update corporate strategy decks without waiting on a co-worker from the finance team for the right piece of information. Customer service teams will be able to use Box AI to surface insights from hundreds of customer feedback surveys to identify key areas for improvement. Box says that it caters to more than 115,000 customers and bringing foundational AI models to where their content is already securely stored will be more useful and valuable. For now, Box AI will only be available to select customers through its upcoming "Design Partner Program". Interested users can also sign up for private beta access when it becomes available. Pricing and other details will be announced later upon general availability.

Samsung is banning its employees from using chatbots like ChatGPT due to security leak by John Callaham More companies are cracking down against the use of generative AI chatbots like ChatGPT, Microsoft's Bing Chat, and Google's Bard. The latest business to ban the use of these kinds of chatbots is Samsung. Bloomberg reports that the company sent out a memo to all employees last week, barring them from using chatbots at work or on devices they use for work. According to the report, Samsung discovered that some of its workers had uploaded secret company source code to ChatGPT. There's no word on exactly what kind of data was leaked. While Samsung employees are allowed to use chatbots on devices they personally own, and they can use them outside of work, the company's memo did ask its workers to not submit any company info to chatbots, nor upload any personal info that could result in a leak of Samsung's intellectual property. The memo added: We ask that you diligently adhere to our security guideline and failure to do so may result in a breach or compromise of company information resulting in disciplinary action up to and including termination of employment. The report says that Samsung is developing its own AI software that will do some of the work that chatbots can do, such as summarizing reports, writing software, and translation. This latest development by a major tech company to ban the use of chatbots at work shows that more and more businesses are becoming concerned about how generative AI could lead to more security issues. This week, Microsoft outlined how it will implement responsible AI practices in its products.

ChatGPT has created a new app sector, with European developers taking the lead by Ishtiaqe Hanif ChatGPT has sparked a growing industry of AI chat apps, with Europe at the forefront. According to a recent study by App Radar, since their launch in November 2022, these innovative apps have garnered an impressive 23.6 million downloads from the Google Play Store. Recently, Call Annie has hit the App Store and gained traction. While OpenAI is yet to start working on GPT-5, GPT-4 is taking over the show. The study identified 40 AI chat apps with over 10,000 users that utilize ChatGPT or similar technology. Europe emerged as the global leader in this nascent sector, with 14 of the apps originating from the continent and boasting a collective 8 million downloads, representing 34% of the entire market. The Americas and Asia were home to nine and seven AI chat apps, respectively, with download figures of 2.7 million and 7.4 million. Additionally, eight apps with 5.2 million downloads had an undetermined country of origin. Thomas Kriebernegg, co-founder and managing director of App Radar, is astounded by the growth of this new sector. He credits the media's fascination with generative AI and the widespread belief that ChatGPT is a groundbreaking technology. The cooperation of developers and a passionate public in Europe drives the growth of this sector. KrKriebernegg notes that European consumers are generally more willing to experiment with new technology compared to other regions, as evident in the higher usage numbers for Fintech apps and solutions in the UK compared to the US. This cultural disposition offers European app developers an advantage in expanding their consumer base at a faster pace before venturing into new territories. Among European countries, Turkey stands out as a promising contender, with four apps, including the top three, accounting for 5.9 million downloads. With a flourishing development ecosystem, a vibrant mobile gaming hub, regulations that favor startups, and a growing list of tech successes such as Getir, Turkey have the potential to become a powerhouse in this segment. However, turning this early advantage into a profitable ecosystem will undoubtedly be a challenging task. Source: The Next Web

New AI language research could turn your thoughts into text in the future by John Callaham Reading a person's mind is something that has been mostly for fiction and for so-called mentalists and psychics. However, a new research paper claims that it has come up with a way to read a person's thoughts and turn them into text. The research comes from scientists working at The University of Texas at Austin. In a press release, the university says the scientists have published a paper on their research in the Nature Neuroscience journal. Their paper says that their breakthrough comes from using a transformer AI model, similar to the ones used by ChatGPT, Bing Chat, and Bard. This new technology does not require any surgical implants on the person to work. The press release says: Brain activity is measured using an fMRI scanner after extensive training of the decoder, in which the individual listens to hours of podcasts in the scanner. Later, provided that the participant is open to having their thoughts decoded, their listening to a new story or imagining telling a story allows the machine to generate corresponding text from brain activity alone. The texts that are generated by a person's brain activity with this method are not exact. Rather, the AI model will create a text message that's supposed to be a general approximation to what the person is thinking about: For example, in experiments, a participant listening to a speaker say, “I don’t have my driver’s license yet” had their thoughts translated as, “She has not even started to learn to drive yet.” Listening to the words, “I didn’t know whether to scream, cry or run away. Instead, I said, ‘Leave me alone!’” was decoded as, “Started to scream and cry, and then she just said, ‘I told you to leave me alone.’” This research will be helpful for people who have suffered strokes, or otherwise cannot communicate normally. The scientists who have developed this technology believe that it be used with more portable brain imaging systems in the future.

ChatGPT rival Google Bard launch was delayed over safety concerns by Ishtiaqe Hanif Google had been secretly working on an AI-driven virtual assistant since mid-2021, well before late 2022, when rival ChatGPT was released according to former Responsible AI team member Blake Lemoine, who discussed about Bard in a recent interview . Bard is a conversational generative AI bot based on LaMDA LLM. Blake told Futurism that OpenAI's offerings are not pushing Google around. What Google has in store is far more advanced in terms of technology, and its trajectory is not being affected. He also said he thinks Google is doing things in ways it believes are safe and responsible. OpenAI turned out to have released their ChatGPT before the curtains on Bard were lifted. Without putting a product name on it, Google has been working on Bard for a couple of years. Blake said, "They were on the verge of releasing something in the fall of 2022. So it would have come out right around the same time as ChatGPT, or right before it. Then, in part because of some of the safety concerns I raised, they deleted it." According to him, Google could've released a product similar to Bard's two years ago. Creative accuracy, and preventing gender, racial, and political bias are important for the Alphabet subsidiary. This has surfaced amid a report claiming Google is rushing out Bard and other AI products with poor ethical guards. Source: Futurism

ChatGPT is allowed again in Italy after it put in some data protection safeguards by John Callaham In March, Italy's data protection agency Garante placed a ban on the use of ChatGPT in that country. The ban was due to concerns that the chatbot AI service from the company OpenAI was in violation of Italy's data protection rules. Today, it was revealed that Italy's government has lifted the ban against ChatGPT after OpenAI agreed to put in some safeguards to protect the personal data of its users. The Associated Press reports that some of those privacy additions include adding a signup form that asks users to confirm their age. TechCrunch clarified that the form asks users if they are 18 years old or above, or if they are between 13 and 17 years old, which in that case they need the permission of a parent or guardian. In addition, ChatGPT users in Italy can now access links to see how the chatbot AI uses its personal info, and how it collects data to train the chatbot. Finally, Italy users can fill in a form to object to having their personal data accessible to ChatGPT for its training. The new privacy additions show that there continue to be concerns about how chatbots like ChatGPT, Microsoft's Bing Chat, and Google Bard use online data to generate the answers to users' questions. While these new features are a start, there will likely be more questions on these matters in the months and even years to come.