DeepMind AI Solves Olympiad-level Math
Plus: Google introduces new ways to search with AI, Apple research presents AIM.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 191st edition of The AI Edge newsletter. This edition brings you Google DeepMind’s breakthrough AI system for geometry problems.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🚀Google Deepmind AI solves Olympiad-level math
🕵️♀️ Google introduces new ways to search in 2024
🖼️ Apple's AIM is a new frontier in vision model training
📚 Knowledge Nugget: GPTs won't make you rich by
Let’s go!
Google Deepmind AI solves Olympiad-level math
DeepMind unveiled AlphaGeometry– an AI system that solves complex geometry problems at a level approaching a human Olympiad gold-medalist. It is a breakthrough in AI performance.
In a benchmarking test of 30 Olympiad geometry problems, AlphaGeometry solved 25 within the standard Olympiad time limit. For comparison, the previous state-of-the-art system solved 10 of these geometry problems, and the average human gold medalist solved 25.9 problems.
Why does this matter?
It marks an important milestone towards advanced reasoning, which is the key prerequisite for AGI. Moreover, its ability to learn from scratch without human demonstrations is particularly impressive. This hints AI may be close to outperforming humans (at least in geometry) or human-like reasoning.
Google introduces new ways to search in 2024
Circle to Search: A new way to search anything on your Android phone screen without switching apps. With a simple gesture, you can select images, text or videos in whatever way comes naturally to you — like circling, highlighting, scribbling, or tapping — and find the information you need right where you are.
Multisearch in Lens: When you point your camera (or upload a photo or screenshot) and ask a question using the Google app, the new multisearch experience will show results with AI-powered insights that go beyond just visual matches. This gives you the ability to ask more complex or nuanced questions about what you see, and quickly find and understand key information.
Why does this matter?
Google is effectively leveraging AI to make searching for information on the go with your smartphone more easy and effortless. So yes, the emergence of Perplexity AI certainly challenges Google’s dominance, but it won’t be easy to completely overthrow or replace it soon. Google might have some tricks up its sleeve we don’t know about.
Apple's AIM is a new frontier in vision model training
Apple research introduces AIM, a collection of vision models pre-trained with an autoregressive objective. These models are inspired by their textual counterparts, i.e., LLMs, and exhibit similar scaling properties.
The research highlights two key findings: (1) the performance of the visual features scale with both the model capacity and the quantity of data, (2) the value of the objective function correlates with the performance of the model on downstream tasks.
It illustrates the practical implication by pre-training a 7 billion parameter AIM on 2 billion images. Interestingly, even at this scale, there were no clear signs of saturation in performance.
Finally, we did not observe any clear signs of saturation as we scale either in terms of parameters or data, suggesting that there is a potential for further performance improvements with larger models trained for even longer schedules.
Why does this matter?
AIM serves as a seed for future research in scalable vision models that effectively leverage uncurated datasets without any bias towards object-centric images or strong dependence on captions.
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: GPTs won't make you rich
It’s been just over a week since OpenAI launched the GPT Store. Now, paying users can share GPTs they’ve made with the world. And soon, OpenAI plans to start paying creators based on GPT engagement.
But with the launch comes an enormous amount of hype.
In this insightful article,
unpacks why you won't make money from GPTs, why the GPT Store is (probably) a distraction, and why - in spite of all that - GPTs are undervalued by the people who need them most.Why does this matter?
GPT Store is cool, but everything is still so experimental that it could easily evolve into something radically different a year from now. It is best not to get too attached to the GPT Store or GPTs in the current incarnation and rather focus on getting the most productivity out of them.
What Else Is Happening❗
💬Amazon’s new AI chatbot generates answers, jokes, and Jeff Bezos-style tips.
Amazon is testing a new AI feature in its mobile apps for iOS and Android that lets customers ask specific questions about products. The AI tool can help determine how big a new shelf is, how long a battery will last, or even write a joke about flash card readers and make a bedtime story about hard drives. (Link)
📺Amazon is bringing its AI-powered image generator to Fire TV.
Fire TV’s new feature is powered by Amazon’s Titan Image Generator. For instance, users can say, “Alexa, create a background of a fairy landscape.” It generates four images that users can further customize in various artistic styles and pick a final image to set as TV background. (Link)
🤝Samsung and Google Cloud partner to bring generative AI to Galaxy S24 smartphones.
The partnership kicks off with the launch of the Samsung Galaxy S24 series, which is the first smartphone equipped with Gemini Pro and Imagen 2 on Vertex AI. It represents a strategic move to enhance Samsung’s technological offerings, providing users with innovative features powered by Google Cloud’s advanced GenAI technologies. (Link)
🚗Android Auto is getting new AI-powered features, including suggested replies and actions.
Google announced a series of new AI features that are launching for Android Auto, which is the secondary interface that brings the look and functions of a smartphone, like navigation and messaging, to your vehicle’s infotainment screen. It will automatically summarize long texts or busy group chats while you’re driving, suggest relevant replies and actions, and more. (Link)
🔍GPT-5 might not be called GPT-5, reveals OpenAI CEO Sam Altman.
At the World Economic Forum in Davos, Altman outlined what he sees as next in AI. The next OpenAI model will do "some things better" than GPT-4 and offer "very impressive" new capabilities. The development of AGI as possible in the near future emphasizes the need for breakthroughs in energy production, particularly nuclear fusion. (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. 😊