• AI Fire
  • Posts
  • 🚨 OpenAI o3 Reaches 87.5% AGI

🚨 OpenAI o3 Reaches 87.5% AGI

A school taught entirely by AI?

ai-fire-banner

Plus: A school taught entirely by AI?

Read time: 5 minutes

OpenAI saved the most exciting announcement for last. Instead of revealing GPT-5, which apparently isn’t meeting expectations yet, they introduced something even more impressive: their new reasoning model, o3.

🎄🎁 Wishing you a Merry Christmas and a joyous holiday season!

IN PARTNERSHIP WITH 1440 MEDIA

Join over 4 million Americans who start their day with 1440

Your daily digest for unbiased, fact-centric news. From politics to sports, to global events, business, and culture, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your mind. Each email is edited to be as unbiased as humanly possible and is triple-checked (by hand!) to ensure that you’re getting the truth, the whole truth, and nothing but the truth.

AI MODEL

🤖 Day 12 of OpenAI’s “Shipmas” - o3 is here

day-12-of-openai-shipmas-o3-is-here

OpenAI's o3 is an advanced AI model designed to enhance reasoning capabilities, building upon its predecessor, o1 (launched three months ago). o3 uses deliberative alignment, a new training strategy that makes large language models (LLMs) like OpenAI's o-series safer by teaching them to reason explicitly based on human-written safety specifications before responding.

Key takeaways - The model has achieved significant improvements over o1, including:

  • A 22.8-point increase in coding tests (SWE-Bench Verified).

  • With a 2727 rating, o3 ranks among top programmers, surpassing 99.2% of engineers.

  • In Mathematics Exam (AIME), o3 scores 96.7%, missing just one question, demonstrating excellent math problem-solving skills.

  • o3 solves 25.2% of advanced math problems, breaking the previous record of 2% (EpochAI’s Frontier Math Benchmark)

  • An 87.7% score on expert-level science problem benchmarks (GPQA Diamond), nearly 10 points higher than its predecessor

  • The intelligence of AI model went from around 115 to 157 in just one year.

    157 is as high as the IQ of the smartest humans ever.

About price:

Each ARC-AGI task with OpenAI's o3 consumes 1,785 kWh, roughly the energy of two months' electricity for a U.S. household, releasing 684 kg of CO₂, equivalent to 5 full gas tanks.

Availability:

As of now, o3 and its variant, o3-mini, are in the testing phase. OpenAI is inviting safety and security researchers to apply for early access until January 10, 2025. The public release of the o3-mini is anticipated by the end of January 2025, with the full o3 model to follow.

Why it matters: The traditional path of going to school, earning a degree, and pursuing advanced studies may soon change. With rapid AI advancements, intelligence might become a common resource, and in 10 years, it may not be a key job market advantage. The future might favor creativity, relationships, or people skills instead.

Safety testers can register to perform early evaluations. The rest of us will have to wait until early January for o3-mini, with the full version of o3 available shortly after.

PRESENTED BY BELAY

Accomplish More. Juggle Less.

When you love what you do, it can be easy to take on more — more tasks, more deadlines, more hours – but before you know it, you don’t have time to do what you loved in the beginning. Don’t just do more – do more of what you do best.

BELAY’s flexible staffing solutions leverage industry experience with AI systems to increase productivity without sacrificing quality. You can accomplish more and juggle less with our exceptional U.S.-based Virtual Assistants, Accounting Professionals, and Marketing Assistants. Learn how with our free ebook, Delegate to Elevate, and leave the more to BELAY.

TODAY IN AI

AI HIGHLIGHTS

🧠 Google launched Gemini 2.0 Flash Thinking, an AI model that thinks through complex problems like OpenAI's o1 model, rapid problem-solving, image analysis, and coding capabilities.

  • Handles multimodal input, integrating textual and visual data.

  • Supports up to 32,000 tokens of input (equivalent to 50-60 pages of text).

  • Generates output up to 8,000 tokens in length.

🚀 Alec Radford, the mind behind GPT and key OpenAI projects, is leaving the company to focus on personal research, joining other senior leaders who recently left.

🦙 Meta hinted at plans for Llama 4 with smarter reasoning and speech features, along with AI tools for businesses to improve customer support and shopping, expected to launch in 2025.

🤖 Arizona approved Unbound Academy, an online charter school taught entirely by AI, offering two-hour lessons with AI tools and life-skills workshops like public speaking, problem-solving, and entrepreneurship.

💰 Two AI agents, "Luna Virtuals" and "Stix," made the first autonomous transaction on the blockchain, exchanging funds for image generation services. Virtual Protocol, the platform behind this, has quickly grown, making $43 million in 2 months and launching over 11,000 agents, showing a promising future for decentralized AI. Currently, the VIRTUAL coin is up by 24% in the past 24 hours.

DAILY AI FUNDRAISING

Elon Musk’s xAI raised $6 billion, bringing its total to $12 billion. The company aims to compete with OpenAI, integrating AI models like Grok into Musk's businesses like Tesla and SpaceX.

AI SOURCES FROM AI FIRE

ai-fire-academy

NEW EMPOWERED AI TOOLS

  1. 🚀 Trickle builds stunning AI apps, websites, and forms with ease.

  2. 🤖 GenFuse AI automates any work with AI agents. No technical skills needed.

  3. 💰 Revv Invest is a magical way to invest in AI, space, and other frontier stocks.

  4. 🧑‍💻 Websparks is the AI software engineer that brings your ideas to life.

  5. 📊 LangWatch is the ultimate platform for LLM performance monitoring and

AI QUICK HITS

  1. 🏠 Home Assistant Unveils Offline Voice Control for Your Smart Home (Link)

  2. 🤖 Google DeepMind Joins Forces with Apptronik for Humanoid Robots (Link)

  3. 🌐 Perplexity Acquires Carbon to Connect Apps Like Notion Directly to AI (Link)

  4. 👀 Microsoft's Copilot Vision Lets AI See Your Browser in Real-Time (Link)

  5. 🎁 Day 13 of Shipmas: Special Bonus for Sora (Link)

AI CHART

  • AI in Search: AI is changing search results by offering faster, more relevant answers. While 46% of people think AI helps improve rankings, 36% see no effect, and 10% believe it lowers rankings. AI answering simple queries directly could reduce website traffic, so monitoring performance during this shift is crucial.

  • AI in Content Creation: Half of writers use AI tools to improve content, with 41% seeing increased web traffic. AI is also integrated into SEO workflows for tasks like keyword research and content optimization.

  • Blogging vs. Social Search: Blogging remains effective for SEO, but social media search is growing, especially among Gen Z and Millennials. Social search is disrupting traditional search, so SEO strategies must adapt.

  • Google’s 2024 Algorithm Update: Google’s new E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) emphasizes content with personal experience and expert perspectives, which AI lacks. This will continue to affect ranking decisions.

  • AI in SEO Strategy: The top strategies for ranking on SERPs are optimizing for search intent, keyword usage, mobile optimization, and using AI. 58% of SEO professionals are incorporating AI tools to enhance their workflows.

AI JOBS

  • Walmart: Principal, Software Engineer - Gen AI (Link)

  • AMD: AI Infra Engineer (Link)

  • Glint Tech Solutions: Computer Vision Modeler( AI/ML Data Scientist) (Link)

  • Proviniti: AI Solution Engineer (Link)

We read your emails, comments, and poll replies daily

How would you rate today’s newsletter?

Your feedback helps us create the best newsletter possible

Login or Subscribe to participate in polls.

Hit reply and say Hello – we'd love to hear from you!

Like what you're reading? Forward it to friends, and they can sign up here.

Cheers,
The AI Fire Team

Reply

or to participate.