Copilot Pro Launch, AlphaGeometry, Meta's Llama 3 & AGI, CES robots, and more

Sponsored by

Generative AI - Short & Sweet

Hi there,

Here are your weekly AI insights. This time, I was lucky to snatch a great deal with MIT. They are not known for living under a rock regarding AI.

Other than that, Microsoft launches Copilot Pro; science will leap forward massively with Google DeepMind’s AlphaGeometry, Meta’s Llama 3/ AGI, and what’s causing open-source LLMs to lag behind GPT-4 (join the discussion). Plus, CES robot news.

Lastly, you can help us immensely by forwarding this email to others. 🙏 

Martin

Innovative technologies are revolutionizing business as we know it, and they’re more accessible than ever. But to truly harness the transformative potential of AI, you need to know how and when to use it. And which pitfalls to avoid.

The six-week Artificial Intelligence: Implications for Business Strategy online short course from MIT Sloan School of Management and MIT Computer Science and Artificial Intelligence Laboratory explores AI’s business applications and challenges. 

Choose this program to:

  • Optimize your business: Leverage AI, ML, and robotics to drive efficiencies, improve productivity, and support your growth.

  • Develop a strategic roadmap: Apply your knowledge to effectively integrate AI into your business.

  • Gain a dual perspective: Benefit from a course designed by two prestigious schools — the MIT Sloan School of Management and the MIT CSAIL.

  • Conveniently build career-critical skills: Follow a program that fits your schedule and benefit from 24/7 support and various payment options.

🌱 Top GenAI News

Microsoft Launches Copilot Pro

Copilot Pro, priced at $20/month/user, integrates advanced AI capabilities across devices, leveraging OpenAI's GPT-4 Turbo. It enhances Microsoft Office Suite applications - Word, PowerPoint, Excel, OneNote, and Outlook - by offering data analysis, Excel coding support, reply suggestions, visual additions, and ever-expanding support functions.

The Copilot GPT Builder will soon enable users to create custom AI assistants (like we demoed here). Note: You need a Microsoft 365 subscription. Competing products include Tabnine, Captain Stack, GPT-Code-Clippy (GPT-CC), Second Mate, and IntelliCode.

AlphaGeometry: An Olympiad-level AI for Geometry

Last Wednesday at 8:15 PM, while preparing for a company presentation, I stumbled upon this post:

LinkedIn screenshot.

These are two major news! Let’s unpack:

  1. The capability. There are an estimated 384 Olympiad gold medalists since 1959(!). These are top mathematicians who dedicated their lives to math/ geometry. And, now, Google DeepMind (GDM) has built AlphaGeometry that can solve complex geometry problems, including those from the International Mathematical Olympiad (IMO).

    The result: tested on 30 IMO problems, it successfully solved 25. Unparalleled! As they are really hard.

  2. Synthetic data. It has been trained solely on AI-generated data, underlining the power of synthetic data.

AlphaGeometry’s workings - a novel way to solve complex problems:

  1. It uses an LLM for intuitive idea generation and a symbolic deduction engine for formal logic and rule verification.

  2. It first employs its symbolic engine for proofs; if unsuccessful, the language model suggests adding new elements to the diagram for further exploration. And so on, until solution.

  3. The development involved creating a custom language with basic geometry rules and training on 100 million machine-generated 'proofs'.

    2-step approach: language model & symbolic engine.

“AlphaGeometry's output is impressive because it's both verifiable and clean…It uses classical geometry rules with angles and similar triangles, just as students do.”

Evan Chen, math coach, and Olympiad gold medalist

Where could this go?

Many ways. Of course, this has implications for mathematics. For instance, AlphaGeometry can assist mathematicians in exploring new theorems in geometry by quickly testing hypotheses and providing proof concepts—a catalyst for science itself.

It could, for instance, assist architects in solving geometric problems in building designs, leading to innovative structures.

It could further have implications in engineering applications, computer graphics, robotics and automation, the gaming industry, and even optimization problems - propelling us forward.

Meta’s Open-Source AGI - Join the Discussion Below

Meta, led by Mark Zuckerberg, is officially developing artificial general intelligence (AGI) with plans to release it as open-source software potentially. To achieve this,

  • Meta is training its next-gen model, Llama 3, and building a massive computing infrastructure, including 350,000 H100 GPUs.

  • Meta/ Zuckerberg links AI with the metaverse, envisioning smart glasses as a key interface.

This is the announcement by Marc, who looks like he’s spent quite some time in the Metaverse already:

Will they open-source it? I have reasons to believe so.

  • They have a strong open-source culture with a respective track record.

  • Meta and IBM have formed the AI Alliance to promote open-source AI.

  • They have Yann, who is an advocate for open sourcing. See this video, for instance.

However, Zuckerberg's control over Meta's stock centralizes the decision to open-source AGI.

On open-source performance (a LinkedIn conversation)

I recently posted that open-source models are behind closed-source models for overall performance, according to the benchmarks.

With my post's initial lack of nuance (guilty!), the comments forced a tremendous, in-depth conversation. Some points:

  • GPT-4 is currently the strongest model, according to the main benchmarks.

  • As long as open-source models are trained and fine-tuned on GPT-4 generated data, they will not surpass GPT-4.

  • Customization and context > overall performance.

Want to contribute to the conversation? Join here.

🌟 Our Favourite Robot Announcements from the CES 2024

Two significant trends stand out: pervasive AI and its integration with robots. Today, we showcase top robot announcements. Last week, we have focused on the AI trends (here). Expect both trends to have a significant impact on our everyday lives.

Robot - Announcements

Samsung Ballie: An AI robot acting as an entertainer, chatbot, and home assistant. It uses vision AI to suggest based on fridge contents (see AI integration above), integrates with LLMs in Samsung phones, and even projects videos or video calls. (Source)

LG AI Agent: A household assistant robot that doubles as a smart home hub. Functions include pet monitoring, security checks, medication reminders, emergency calls, and mood-reading based on voice and facial expressions. (Source)

ElliQ: Designed by Intuition Robotics, this voice-operated AI care companion is popular among the elderly. It assists with daily check-ins, wellness goals, and more. (Source)

Roborock S8 MaxV Ultra: A vacuum and mop robot with a robotic arm. Operable by voice commands, it avoids obstacles and allows owners to make calls through its built-in camera. (Source)

🐦️ IMO, The Best AI-Generated Clip (Click Screenshot)

Click on the image to go to the Reddit link. It’s the best AI-generated video so far.

Wait! Tell me, what topic area are you most excited about? 😏

Login or Subscribe to participate in polls.

🎥 Newsletter as a Video

🙂 Hot Notes

Thank you so much for reading