OpenAI Unveils GPT-4o Long Output: A 16X Boost in AI Response Capacity
OpenAI has introduced GPT-4o Long Output, an experimental model offering a significant increase in output capacity. This new variant can generate up to 64,000 tokens of output, a 16-fold increase from the original GPT-4o’s 4,000 token limit. The model maintains a 128,000 token context window, allowing users to balance input and output lengths. Priced at $6 per million input tokens and $18 per million output tokens, it aims to serve applications requiring detailed responses, such as code editing and writing improvement. Currently in alpha testing with select partners, this development showcases OpenAI’s commitment to addressing user needs for more extensive AI-generated content.
Canva Expands AI Capabilities with Leonardo.ai Acquisition
Canva has acquired Leonardo.ai, a generative AI startup, to enhance its AI technology stack. The acquisition brings Leonardo.ai’s 120 employees and advanced AI tools to Canva, including real-time image generation from text and sketch prompts. Leonardo.ai will continue to operate independently while benefiting from Canva’s resources. This move is part of Canva’s strategy to strengthen its Magic Studio generative AI suite and expand its offerings for enterprise users. The acquisition, Canva’s eighth overall, demonstrates the company’s commitment to integrating cutting-edge AI capabilities into its platform, aiming to maintain its position as a leading visual communication tool.
OpenAI Launches Advanced Voice Mode for ChatGPT, Prioritizing Safety and Realism
OpenAI has begun rolling out ChatGPT’s Advanced Voice Mode to a select group of ChatGPT Plus users. This new feature, powered by GPT-4o, offers hyper-realistic audio responses with improved latency and emotional recognition. The company has implemented strict safety measures, including the use of preset voices and filters to prevent impersonation and copyright infringement. OpenAI’s cautious approach aims to avoid controversies surrounding deepfakes and legal issues. The gradual release allows for close monitoring of usage, with plans to expand access to all Plus users by fall 2024.
Baidu Unveils Self-Reasoning AI to Combat Hallucinations in Language Models
Baidu has developed a novel self-reasoning framework for AI, addressing the challenge of factual accuracy in large language models. This approach enables AI to critically evaluate its own knowledge and decision-making processes, potentially reducing “hallucinations” or incorrect information generation. The system assesses information relevance, selects and cites pertinent documents, and analyzes its reasoning path to produce well-supported answers. In evaluations, it outperformed existing models and achieved results comparable to GPT-4 with significantly less training data. This breakthrough could lead to more reliable and transparent AI systems, with potential applications in industries requiring high trust and accountability.
Samsung Forecasts Strong AI-Driven Chip Demand as Q2 Profits Surge
Samsung Electronics reported a significant increase in second-quarter operating profit, driven by rebounding semiconductor prices amid the AI boom. The company expects AI servers to dominate the memory market in the latter half of 2024 as major cloud providers expand their AI investments. Samsung’s chip division returned to profitability, with high-end DRAM chips like HBM experiencing strong demand. The tech giant anticipates continued growth in the premium smartphone segment, fueled by AI-enabled devices. Despite challenges in the mobile division, Samsung remains optimistic about overall smartphone demand in the coming months.
