The development of the Mojo programming language was led by Chris Lattner, who was also one of the key drivers of Apple’s Swift programming language.

Lattner specialized in Xcode development during his tenure at Apple, then spent time at Tesla before joining Google’s Brain AI program in 2017.

Lattner co-founded Modular AI in 2022 to develop the Mojo programming language.

The Mojo language was made available for download in September of this year, and will initially be supported only on Linux. In addition to a compiler, the Mojo SDK includes a suite of developer and IDE tools for building and iterating on Mojo applications.

Since the launch of the Mojo programming language on May 2, more than 120,000 developers have signed up to use the Mojo Playground, and more than 19,000 developers are actively discussing Mojo on Discord and GitHub, Modular AI said.

Mojo is described as a high-performance “Python++” programming language for computation, aimed at AI developers, which will evolve into a superset of Python over time.

Currently, Mojo already supports seamless integration with any Python code and has an extensible programming model for performance-critical systems, including gas pedals (e.g., GPUs) that are prevalent in AI.

THIS IS A SPONSOR PROMOTION: >>>>>>>>>>>>>

Geekwills is an online shop that connects consumers with millions of products and brands around the world with the mission to empower them to live their best lives. Geekwills is committed to offering the most affordable quality products to enable consumers and sellers to fulfill their dreams in an inclusive environment.

The post AI programming language Mojo lands on the Mac platform appeared first on TechGoing.

Adobe has collaborated on a solution called Content Credentials, which involves a number of partners such as the BBC, Microsoft, Nikon and Truepic.

With the solution, Adobe adds information about the source of an image, video, or PDF file to the metadata, which can be used to indicate whether AI algorithms were used in the creation or editing of that data file.

The information will be automatically added to Adobe Photoshop and Premiere, and Microsoft plans to support it in Bing Image Creator.

When content is displayed in applications or browsers that support this feature (implemented in JavaScript), a logo in the form of the lowercase letter “cr” will appear in the upper right corner of the image.

When the user clicks on it, a pop-up window appears with information about the image. Problematic files can also be uploaded to a special website, which will display the necessary metadata.

Adobe isn’t the only company making good on its promise to provide AI content labelling after Google launched its own solution called SynthID. Recently American scientists debunked the validity of this approach, arguing that it’s not difficult to forge metadata.

Embedded the video below:

The post Adobe launches cr LOGO to identify AI content appeared first on TechGoing.

▲ image source Stability AI

The Stable LM 3B model has 3 billion parameters, focuses on text generation, has an autoregressive system based on the Transformer decoder architecture, and uses several open-source large-scale datasets for training.

▲ Image courtesy of Stability AI

The model uses 3 billion parameters and 256 NVIDIA A100 40GB GPUs to train it, which means that although it has fewer parameters than similar large models, the performance is still impressive, and it is more suitable for mobile platforms due to the smaller size and lower power consumption of the model.

In addition, the model has multi-platform compatibility and allows for fine-tuning according to specific needs. Currently, the model has been open-sourced on the Hugging Face platform, which makes it easy for developers to use and improve it, so interested parties can go there to learn more about it.

The post Stability AI Launches LM 3B, Claimed to Bring High-Performance Experiences to Mobile Users appeared first on TechGoing.

On September 21st, OpenAI launched DALL・E 3, which not only made a huge leap in drawing ability, but also “understood” human speech. The newly upgraded DALL・E 3 has also become the undisputed leader in the field of AI painting.

Netizens who joined in the fun also posted comparison pictures of the same prompt using DALL・E 3 and Midjourney.

It can be seen that DALL・E 3 does a better job than Midjourney in “understanding the intention”, and the images it generates are more accurate.

This may be because, in this update, OpenAI announced that DALL・E 3 is integrated with ChatGPT.

This will mean that we have to say goodbye to complicated prompt projects. In the future, we only need to describe the images we want to ChatGPT, and we can generate a large wave of beautiful and different styles of images!

As soon as the news of DALL・E 3’s closed beta came out, all kinds of talk about Midjourney being slapped became rampant on the Internet.

Midjourney looked at this wave and came back with a piece of bombshell news:

• Midjourney will launch an even more amazing version 6 within this year, which will not only make a huge leap in the quality of image generation, but also add social features.

In addition, Midjourney will also launch 3D and video generation, although we will not see these features within this year.

After the release of DALL・E 3, Midjourney CEO David Holz was asked what he thought of OpenAI’s DALL・E 3. Holz responded “arrogantly”:

• Midjourney will continue to provide the highest-quality images.
• In this comparison between the DALL・E 3 and Midjourney V5, you can actually see that the DALL・E 3 is not far ahead in terms of image quality, although it does follow prompts and render text better.

Although Midjourney V6 has not been officially launched yet, there are already rumours about various version updates on the Internet!

Let us take a quick look next.
Midjourney V6 updates

1. Independent website development

Midjourney will be spinning off from Discord and launching its own independent website. And will integrate user-friendly social mechanisms on its homepage.

2. Mobile Application Development

Midjourney expects to expand its reach through mobile apps. iOS users are known to get the app natively, while Android users will download it through other channels.

For users who prefer to browse the web, the Midjourney website will have website pages optimized for use on mobile devices.

3. Data center upgrade

Midjourney will update its data center to provide a faster and more efficient user experience. This results in faster image generation, especially beneficial for professional and SVIP users. However, due to limited computing resources, it may not be immediately available to all user levels.

4. Add Turbo mode

Midjourney provides a feature called “Turbo Mode” that is currently available to users of all levels. Performance in this mode is four times faster than standard mode. Users can activate Turbo mode by adding the command –turbo or using /turbo.

Netizens predict the updated features of Midjourney V6

1. Higher image resolution

Images generated by Midjourney V6 will have a resolution of 2048×2048. This leap in resolution not only enhances the richness of the generated image content, but also enables complex designs and nuances that were previously unattainable.

2. Further improved natural language processing

Midjourney V6 can better understand users’ text input and achieve the same communication and communication as fellow artists. This will result in more accurate and detailed image generation.

3. Video generation function

Midjourney V6 will seamlessly transform static images into dynamic and moving scenes. Videos will be able to carry more content and plot than pictures, helping creators better tell their stories.

4. 3D modeling function

Previously, delays in developing 3D features for Midjourney were due to a lack of data on 3D objects. But in later versions, Midjourney will launch 3D functionality. Compared with 2D, the 3D generated by Midjourney will have almost no quality loss.

5. More precise generation control

Being able to change specific elements in an image without changing the entire composition, this kind of fine-tuning control will bring more flexibility to the creation, giving the creator more room for adjustment, and can push the resulting image into new directions. high.

And, the updated Midjourney V6 more accurately reflects prompt parameters in generated images, giving creators greater control over their creations.

6. Drag and drop controls

New features that Midjourney is working on, like DragGAN, a new AI editing tool that allows you to make changes to images using simple drag-and-drop controls on them, may only be seen in the V7 release.

The post Midjourney V6 is coming, competing for the title of strongest mapping AI on the surface appeared first on TechGoing.

Summarizes as follows:

Dream Screen: AI can be used to generate videos and photos that creators can use as backgrounds for YouTube shorts and then insert their own content. Initially, creators will be able to enter prompts or commands to generate backgrounds, but YouTube hopes that creators will eventually be able to use AI tools to remix and orchestrate existing content to create new content.

AI Insights (Video Topic Ideas and Outline Generator): This feature will be integrated into the YouTube Studio creator management platform. AI can personalize video topics based on the creator’s personal situation and audience trends.

AI-based music recommendation system: It can recommend appropriate background music or audio based on the creator’s description of the video’s characteristics and key points.

Aloud: Also integrated within YouTube Studio, AI-generated dubbing of videos (other than their native language) in another language with just one click allows creators to pre-check it before adding it to the video. The tool is currently being tested on a small scale and will be available to a wider audience next year.

The post YouTube launched a series of AI tools that generate backgrounds and dubbing appeared first on TechGoing.

IT House learned from the Wall Street Journal that the function of this model will be several times more powerful than the commercial version of Llama 2. It is planned that the model (details are subject to change) will help other companies build services that produce complex text, analysis and other output.

Llama 2 is an open-source AI language model launched by Meta in July. It is distributed by Microsoft’s cloud Azure service and competes with OpenAI’s ChatGPT and Google’s Bard.

Meta reportedly formed a team earlier this year to build the model, with the goal of accelerating the creation of AI tools that can mimic human expressions.

The post Meta plans to develop a new large-scale language model early next year appeared first on TechGoing.

Short videos have now become an important form of streaming media in the mobile era. According to reports, Samsung hopes to simplify the operations of users to shoot, produce, and optimize short videos in post-production through the cooperation between the two parties. It is estimated that a series of filters, beauty tools, and subtitles will also be provided. Options and other tools, and also provide AI-related functions.

Note: Irrelevant Labs is a start-up company headquartered in Washington, USA. Its main business is to introduce various AI tools to phones.

Samsung has now strengthened the AI ​​layout on the smartphone side, and has cooperated with Qualcomm and Meta on Llama 2.0 LLM, which means that large language models can be run locally on phones.

The post Samsung invests in Irreverent to introduce short video AI generation tool on phone appeared first on TechGoing.

The biggest change is a new section on using Microsoft’s artificial intelligence services, which involves using its Bing Chat chatbot as well as Windows Copilot and Microsoft 365 Copilot services. Here’s the full text of that section:

1. Reverse engineering. Artificial intelligence services must not be used to discover models, algorithms, and any underlying components of the system. For example, you cannot try to determine and remove the weights of the model.
2. Extract data. Do not use web scraping, web harvesting, or web data extraction methods to extract data from the AI Service unless expressly permitted.
3. Restrictions on the use of AI service data. Do not use the AI Services or data from the AI Services to create, train or improve (directly or indirectly) any other AI Services.
4. Use your content. As part of providing the AI service, Microsoft will process and store your input to the service and the output of the service in order to monitor and prevent abusive or harmful use of or output from the service.
5. Claims by third parties. You are solely responsible for responding to any third-party claims related to your use of the AI Services (including, without limitation, copyright infringement or other claims related to content output during your use of the AI Services) in accordance with applicable law.

In addition, the definition of “your content” in the agreement has been expanded to include “content generated by you using our artificial intelligence services”. Additionally, the “Code of Conduct” section now contains information on the use of artificial intelligence services.

There are a number of other smaller changes to the service agreement, including a new Microsoft Storage section detailing what OneDrive and Outlook.com are for. There have also been some changes to the Microsoft Rewards section to reflect that it is now available globally.

The post Microsoft updates service agreements to prevent data scraping via AI services appeared first on TechGoing.

In 2021, Microsoft introduced Jenny Multilingual, a female voice that can naturally speak many different languages. Today, Microsoft announced an upgrade to Azure AI text-to-speech. Not only does it add more language support, it also adds a male voice — Ryan Multilingual.

In a blog post, Microsoft said that this addition not only provides users with more options, but also enables them to create inclusive and diverse user experiences. By introducing male voices, Azure Text-to-Speech ensures that the technology adapts to different scenarios and user preferences, enabling developers to design solutions that resonate with a wider range of end users. Additionally, both new voices feature automatic language prediction that automatically recognizes the input language based on the input text and adjusts the speech output accordingly, eliminating the need for manual tagging. In addition.

This update expands the number of supported languages from 14 to 41, as listed below:

ar-EG

ar-SA

ca-ES

cs-CZ

da-DK

de-AT

de-CH

de-DE

en-AU

en-CA

en-GB

en-HK

en-HK

en-IN

en-US

en-US

es-MX

(b) the organization of the work of the Committee on Conventions and Recommendations (CRIC)

fi-FI

fi-FI

fr-CH

fr-FR

hi-IN

hu-HU

hu-HU

it-IT

ja-JP

ko-KR

nb-NO

nl-BE

nl-NL

pl-PL

pt-BR

pt-PT

ru-RU

sv-SE

th-TH

tr-TR

zh-CN

zh-HK

zh-TW

In addition it is noted that Microsoft also said it has developed new systems that automatically change the pronunciation, intonation and phonology of its AI voices to best match the specific language of the input. This breakthrough enables users to easily switch between languages during text input, creating a more immersive and authentic multilingual experience, Microsoft said.

The new male voices, language support and adaptive system are currently in public preview in Azure’s Eastern U.S., Western Europe and Southeast Asia regions.

The post Microsoft Azure AI text-to-speech service upgrade, male voice and more language support appeared first on TechGoing.

▲ Image source 9To5Google

For example, foreign media said that when a user enters “The dogs aren’t outside today it is too hot grammar check” in the Google search box, after that, Google will prompt the user that there may be problems in spelling, and they should enter “The dogs are outside today it is too hot grammar check”.

It is reported that Google will bold and emphasize the grammatical errors in the sentence. If the search engine does not detect grammatical errors, a “green mark” will appear to inform the user that the grammar is correct, and the user can also hover the mouse over the corrected grammatical error. , use the shortcut “Copy” button to copy the corrected sentence.

Foreign media revealed that the support page for this feature seems to have been online for at least a few weeks, and they also contacted Google spokesperson Colette Garcia. Garcia said that Google had actually quietly rolled out this “syntax check” feature at the end of June.

 When a user enters a sentence with grammatical errors, the search engine will automatically give the user the correct spelling scheme.

Colette Garcia also claimed that the “grammar check” function uses AI to work, but he also warned that the current “grammar check” function may not be 100% accurate, and it will take some time to improve.

The post Google launched an AI grammar detection tool to help users improve their writing skills appeared first on TechGoing.