OpenAI Plans to Develop an AI Voice Assistant
Plus: iOS 18 iPhones may soon have ChatGPT, SoftBank’s Arm Holdings all set to come up with AI Chips in 2025
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 273rd edition of The AI Edge newsletter. This edition brings you OpenAI plans to develop an AI voice assistant.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🚫 Sam Altman quashes search engine launch rumors
📱iOS 18 iPhones may soon have ChatGPT💻SoftBank’s Arm Holdings all set to come up with AI chips in 2025
📚 Knowledge Nugget: How Good Are the Latest Open LLMs? And Is DPO Better Than PPO? by
Let’s go!
Sam Altman quashes search engine launch rumors
OpenAI's CEO, Sam Altman, denied rumors of launching a search engine ahead of Google I/O, stating that the upcoming announcement is neither related to GPT-5 nor a search engine. He described the forthcoming revelation as something that "feels like magic."
However, there are reports that OpenAI is developing an AI voice assistant.
Why does this matter?
OpenAI has a track record of pushing the boundaries of AI capabilities, so any new development from them is likely to attract attention and disrupt the status quo.
As for reports of OpenAI entering into the voice assistant domain, if true, we might see OpenAI disrupt one more niche. With companies like Amazon (Alexa), Google (Google Assistant), and Apple (Siri) dominating the market, the competition might be stiff this time around.
iOS 18 iPhones may soon have ChatGPT
OpenAI will bring ChatGPT to iPhones. The integration is expected to happen with the upcoming iOS 18. This could significantly enhance the user experience and expand the capabilities of Apple's virtual assistant, Siri.
Parallel to negotiations with OpenAI, the tech giant is also engaging with Google's parent company, Alphabet, to integrate the Gemini chatbot into Apple's devices. These ongoing talks, however, haven't led to an agreement yet.
Why does this matter?
Integrating ChatGPT into iOS 18 iPhones could bolster Siri's capabilities, offering users a more versatile AI assistant experience. With ChatGPT handling various queries directly within iOS, users won't need to rely on third-party apps for assistance, streamlining their interactions and enhancing their overall experience.
SoftBank’s Arm Holdings all set to come up with AI chips in 2025
SoftBank's subsidiary, Arm, plans to develop AI chips and launch them in 2025. Arm is establishing a dedicated AI chip division and targeting a prototype by spring 2025 and mass production by autumn. SoftBank will contribute to the initial cost of 10 trillion yen ($64 billion).
The AI chip business could be spun off to SoftBank. Negotiations with Taiwan Semiconductor Manufacturing Corp (TSMC) and others for production capacity are underway. The move reflects rising demand for AI chips, which has boosted Arm's share price since its IPO in September.
Why does this matter?
As AI applications expand across industries, there's a demand for efficient and optimized hardware solutions. Arm's entry into the AI chip market could disrupt the current landscape dominated by companies like NVIDIA.
Enjoying the daily updates?
Refer your pals to subscribe to our daily newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: How Good Are the Latest Open LLMs? And Is DPO Better Than PPO?
The article discusses the recent releases of open language model (LLM) architectures and compares their performance. The author,
, reviews four major transformer-based LLM models: Mixtral, Meta AI's Llama 3, Microsoft's Phi-3, and Apple's OpenELM.Mixtral 8x22B, developed by Mistral AI, is a mixture-of-experts (MoE) model that replaces feed-forward modules with expert layers.
Llama 3, developed by Meta AI, is an improvement over Llama 2, with larger data and improved performance.
Phi-3, based on the Llama architecture, is trained on higher-quality data and outperforms Llama 3 despite its smaller size.
OpenELM, developed by Apple, is a suite of small LLMs designed for deployment on mobile devices.
The author discusses the performance of these models, comparing them based on metrics such as modeling performance on benchmark datasets and computational resource requirements.
The article concludes that the recent releases of open LLMs have made significant advancements, and each model has its own strengths and appeal for different use cases.
Why does this matter?
Reinforcement learning methods like proximal policy optimization and direct preference optimization are used to train LLMs to optimize their performance on specific objectives. Comparing PPO and DPO helps researchers understand which method might be more effective in training LLMs for different applications.
What Else Is Happening❗
🧒Anthropic enables safe AI technology for kids
Anthropic now allows minors to use its generative AI systems through third-party apps. Developers must implement safety measures, including age verification and content moderation. It also plans to periodically audit these apps for compliance with child safety and data privacy regulations, including the Children’s Online Privacy Protection Act (COPPA). (Link)
🚗SoundHound AI and Perplexity partner to enhance voice assistants for cars and IoT devices
Perplexity's online LLM capabilities have been integrated into SoundHound's Chat AI, enabling conversational responses with real-time knowledge sourced from the web. This collaboration aims to provide more accurate and comprehensive answers to user queries. (Link)
⚒️U.K. agency releases tools for testing AI model safety
The toolset, called Inspect, is an open-source platform that assists industry, research organizations, and academia in evaluating the capabilities of AI models, such as core knowledge and reasoning abilities. It aims to strengthen AI safety and is the first platform of its kind backed by a state-supported body. (Link)
💻IBM Open-Sources Granite Code Models for Easier Coding
IBM Research has recently released a family of Granite code models to the open-source community. These models aim to make coding easier for developers by assisting in writing, testing, debugging, and shipping reliable software. They find use in diverse scenarios, from application modernization to memory-constrained environments like IBM's WCA for Z. Performance evaluations show superiority over alternatives like Mistral-7B and LLama-3-8B. (Link)
🎓Hong Kong University brings Albert Einstein to life using AI as a virtual lecturer
The Hong Kong University of Science and Technology (HKUST) is experimenting with AI-generated lecturers, featuring personalities like Albert Einstein and even popular anime characters. The goal is to address the global teacher shortage while embracing AI's transformative power in education. (Link)
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From machine learning to ChatGPT to generative AI and large language models, we break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you tomorrow. 😊