NVIDIA's New Dataset Sharpens LLMs in Math
Plus: Apple's AI updates to Spotlight and Xcode, Google open-sources Magika.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 213th edition of The AI Edge newsletter. This edition brings NVIDIA’s new dataset to improve LLMs at math.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🚀NVIDIA's new dataset sharpens LLMs in math
🌟 Apple is working on AI updates to Spotlight and Xcode
🤖 Google open-sources Magika, its AI-powered file-type identifier
📚 Knowledge Nugget: AI, again by
Let’s go!
NVIDIA's new dataset sharpens LLMs in math
NVIDIA has released OpenMathInstruct-1, an open-source math instruction tuning dataset with 1.8M problem-solution pairs. OpenMathInstruct-1 is a high-quality, synthetically generated dataset 4x bigger than previous ones and does NOT use GPT-4. The dataset is constructed by synthesizing code-interpreter solutions for GSM8K and MATH, two popular math reasoning benchmarks, using the Mixtral model.
The best model, OpenMath-CodeLlama-70B, trained on a subset of OpenMathInstruct-1, achieves a score of 84.6% on GSM8K and 50.7% on MATH, which is competitive with the best gpt-distilled models.
Why does this matter?
The dataset improves open-source LLMs for math, bridging the gap with closed-source models. It also uses better-licensed models, such as from Mistral AI. It is likely to impact AI research significantly, fostering advancements in LLMs' mathematical reasoning through open-source collaboration.
Apple is working on AI updates to Spotlight and Xcode
Apple has expanded internal testing of new generative AI features for its Xcode programming software and plans to release them to third-party developers this year.
Furthermore, it is looking at potential uses for generative AI in consumer-facing products, like automatic playlist creation in Apple Music, slideshows in Keynote, or Spotlight search. AI chatbot-like search features for Spotlight could let iOS and macOS users make natural language requests, like with ChatGPT, to get weather reports or operate features deep within apps.
Why does this matter?
Apple’s statements about generative AI have been conservative compared to its counterparts. But AI updates to Xcode hint at giving competition to Microsoft’s GitHub Copilot. Apple has also released MLX to train AI models on Apple silicon chips easily, a text-to-image editing AI MGIE, and AI animator Keyframer.
Google open-sources Magika, its AI-powered file-type identifier
Google has open-sourced Magika, its AI-powered file-type identification system, to help others accurately detect binary and textual file types. Magika employs a custom, highly optimized deep-learning model, enabling precise file identification within milliseconds, even when running on a CPU.
Magika, thanks to its AI model and large training dataset, is able to outperform other existing tools by about 20%. It has greater performance gains on textual files, including code files and configuration files that other tools can struggle with.
Internally, Magika is used at scale to help improve Google users’ safety by routing Gmail, Drive, and Safe Browsing files to the proper security and content policy scanners.
Why does this matter?
Today, web browsers, code editors, and countless other software rely on file-type detection to decide how to properly render a file. Accurate identification is notoriously difficult because each file format has a different structure or no structure at all. Magika ditches current tedious and error-prone methods for robust and faster AI. It improves security with resilience to ever-evolving threats, enhancing software's user safety and functionality.
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: AI, again
This time it's personal.
Or is it?
engages with this subject (after a sad encounter with an artist on X) and delves into the 3 levels of AI:It is simply more sophisticated methods of information aggregation and data manipulation, wrapped in “humanized” form.
The second level, which the tech industry monetizes. Users submit frequent, small tasks like marketing copy, agendas, summaries, or code, which software handles by interpreting vast amounts of data.
The truly generative arena of AIs that can produce large pieces of text or “artistic” images to order.
He explores the implication of these in creative fields, such as credit vs compensation and pragmatic challenges. He also reflects on their own stance towards AI.
Why does this matter?
The article invites readers to consider their own perspectives on these issues. The matters discussed are significant because they touch on the evolving role of AI in creative industries, the ethical considerations surrounding AI-generated content, and the challenges faced by creators in adapting to technological advancements. It could impact the future of creativity, IPR, and the relationship between humans and tech.
What Else Is Happening❗
💰SoftBank’s founder is seeking about $100 billion for an AI chip venture.
SoftBank’s founder, Masayoshi Son, envisions creating a company that can complement the chip design unit Arm Holdings Plc. The AI chip venture is code-named Izanag and will allow him to build an AI chip powerhouse, competing with Nvidia and supplying semiconductors essential for AI. (Link)
🔊ElevenLabs teases a new AI sound effects feature.
The popular AI voice startup teased a new feature allowing users to generate sounds via text prompts. It showcased the outputs of this feature with OpenAI’s Sora demos on X. (Link)
🏀NBA commissioner Adam Silver demonstrates NB-AI concept.
Adam Silver demoed a potential future for how NBA fans will use AI to watch basketball action. The proposed interface is named NB-AI and was unveiled at the league’s Tech Summit on Friday. Check out the demo here! (Link)
📑Reddit signs AI content licensing deal ahead of IPO.
Reddit Inc. has signed a contract allowing a company to train its AI models on its content. Reddit told prospective investors in its IPO that it had signed the deal, worth about $60 million on an annualized basis, earlier this year. This deal with an unnamed large AI company could be a model for future contracts of similar nature. (Link)
🤖Mistral quietly released a new model in testing called ‘next’.
Early users testing the model are reporting capabilities that meet or surpass GPT-4. A user writes, ‘it bests gpt-4 at reasoning and has mistral's characteristic conciseness’. It could be a milestone in open source if early tests hold up. (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. 😊