Meta AI launched MAGNeT, a text-to-audio model that balances speed and quality, achieving a 7x speed improvement. It offers potential for music production, sound design, and assistive technologies. MAGNeT's innovative design sets it apart, with features like custom masking schedules and progressive decoding for optimized learning. It excels in efficiency and quality, representing a new standard in text-to-audio synthesis.

Julie Sweet, CEO of Accenture, announced $1 billion yearly for AI training. GenAI progress is rapid; collaboration with governments and leadership is key. IBM's Arvind Krishna emphasizes retraining talent and collaboration for successful AI implementation. AI could generate $40 trillion in productivity annually by the end of the century.

Stability AI introduces Stable Code 3B, a 2.7 billion parameter decoder model for AI software development. It offers accurate code completion, running offline on regular laptops. With an extended context length, it supports 18 programming languages, including Python and Javascript, for multi-file projects. Available for both non-commercial and commercial use.

An international research team is testing AI-based therapy for tinnitus patients in a study published in Frontiers in Audiology. While AI shows promise, human therapist support remains crucial for better outcomes in managing tinnitus-related mental health symptoms.

Adobe is launching new AI-driven audio editing features in Premiere Pro to streamline editing tasks. Updates in the beta version offer interactive fade handles, AI-driven audio tags, improved waveform adjustment, and more visual enhancements for a customizable editing experience. The beta is available to all Creative Cloud subscribers for testing before official release.

Microsoft CEO Satya Nadella defended the investment in OpenAI, saying partnerships are crucial to compete in AI. ChatGPT showcases AI advancements but faces scrutiny from EU and UK regulators. Nadella emphasizes the importance of high-risk decisions for AI breakthroughs and seeks a stable partnership with OpenAI. He also acknowledges the challenge of limiting AI risks in elections.

Locofy's Lightning tool converts design prototypes from Figma and AdobeXD into code seamlessly. Raised $7.5M in funding to expand operations. Lightning integrates AI and heuristics for a five-step design-to-code process. Pricing structure is adaptable based on usage needs.

OpenAI CEO Sam Altman downplays fears of AI impact on jobs, stating AGI's development in the near future. He warns against overhyping GPT-4 model, emphasizing AI as a valuable tool rather than a job threat. Altman's leadership role at OpenAI temporarily in question due to concerns over AI safety.

CES 2024 attracted 130,000 attendees with innovations from 311 Fortune 500 companies. Key trends included AI, sustainability, and metaverse. Exciting products unveiled include Rabbit R1, Clicks iPhone Keyboard Case, LG Signature T OLED TV, Evie Ring, XREAL Air 2 Ultra AR Sunglasses, Somnox 3 Sleep Robot Pillow, Belkin Pro Auto Tracking Stand, Nanoleaf Skylight Modular Ceiling Light, Razer Iskur v2 Gaming Chair, and Proclaim Custom-Jet Oral Care System.

OpenAI lifts ban on ChatGPT for military use, now prohibits harm to individuals. Experts concerned about relaxed restrictions on military applications and potential impact on civilians in conflict zones.

Microsoft has opened up access to its AI-powered Office features for businesses of all sizes, removing the 300 user requirement. Copilot, powered by GPT-4, now offers a $30 option for all enterprises, changing the way documents are created and edited in Microsoft 365 applications.

Microsoft has launched AI-powered Copilot Pro for consumers, offering features in Office apps like Word, Excel, PowerPoint, and more for $20/month. This includes access to OpenAI models, improved design tools, and building custom Copilot GPT. Advanced users can benefit from faster performance and new features, with more to come.

Mercedes-Benz partners with Will.i.am to launch MBUX Sound Drive, connecting music with driving. Volkswagen integrates ChatGPT into its infotainment system. BMW introduces Alexa-driven LLM feature. Google enhances Android Auto for EVs. Qualcomm and JPMorgan Chase collaborate on car wallet for streamlined payments.

Large language models (LLMs) excel in natural language processing, particularly in fact-based question answering. However, evaluating answers accurately across different granularity levels poses challenges. To address this, Google proposes GRANOLA QA, a multi-granularity QA evaluation setting, considering both accuracy and information content. Using a knowledge graph and zero-shot prompts, LLMs generate answers at various granularity levels. Experimental results show a significant improvement in

CEOs are feeling optimistic about the global economy but fear survival without reforms due to climate change and AI pressures. PwC survey of over 4,700 CEOs shows concerns on economic downturn, geopolitical risks, and ability to withstand changes. Most see AI as crucial but also a source of cybersecurity risks and misinformation. Transformations are necessary for survival in the next decade, with focus on addressing challenges and opportunities of climate transformation.

Researchers at Dartmouth College are working on methods to identify and mitigate biases embedded in large language models, aiming to reduce stereotypes without compromising language capabilities. Their approach involves fine-tuning models with a dataset filled with stereotypes to target specific attention heads. This technique, not model-specific, can be adjusted for different applications, but is limited to transparent models. Adapting this approach to black-box models is their next goal.

The IMF's analysis reveals that AI could impact 40% of global jobs, with developed economies facing higher risks but also greater benefits. Emerging markets may struggle with AI integration, potentially widening global inequality. Prioritizing innovation, integration, and regulation is crucial for all economies to leverage AI effectively.

2024 will see a surge in global elections involving half the world's population. OpenAI's new measures aim to combat AI-generated misinformation, ensuring transparent voting information. The company is committed to enhancing security and accuracy in the upcoming crucial elections.

OpenAI's ChatGPT Enterprise Edition received positive feedback, with 260 companies signing up in 4 months. Enhanced features like data encryption and privacy protection cater to business needs. Pricing is flexible, and a new paid tier for small teams is available at $25/user/month. Microsoft recommends using ChatGPT Enterprise Edition for higher privacy and security. The GPT store offers customized chatbots for businesses.

Parallel text-to-speech (TTS) models are widely used for real-time speech synthesis due to their control and speed advantages. NVIDIA introduced Incremental FastPitch, reducing latency while maintaining speech quality. It combines block-based FFT modules, attention masks, and past model states for better performance in streaming TTS. The model is trained on the Chinese Standard Mandarin Speech Corpus and achieves results comparable to parallel FastPitch.