Search the Community

OpenAI reveals how it will try to stop the use of its tools for false election information by John Callaham Today, the 2024 US Presidential election officially gets underway with caucuses in the state of Iowa. Generative AI company OpenAI has decided to use today to outline how it plans to prevent groups from using its tools like ChatGPT and DALL-E to create and distribute 'deepfakes" and other false information that could be used to disrupt the US presidential election, and indeed elections all over the world this year. In a blog post, OpenAI stated it will make sure its tools are used for "accurate voting information, enforcing measured policies, and improving transparency". It added: We have a cross-functional effort dedicated to election work, bringing together expertise from our safety systems, threat intelligence, legal, engineering, and policy teams to quickly investigate and address potential abuse. OpenAI says it will not allow its tools to be used for any political campaigns or for any lobbying purposes. Also, it states that it will not allow any chatbots to be created with its tools and services that are designed to simulate chatting with real candidates or government groups. The company says it will block the use of its tools to distribute efforts to misrepresent how people should vote in the election, or discourage people from voting. It will also allow people to report possible violations to OpenAI. In addition, OpenAI is working on tools to help spot AI-created deepfake images made via DALL-E. A new tool, called a "provenance classifier" is currently under development that is supposed to help find AI-made images " even where images have been subject to common types of modifications." A group of testers that will include journalists, will be able to try out this tool "soon". A recent report claimed that Microsoft's Copilot chatbot, which uses technology developed by OpenAI, frequently offered false answers to questions about some 2023 elections. In November, Microsoft announced plans to offer a new tool that would help political parties show that images created by them, such as ads and videos, are real and have not been changed by AI tools.

New report claims OpenAI's board is in talks to bring Sam Altman back as CEO by John Callaham Friday's announcement that OpenAI's board of directors fired its co-founder and CEO Sam Altman out of the blue shook the entire tech industry. Now there's a new report claiming that the company's board is actually in talks to bring Altman back to his former role. The report comes from The Verge, citing unnamed sources. There's no word on exactly why OpenAI's board might be willing to do a 180 on Friday's decision. It does claim Altman himself is "ambivalent” about returning to the CEO role. It also claims that if he did return, there would have to be "significant governance changes" at the company. OpenAI's board stated on Friday that Altman was shown the door because "he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities." After his firing, another OpenAI co-founder and its President Greg Brockman announced his resignation from the company, after being told he would no longer be the board's chairman. In the 24 hours and change since Altman's firing, there have been numerous reports about why Altman was fired. One report from Bloomberg (via Yahoo) claimed that Altman had disagreements with the board on the safety of its AI products like ChatGPT, along with both the speed and the commercialization of those AI products. Another Bloomberg report (via Yahoo) stated Altman specifically clashed with OpenAI's chief scientist and board member Ilya Sutskever, and it's possible he convinced the other board members to fire Altman. Currently, the company's chief technology officer Mira Murati is serving as its interim CEO. The same report claimed that OpenAI's major technology and investment partner Microsoft was "blindsided" by the decision to fire Altman. Officially, Microsoft's CEO Satya Nadella issued a statement claiming the company was "committed to our partnership" with OpenAI. However, Bloomberg's report claims that privately, Nadella was "furious" at the board's decision to get rid of Altman.

Microsoft Teams will soon add generative AI-based custom backgrounds for annoucement posts by John Callaham Back in May, Microsoft rolled out a new and large collection of virtual backgrounds that Microsoft Teams member could use while they were in a video meeting. Now it looks like Microsoft is planning to offer Teams members some more ways to create backgrounds, at least for announcement posts in Teams channels. In a new entry on the Microsoft 365 Roadmap this week, Microsoft confirmed it is working on a way for Teams users to create their own custom background, for those announcement posts. The roadmap entry says it should only take "just a few clicks" to make those backgrounds in Teams. In addition, Teams Premium members will be getting an additional wrinkle to this feature. They will have access to OpenAI's DALL-E generative AI art creation service. This will allow those Premium members to just type in some text prompt to make those custom backgrounds for announcement posts. The Roadmap listing claims this feature will be available for PC, Mac, iOS, Android and web Teams users. It's currently listed as rolling out sometime in November, although keep in mind that Microsoft 365 Roadmap listings can change their dates without warning. Right now, the roadmap entry doesn't state exactly how this new custom background feature will work in Teams, so we may have to wait a bit before learning more about this feature. We would also expect that the use of DALL-E for custom backgrounds will be subject to certain content guide rails inside Microsoft Teams. Earlier this month, Microsoft officially launched the new Teams app in general availability for both Windows and Mac. Microsoft claims that the performance of the new Teams apps is up to twice as fast compared to the "classic" Teams apps, while also using less memory at the same time. The new Teams app will eventually become the default version for all users.

Bing Image Creator is taking longer than usual to make art; Microsoft is adding more GPUs by John Callaham In late September, Microsoft announced a ton of software updates related to its AI efforts. One of them was that Bing Image Creator, its generative AI-based art maker, had been officially upgraded with the DALL-E 3 model from OpenAI as of late Friday, September 29, and was available for all of its users. However, it looks like the addition of DALL-E 3 in Bing Image Creator has caused a big boost in the number of people accessing it, and as a result, the service can sometimes be slow in making images. This has been noted by a number of people on X (formerly Twitter), including @DmitryForLeader who wrote that he had to wait over two hours to get an image. Others have posted reports of similar long waits to see their artwork generated. It looks like Microsoft is very aware of this issue and is working to fix it ASAP. Mikhail Parakhin, the new head of Microsoft's Windows and Web Services team, posted his own X message which stated that they didn't expect as much interest in DALL-E 3 in Bing Image Creator from its users this past weekend. Folks, we know DALL-E 3.0 generation right now is taking longer than normal. We expected some strong interest, but we didn't expect THAT much, especially given it's a weekend. Bringing more GPUs in, will be better soon. — Mikhail Parakhin (@MParakhin) October 1, 2023 Parakhin stated that Microsoft is "bringing more GPUs in" and that should help to decrease the wait time "soon". When asked in a follow-up message just how many GPUs are being added, Parakhin responded with, "Measured in thousands of H100-equivalent." That's a reference to NVIDIA's H100 GPUs, which have been made and bought by the boatload by Microsoft and other companies for use in their data centers specifically for generative AI services like Bing Chat and Bing Image Creator. Those sales have also caused NVIDIA's stock price to skyrocket in 2023.

OpenAI announces DALL-E 3 image generator that can integrate with ChatGPT by Omer Dursun The AI image generator race is heating up. We recently reported that Microsoft Paint may get an AI-powered art generator feature, while Adobe has commercially released Firefly generative AI for Photoshop, Illustrator, and Express. Today, OpenAI has announced the third generation of its popular AI art generator, DALL-E. The new version, dubbed DALL-E 3, comes with significant upgrades to improve safety and makes the tool more accessible and collaborative. A key feature is the integration of DALL-E with OpenAI's ChatGPT conversational agent. Users will soon be able to use ChatGPT to help craft text prompts to guide DALL-E's image generation rather than having to develop their own detailed prompts. In a demonstration, OpenAI researchers showed how ChatGPT could generate a long paragraph describing a prompt for a ramen restaurant logo, which DALL-E then used to produce several concept designs. Our new text-to-image model, DALL·E 3, can translate nuanced requests into extremely detailed and accurate images. Coming soon to ChatGPT Plus & Enterprise, which can help you craft amazing prompts to bring your ideas to life:https://t.co/jDXHGNmarT pic.twitter.com/aRWH5giBPL — OpenAI (@OpenAI) September 20, 2023 The company believes this new feature will allow more people to benefit from AI art generation as they won't need highly developed writing skills to utilize DALL-E. The integration with ChatGPT opens up prompt engineering support. In addition, OpenAI says it has focused heavily on safety improvements in DALL-E 3 through mechanisms like input classifiers trained to avoid objectionable content. The company worked with external evaluation teams to test the system's robustness. DALL-E 3 also reportedly improves on avoiding recreating copyrighted works or images of public figures when prompted. OpenAI will roll out DALL-E 3 access over the coming months, starting with ChatGPT Plus and Enterprise subscribers in October. The company did not provide a timeline for a fully public release. In case you missed it, since March, Microsoft Edge has a feature called Bing Image Creator powered by the DALL-E model.

Microsoft Paint may add an AI art generator in the future similar to Bing Image Creator by John Callaham Microsoft just updated its Paint app for Windows 10 and 11 for some members of the Windows Insider Program. However, it looks like we might get some additional features for the app in the near future. Windows coding enthusiast "PhantomOcean3" posted on his X (formerly Twitter) account that he found some code strings that seem to indicate there's an AI image generation feature coming for Paint called Cocreator: The latest Paint app update in Dev/Canary (11.2308.18.0) includes strings for a new image generation feature: Cocreator Describe what you'd like to create Try "a cat walking in a wood" or "A house made out of leaves" Choose a style — PhantomOcean3 💙💛 (@PhantomOfEarth) September 18, 2023 In a follow up post, he stated the feature is based on OpenAI's DALL-E AI image generator. That's the same one that Microsoft already uses for its Bing Image Creator service in Bing Chat. It's possible we could learn a lot more about Microsoft's plans to add AI features in Paint and other Windows 11 apps in just a few days as part of the company's press event in New York City on September 21. The Paint app for Windows 10 and 11 has received quite a few updates in the Insider channel over the past few months. That includes adding the long-awaited dark mode to the app this past summer. Some of the other new additions include center canvas support in mid-August, followed by a background removal feature in early September. Today's Insider update for Dev and Canary channel members put in a way to add, remove, and manage layers on the canvas. There's also new support for transparency "including the ability to open and save transparent PNGs" according to Microsoft. All of these new features coming in the space of a short time span seems to indicate that Microsoft is trying to improve its core apps for Windows on a more frequent basis.

ChatGPT creator OpenAI to hold its first developer conference on November 6 by Omer Dursun OpenAI has announced plans to host its first developer conference, OpenAI DevDay. The one-day event, scheduled for November 6, will bring together developers worldwide for technical sessions, product demos, and a live-streamed keynote speech in San Francisco. Through the conference, OpenAI aims to strengthen its relationships with the developer community, which has been key to the rapid adoption of its AI technologies. The one-day event will bring hundreds of developers from around the world together with the team at OpenAI to preview new tools and exchange ideas. In-person attendees will also be able to join breakout sessions led by members of OpenAI's technical staff. We're looking forward to showing our latest work to enable developers to build new things. The company opened up access to its models through an application programming interface (API) in 2020, which Microsoft was an early adopter. Today, over 2 million developers currently utilize OpenAI models like GPT-4, GPT-3.5, Dall-E, and Whisper in their applications and products. OpenAI DevDay will likely include updates on the ChatGPT conversational agent, leveraging GPT-4 and GPT-3.5. With AI giants like Google and Microsoft racing to keep pace, OpenAI is positioning itself at the forefront of AI innovation. Recently, the company revealed ChatGPT Enterprise, which offers a number of features that are not currently available for normal ChatGPT users. It follows Microsoft's first announced Bing Chat Enterprise. On the other hand, four tech companies (Microsoft, OpenAI, Google, and Anthropic) have teamed up to create the Frontier Model Forum in July. The goal of this group is to ensure "safe and responsible development" of generative AI, specifically for frontier AI models. Registration details for the gathering have not yet been announced, however, DevDay's website says in-person developer registration will open in the coming weeks. The company also plans to livestream DevDay's keynote presentation to make it accessible to developers worldwide.

Adobe adds AI features to Photoshop with new Generative Fill feature by John Callaham Earlier this year, Adobe announced Firefly, a set of generative AI tools and features for its many Creative Cloud apps. Today, Adobe revealed that one of those Firefly tools, Generative Fill. is being added to the beta version of its popular Photoshop image editing app. In a press release, Adobe stated that Generative Fill for Photoshop was designed to help photo editors add, remove, or extend content in images. It stated: Generative Fill automatically matches perspective, lighting and style of images to enable users achieve astounding results while reducing tedious tasks. Generative Fill expands creative expression and productivity and enhances creative confidence of creators with the use of natural language and concepts to generate digital content in seconds: One thing about Adobe's Firefly platform is that it uses artwork that comes from the company's own Adobe Stock images, along with openly licensed and public domain content. This allows features like Generative Fill to use content without having to deal with copyright restrictions. In addition, Adobe says it follows its own AI Ethics rules for labeling content made with generative AI as such. It states: Generative Fill supports Content Credentials, serving an essential role in ensuring people know whether a piece of content was created by a human, AI-generated or AI-edited. Content Credentials are like “nutrition labels” for digital content and remain associated with content wherever it is used, published or stored, enabling proper attribution and helping consumers make informed decisions about digital content. The Generative Fill feature in Photoshop is available on the desktop beta app today. It's also available on the module within the Firefly beta app. It is expected to be generally available sometime in the second half of 2023. It will join the many art AI creation tools already out there, including Microsoft's Bing Image Creator and Midjourney.

A major photography prize was given to a man who used AI to generate it by John Callaham Earlier this month, it was revealed that the prestigious Sony Photography Awards had named Boris Eldagsen as the winner of its creative open category for an image he submitted called The Electrician. Eldagsen then admitted to Sony that the photo was not a photo at all. It was an image he created via unnamed AI generators. The photo was part of an online art collection from Eldagsen that he called Pseudomneisa. The page states: These images were imagined by language and re-edited more between 20 to 40 times through AI image generators, combining “inpainting”, “outpainting” and “prompt whispering” techniques. So why did he submit one of these images to a major photography contest? In an interview with The Register, Eldagsen stated he wanted to start a conversation about how AI will change how we look at art and photography in the future. He stated: I applied as a cheeky monkey, to find out if the competitions are prepared for AI images to enter. They are not. We, the photo world, need an open discussion. A discussion about what we want to consider photography and what not. Is the umbrella of photography large enough to invite AI images to enter – or would this be a mistake? At first, the competition's organizers still named him as a winner even though he admitted the image's true origins. Eldagsen still refused the prize for winning the category, which included $5,000, some Sony camera products, and a trip to London to view the winning photos. A spokesperson for the competition told The Guardian in a statement that it "felt that his entry fulfilled the criteria for this category, and we were supportive of his participation." It added: We recognise the importance of this subject and its impact on image-making today. We look forward to further exploring this topic via our various channels and programmes and welcome the conversation around it. As we see more and more AI image generators launch, like DALL-E, MidJourney, and Microsoft's Bing Image Creator, we are seeing more and more debate about how these services could be used to spread misinformation with "deep fake" photos. Google recently revealed plans to launch its own AI image and video creators, but added that they will not be programmed to create images with humans.

AI art generator Midjourney is ending its free trial due to "abuse" of its images by John Callaham Even before chatbots like ChatGPT, Bard, and Bing Chat launched, AI was being used by more and more people to create art and images from just a few text prompts. This week, one of the biggest names in this field, Midjourney, announced a sudden end to its free trial, and the reasons could have implications for other similar programs. According to The Washington Post, Midjourney CEO David Holz stated that the company's own preventive measures have not been enough to keep the service from being hit with "extraordinary demand and trial abuse". For now, Midjourney is a paid service, with access only available if users pay $10 a month. The abuse that Holz refers to are the very realistic-looking "deepfake" images of real people that have been generated by Midjourney users. They have raised concerns that they could be used by hostile actors to create dissent online. It also put other AI art generators in the spotlight, including Microsoft's recently launched Bing Image Creator. Microsoft admitted that it put strong restrictions inside the generator. In fact, it banned making images when users just typed "Bing" in the prompts. It shows that it's an all-new world when it comes to making artwork with just a few words, and no one really knows how far it will go in the future. Source: The Washington Post

Microsoft's Bing Chat adds Bing Image Creator for making AI generated art by John Callaham Microsoft continues to put in additional features to the new Bing and Bing Chat. Today, the company confirmed that a small number of Bing Chat preview users can now check out Bing Image Creator, which will let them see AI-made artwork generated by their prompts in Bing Chat. Microsoft states: For those in the Bing preview, Bing Image Creator will be fully integrated into the Bing chat experience, rolling out initially in Creative mode. By typing in a description of an image, providing additional context like location or activity, and choosing an art style, Image Creator will generate an image from your own imagination. It’s like your creative copilot. Just type something like “draw an image” or “create an image” as a prompt in chat to get creating a visual for a newsletter to friends or as inspiration for redecorating your living room. The AI art generator is using an advanced version of OpenAI's DALL∙E model. In addition to the normal Bing Chat platform, You can access Bing Image Creator in the Microsoft Edge web browser: To use Bing Image Creator in Edge, simply click the Bing Image Creator icon in the sidebar to create your image or invoke from Bing chat in Edge. Microsoft stated that for now, only a small number of Bing Chat preview users will be able to use Bing Image Creator, but it will be rolled out to more users in the near future. While it's only available in Creative mode, it will be supported by Balanced and Precise mode sometime in the future as well. However, Microsoft is allowing anyone to check out the preview version of Bing Image Creator on its own website. Microsoft also announced that it's adding even more features to the new Bing: Stories and Knowledge Cards 2.0: Stories provide a more engaging way to search and interact with content, offering images and short videos. Also new to Bing users today, Knowledge Cards 2.0 is an AI-powered infographic-inspired experience that provides fun facts and key information at a glance. It’s been updated to include interactive, dynamic content like charts, graphs, timelines, visual stories and more. Microsoft added that over more than 100 million chats have been made in the new Bing since its public preview launch.

OpenAI sees challenges with full AGI and outlines plans to address them by Karthik Mudaliar Image by DALL-E Artificial general intelligence (AGI) is a term that refers to AI systems that can perform any task that humans can do, or even better. Achieving AGI is one of the most ambitious and controversial goals in the field of AI, as it could have profound implications for humanity's future. OpenAI, the AI research company behind ChatGPT, has recently published a blog post that gives us an idea how the company will be pulling off with the challenges for creating and managing AGI. OpenAI's wants AGI to "empower humanity to maximally flourish the universe". With AGI, there could be potential risks and uncertainties such as misuse, accidents, and even societal disruptions. To tackle this, OpenAI has proposed a set of principles that guide its work on AGI, intended for the short term - Deploying and learning from less powerful AI systems in the real world to gain experience and feedback Democratizing access to AI benefits and governance Developing new alignment techniques to ensure that AI systems act in accordance with human values and preferences Strengthening the institutions of the world to agree on wide bounds for AI use and regulation Sharing information about AI progress with the public and seeking input from diverse stakeholders There are certainly other factors as well that will shape how the development of AGI is impacting humanity. These factors include the timeline (how long it will take to create AGI), takeoff speeds (how fast AGI will improve itself), coordination (how well different actors will cooperate or compete on AGI development), and safety (how well aligned AGI will be with human goals). OpenAI says that a shorter timeline and slower takeoff speeds will allow more time for learning, adaptation, and coordination among governments, firms, and the overall general population to gradually understand what's happening and experience both the upsides and downsides of the technology. With that said, OpenAI also warns about the fact that things are not predictable and even the company itself doesn't have all the answers and solutions to potential problems in the future, especially after AGI becomes mainstream. The company hopes that for the long term, a conversation about how to govern these systems, how to fairly distribute their benefits, and how to fairly share access is a must. OpenAI also emphasized upon how the company's structure is aligned with its mission. The company's charter outlines its principles and goals for developing AGI. It also has a cap on the returns its shareholders can earn, and the company is actually governed by a nonprofit board that oversees its activities. The board can also override any for-profit interests if they conflict with the company's mission. You can read more about how OpenAI plans for the future of AGI by clicking here. Are you with or against the idea of AGI? Let us know your thoughts in the comments section below.

OpenAI reportedly is working on a ChatGPT mobile app along with AI-created videos by John Callaham OpenAI's ChatGPT app continues to send huge ripples throughout the entire tech industry. However, those ripples could soon turn into tidal waves. According to a new article on Semafor, using unnamed sources, OpenAI is working on a mobile app version of ChatGPT, along with a new feature that will allow users to make videos with AI assistance, similar to the image generation AI app Dall-E. This report follows up on previous rumors that Microsoft plans to incorporate ChatGPT into its long-suffering Bing search engine. This new report claims that OpenAI plans to introduce a new and faster version of ChatGPT, called GPT-4, into Bing " in the coming weeks", The report claims that GPT-4 creates answers to questions much faster than the current GPT-3 version and those answers also look more like a human wrote it. The upcoming mobile app, according to the report, should give ChatGPT an even bigger base to provide feedback to OpenAI so the AI service can offer better answers. There's no word on when the mobile version will be released. The report also does not give any more details about the new AI-assisted video creation tool. The article does reveal that each search made on ChatGPT costs a couple of cents, which means it's extremely expensive to run the program. Microsoft and OpenAI will have to find new ways to keep the ChatGPT servers running and make them run more efficiently. Indeed, it's not uncommon to go to the ChatGPT web site and find that it's not working due to too many users at once. Neither Microsoft nor OpenAI has commented on Semafor's new report.

Microsoft promises to give customers access to ChatGPT as a service on Azure soon by Usama Jawad ChatGPT has been making a lot of waves in mainstream media lately. The capabilities and potential use-cases demonstrated by the GPT-3 large language model have managed to wow consumers and enterprise customers alike. Naturally, Microsoft, which invested $1 billion in ChatGPT's founding company OpenAI - with reports of more financing to follow, wants to capitalize on this momentum. As such, it has committed to launch ChatGPT on Azure soon. Microsoft has announced that its Azure OpenAI service is now generally available, which means that customers can access complex AI models such as GPT-3.5, Codex, and DALL•E 2 on the cloud. For those unaware, GPT is the family of large language models powering ChatGPT, Codex is the model that is used to convert natural language to programming code in GitHub Copilot, and DALL•E can generate images after receiving only textual inputs. As such, Microsoft has boasted that Azure is the "best place to build AI workloads" for customers of all sizes across various industries looking to save time on operational activities and offering streamlined AI-powered services to end-users. It claims that customers can build AI applications on Azure by leveraging purpose-built AI-optimized infrastructure with enterprise-grade functionalities. That said, Microsoft has noted that it is committing to innovating in the AI space in a responsible way - many people think that this isn't the case right now at least. It emphasizes: We have taken an iterative approach to large models, working closely with our partner OpenAI and our customers to carefully assess use cases, learn, and address potential risks. Additionally, we’ve implemented our own guardrails for Azure OpenAI Service that align with our Responsible AI principles. As part of our Limited Access Framework, developers are required to apply for access, describing their intended use case or application before they are given access to the service. Content filters uniquely designed to catch abusive, hateful, and offensive content constantly monitor the input provided to the service as well as the generated content. In the event of a confirmed policy violation, we may ask the developer to take immediate action to prevent further abuse. We are confident in the quality of the AI models we are using and offering customers today, and we strongly believe they will empower businesses and people to innovate in entirely new and exciting ways. Although ChatGPT will be available as a service on Azure for customers to build upon soon, the Redmond tech giant hasn't really defined a concrete release date yet.

Shutterstock integrates OpenAI's image generation into its platform by Steve Bennett Shutterstock has announced today that it will be expanding its partnership with OpenAI, the creators of the AI image generator known as DALL-E. The platform is looking to launch a new tool for users of the site, dubbed Shutterstock.ai, which will allow for AI generated images to generate revenue for creators, while also training the AI itself. OpenAI and Shutterstock originally entered into a strategic partnership in 2021, which helped to train and improve DALL-E's capabilities: "The data we licensed from Shutterstock was critical to the training of DALL-E,” said Sam Altman, OpenAI’s CEO. “We’re excited for Shutterstock to offer DALL-E images to its customers as one of the first deployments through our API, and we look forward to future collaborations as artificial intelligence becomes an integral part of artists’ creative workflows.” The platform itself will not just provide a new income stream for creators, but will also allow for Shutterstock to compensate those who have helped to further the development of the AI significantly. Shutterstock, with this new partnership, has also integrated OpenAI's seamless image generation into the platform, and allows its customers to generate images immediately, which they have branded with the strapline "Creativity at the Speed of Your Imagination." Shutterstock also confirmed that it is committed to ensuring that the development of the AI follows ethical best practices, and that it will be helping to develop policy surrounding usage rights, even for AI generated content. Source: Shutterstock

Microsoft brings DALL·E 2 to Azure OpenAI, launches new employability program, and more by Usama Jawad Just as it does at each one of its developer-oriented Ignite conferences, Microsoft has announced a bunch of new capabilities for its Azure cloud in the 2022 edition as well. Starting off with Azure Cognitive Services, DALL·E 2 is now available for select customers through a preview in Azure OpenAI. For those unaware, the model allows the generation of images via a text or image input. Better text summarization capabilities in over 90 languages are now present in Azure Cognitive Service for Language too. Meanwhile, Image Analysis 4.0 and Spatial Analysis on the Edge are two tools ready for preview in the computer vision space. They will offer better feature extraction and analysis with a focus on safety and security. Furthermore, Azure Cognitive Service for Speech is getting new customization features such as more languages and emotions too. You can read more about all the updates to Azure Cognitive Services here. Azure Form Recognizer has a couple of preview capabilities as well, including content extraction from contracts and support for 275 languages. On the Azure Machine Learning side, we have registries, Azure Container for PyTorch, Azure Data Science Virtual Machines for PyTorch, and Apache Spark pools integration available as preview. The generally available capabilities include automated machine learning features and Python SDK v2. The Responsible AI dashboard is set to become generally available soon too. Microsoft had some updates to share regarding its low-code industrial AI development platform "Project Bonsai" as well. There are a bunch of new capabilities in preview including action masking, training from data with OfflineRL, model-based training acceleration, Assessment 2.0, and Bonsai Samples in Azure Marketplace. It has also added Tata Consultancy Services (TCS) as a partner in this initiative. Coming over to the Azure Data side of things, Microsoft announced a bunch of new partner solutions in multiple domains for its Intelligent Data Platform. This includes MongoDB and YugabyteDB for databases, Qlik, Informatica, Fivetran, and more for analytics, and Onetrust, Delphix, and others for data governance. On the databases end, Azure CosmosDB has better performance through the Citus engine and support for documents larger than 16MB in Azure CosmosDB for MongoDB. Meanwhile, Azure Database for PostgreSQL has a preview feature in tow in the form of support for customer-managed encryption keys (CMEK) for data at rest. And Azure Database for MySQL now sports a preview for Autoscale IO. As the name suggests, it manages IO dynamically and you only pay for the IO that you use. As expected, Azure Synapse Analytics is also getting more capabilities soon. This includes a pipeline template for connectivity to Microsoft 365 data and an improvement to the Copy Activity functionality through a tool called Mapping Data Flows. Meanwhile, R language is now supported in preview form in Azure Synapse while a new version of SynapseML is generally available. Azure Data Explorer is adding support for new data ingestion types, including Amazon S3, Azure Event Grid, Azure Synapse Link, and OpenTelemetry Metrics. It can also read and write in the open standard Parquet and Delta formats. Additionally, the SAP Change Data Connector (CDC) for Azure Data Factory is now generally available. Finally, Microsoft Purview Data Governance has new features such as SQL Dynamic lineage, Metamodels, and machine learning-based classifications for more data formats in preview. Shifting gears to Azure hybrid, multi-cloud, and edge, we have a single sign-on experience (SSO) experience integrated with Azure Active Directory (AAD) on Azure Arc-enabled SQL Server along with better security through Microsoft Defender. There are more hybrid deployment options for Azure Kubernetes Service (AKS) as well. Meanwhile, Azure Stack HCI has a new release featuring improved Azure remote support, ​Azure Marketplace for VM self-service, and ​Graphics processing unit partitioning (GPU-P) with live migration in preview, along with intent-based network automation with Network ATC being generally available. On the Azure infrastructure track, Azure Automanage is available for for Azure Arc-enabled servers and Azure virtual machines, and there are flexible orchestration options for Virtual Machine Scale Set and Spot Virtual Machines for more cost-saving too. Similarly, Azure Premium SSD v2 is generally available while Azure Elastic SAN (storage area network) is now in preview. Over on the monitoring side, Azure Monitor has cost optimizations for logs and predictive autoscale for Virtual Machine Scale Sets. Azure Monitor Agent (AMA) now supports Windows 10 and Windows 11 through the Windows client installer and also has an agent migration tool, now in preview. In the same vein, alert rule recommendations will be offered too. In the networking space, Azure Domain Name System (DNS) private resolver and Azure public multi-access edge compute (MEC) are generally available. Microsoft is also expanding Azure Hybrid Benefit and is offering the Windows admin center through the Azure portal as well. Additionally, Microsoft will soon offer a savings plan for compute that will enable customers to take advantage of a fixed hourly rate for all compute as long as they sign up for a commitment. Finally, Nutanix Cloud Clusters are generally available too. Customers of Azure VMware Solution will likely be pleased to find out that stretched clusters now provide 99.99% private cloud uptime in preview. Customer managed keys (CMK) are supported as well, offering easier management, rotation, and integration with Azure Monitor and Azure Log Analytics. Finally, when it comes to DevOps and other developer tools on Azure, we have a whole bunch of updates. We have listed them below for brevity and you can view more details via the embedded link, where available: Azure Kubernetes Fleet Manager is a new service in preview to simplify cluster management and organization Independent software vendors (ISVs) can now create one-click Kubernetes apps Azure Monitor has a new managed service for Prometheus and Java profiling with Java Flight Recorder (JFR) Visual Studio images for Dev Box are now generally available on Azure Marketplace Azure Deployment Environments are now in preview GitHub Advanced Security for Azure DevOps is in private preview Visual Studio 2022, v17.3 is generally available with new toolkit for building Microsoft Teams apps, performance improvements for C++ development, and tooling for .NET Multi-platform App UI (MAUI) Microsoft has partnered with Postman on Azure API Management Azure Communication Services is introducing a preview for call automation next month Service Bus explorer capabilities are now generally available Azure App Service has a bunch of new features including support for Go and larger SKU options Azure Managed Confidential Consortium Framework is now in preview Last, but definitely not the least, Microsoft is launching an employability program to bridge the gap between job seekers and the Microsoft partner and customer ecosystem. The former will be able to upskill themselves through self-paced trainings on Microsoft Learn and will be able to showcase this proficiency on LinkedIn. They will also be invited to a private LinkedIn group where employers will be able to directly engage with them.

Microsoft unveils new Designer, Microsoft Create, Bing Image Creator, and Clipchamp by Sayan Sen At its Ignite 2022 event today, Microsoft has unveiled a bunch of tools for creators. The biggest announcement is its new Designer app. While we already had a pretty good idea of what the new app was about from previous leaks which included several different images, the official announcement confirms some of those expectations and rumors. Designer is a graphic design app in Microsoft 365 that will help users create social media posts, invitations, digital postcards in a fly as it will powered by OpenAI's DALL.E 2, among others. Basically the company says Designer is PowerPoint on steroids where the AI does most of the heavy lifting. With its help, users can expect over eight billion AI-generated slide designs. Microsoft has confirmed that Designer integration is coming to Edge as well. The free web preview is available at this link starting today. Since it is driven by AI, the app is expected to get better over time. Designer web preview While we are on the topic of AI-generated images, Microsoft is also bringing a new "Image Creator from Microsoft Bing" feature that is similar to Designer, except his will be in Bing and help create completely new images. It will be a gradual rollout and once available, users can access the feature via the Images tab in Bing (https://www.bing.com/create) or via the Image Creator icon on the sidebar in Edge. Up next, we have an announcement related to Clipchamp, Microsoft's inbox video editor. The tool is now a part of the Microsoft 365 family of apps and will help complement Designer when video editing tasks are involved. As a reminder, premium Clipchamp features are now available to Microsoft 365 Personal and Family subscribers and Clipchamp Essentials subscribers. Microsoft has also launched a new "Microsoft Create" website. The Redmond company describes it as the "ultimate creator launchpad" for creating videos, graphic designs, documents, presentations, and more. Essentially, the Create website will provide a curated list of various articles, videos, templates and more, from similar creators that will help guide new users.

Powerful AI-based image generator DALL·E to become available to a million more people by Usama Jawad DALL·E is an extremely powerful AI tool based on GPT-3 that can create images out of practically any text you pass to it as an input. For example, it could even construct a completely ludicrous image of a dog cutting trees if you pass DALL·E a good enough description of what you want the AI model to draw. If you're on social media, you may have seen image generations from DALL·E's unofficial and less powerful counterpart DALL·E Mini. Up until now, access to DALL·E has only been given to a handful of people but it will soon become available to a million more. This announcement was made by OpenAI, the company behind DALL·E. DALL·E has entered beta testing phase. It's a closed beta of sorts because access will only be given to one million people on the the waitlist, as selected by OpenAI. Each individual will receive 50 free credits in their first month and then 15 complimentary credits in subsequent months. They will also have the option to purchase 115 credit-increments for $15 each. A single credit can be used to prompt DALL·E once, and in return, it will generate four images based on the text you passed to it. So if you purchase 115 credits for $15, you will be able to send it 115 prompts and receive 460 images in total. A single credit can also be used for an edit or variation prompt, but that will return three images. What's even more interesting is that DALL·E users will immediately get commercial rights to any images created from their prompts. This includes reprinting, selling, and merchandising. Theroretically, a character that you conjure up using free credits could potentially be worth millions commercially. OpenAI has noted that it will be taking the following measures to ensure that DALL·E is safe to use and that it is not misused for malicious purposes: Curbing misuse: To minimize the risk of DALL·E being misused to create deceptive content, we reject image uploads containing realistic faces and attempts to create the likeness of public figures, including celebrities and prominent political figures. We also used advanced techniques to prevent photorealistic generations of real individuals’ faces. Preventing harmful images: We’ve made our content filters more accurate so that they are more effective at blocking images that violate our content policy — which does not allow users to generate violent, adult, or political content, among other categories — while still allowing creative expression. We also limited DALL·E's exposure to these concepts by removing the most explicit content from its training data. Reducing bias: We implemented a new technique so that DALL·E generates images of people that more accurately reflect the diversity of the world’s population. This technique is applied at the system level when DALL·E is given a prompt about an individual that does not specify race or gender, like "CEO". Monitoring: We will continue to have automated and human monitoring systems to help guard against misuse. Finally, it is important to note that artists in need of financial assistance should fill out this form to get more details about subsidized access. One million more people on DALL·E's waitlist should be able to access the tool within the next few weeks.

OpenAI's DALL·E borrows from the GPT-3 and creates high-fidelity images from text by Ather Fawaz Last year, OpenAI released GPT-3, the largest transformer model to date with over 175 billion parameters. The model demonstrated great prowess in generating text from a given context and OpenAI licensed it exclusively to Microsoft for providing the computational backend required to host and run the model for its customers. Building on this, OpenAI have announced a distilled, 12-billion parameter version of GPT-3 today. Dubbed DALL·E, the new transformer model borrows heavily from GPT-3 but combines its abilities with ImageGPT (a model that completed half-complete images provided as input to the model). As such, DALL·E specializes in generating images from a given caption. The name DALL·E is a portmanteau of the artist Salvador Dalí and Pixar’s famous WALL·E. The model receives a stream of up to 1280 tokens containing both the text and image as a single stream of data. After preprocessing this, the model is then trained using maximum likelihood to generate all tokens sequentially. Once trained, DALL·E creates images for a variety of sentences that explore the compositional structure of language. Some of the samples for that are shown below. Image via OpenAI As demonstrated by the samples above, DALL·E extracts the essential information from a sentence and translates that into images. Impressively, OpenAI also provides a handy interface on its blog post to generate images of your own liking. We tried out a bunch of them all with promising results. For example, when the input to the model was "a triangular yellow manhole cover in the shape of a triangle", DALL·E produced the following array of images. Impressively, the model can extract temporal as well as geographical information from the provided text. Similarly, OpenAI says that the model offers a degree of controllability over the attributes and positions of a small number of objects as well. To illustrate, for the caption "a capybara made of voxels sitting in a field", the model produced the following results. And for "a plain white cube looking at its own reflection in a mirror. a plain white cube gazing at itself in the mirror", the model generated the following images. DALL·E also borrows the zero-shot learning capabilities of the GPT-3 and extends those to the visual domain. It can perform several kinds of image-to-image translation tasks when prompted in the right way. When prompted to create "the exact same cat as the top as a sketch on the bottom", the model produced the following images. How well the model encapsulates provided information, depends on the phrasing of the input caption. However, for a greater number of objects, DALL·E starts confusing the associations between the objects and their colors, and the success rate decreases sharply. In such cases, even the rephrasing of captions does not yield better results. Moving forward, OpenAI plans to provide more details on DALL·E's training and architecture in an upcoming paper. But the firm's work shows that transformer models can be a worthy competitor to Generative Adversarial Networks (GANs) in the visual domain. "We’ve found that [DALL·E] has a diverse set of capabilities, including creating anthropomorphized versions of animals and objects, combining unrelated concepts in plausible ways, rendering text, and applying transformations to existing images," the firm wrote in its blog post.