Howdy. It’s Barsee again.

Happy Thursday, AI family, and welcome back to AI Valley.

In today’s edition:

• Alibaba new model outperforms DeepSeek-R1

• OpenAI plans to introduce $20k/month PhD-level AI agents

• Elon Musk loses bid to block OpenAI’s For-Profit conversion

• Amazon new reasoning Al model to rival OpenAI and Anthropic

• Plus trending AI tools, posts, and resources

Ready, set, go…

NEW AI MODEL

Alibaba new model outperforms DeepSeek-R1

Alibaba launches QwQ-32B, a new open-source reasoning model that outperforms larger AI systems through advanced reinforcement learning, following their initial QwQ release in November 2024.

Here's what you need to know:

• Despite having only 32 billion parameters (compared to DeepSeek-R1's 671 billion), QwQ-32B surpasses models like DeepSeek-R1 and o1-mini in mathematical benchmarks (e.g., AIME, MATH) and scientific reasoning tasks (e.g., GPQA).

• The model achieves this through a clever two-step training process: first, it learns to solve math and coding problems by checking if its answers are correct, then it improves its general abilities by receiving feedback on how well it follows instructions and reasons.

• QwQ-32B requires just 24GB of GPU memory, a fraction of DeepSeek-R1's 1500GB (equivalent to 16 high-end GPUs). This makes it far more accessible and cost-effective for businesses.

• The model includes agentic capabilities, allowing it to dynamically adjust its reasoning processes based on environmental feedback. Recommended settings for optimal performance include a temperature of 0.6 and TopP of 0.95.

• QwQ-32B is released under the Apache 2.0 license, meaning businesses can freely download, modify, and use it in commercial applications without restrictions or fees.

Why it matters: 

QwQ-32B showcases how efficiency and advanced training techniques can outperform simply scaling up model size. For enterprises, this represents a major shift in AI capabilities, offering a powerful tool for complex problem-solving that is both accessible and customizable. 

TOGETHER WITH RYSE

The Smart Home disruptor with 200% growth..

No, it’s not Ring or Nest—meet RYSE, the company redefining smart shade automation, and you can invest before its next major growth phase.

With $10M+ in revenue and distribution in 127 Best Buy locations, RYSE is rapidly emerging as a top acquisition target in the booming smart home industry, projected to grow 23% annually.

Its patented retrofit technology allows users to automate their window shades in minutes, controlled via smartphone or voice. With 200% year-over-year growth, demand is skyrocketing.

Now, RYSE’s public offering is live at just $1.90/share and you can earn up to 25% in bonus shares.

Find Out More

SIDE UPDATES

🧠 OpenAI plans to introduce $20k/month PhD-level AI agents

OpenAI is reportedly developing advanced AI agents capable of handling specialized tasks like academic research and software development, with prices reaching up to 20,000 per month. The highest-tier agent is expected to perform advanced academic research, analyze complex datasets, and contribute to high-level problem-solving. The mid-tier agent will focus on software development, assisting with coding, debugging, and automation, while the entry-tier model is designed for professionals needing AI support in knowledge-based tasks.

⚖️ Elon Musk loses bid to block OpenAI’s For-Profit conversion

A federal judge denied Elon Musk’s motion to stop OpenAI from becoming a for-profit entity, citing insufficient evidence. Musk’s lawsuit accused OpenAI of exploiting his contributions and shifting its mission, but the judge found his claims unsupported. While early emails were “highly suggestive,” they didn’t meet the legal burden. The trial may be expedited to fall 2025 if Musk drops other allegations. This follows Musk’s $97 billion offer to buy OpenAI’s nonprofit arm, complicating its conversion amid investor pressure.

🤖 Amazon is working on a new reasoning Al model to rival OpenAI and Anthropic

Amazon is working on Nova, an advanced reasoning AI model set to launch by June. Nova employs a "hybrid reasoning" approach, combining fast responses with deep analytical thinking, to deliver top-tier performance at a competitive price. Amazon aims for Nova to rank among the top five AI models in benchmarks, particularly excelling in software development and math. The model is positioned to rival offerings like OpenAI’s o1 and Anthropic’s Claude 3.7 Sonnet.

🤝 Scale AI secures military contract in major step toward the era of ‘agentic warfare’

Scale AI has secured a multi-million-dollar deal with the U.S. Department of Defense’s Defense Innovation Unit (DIU) for “Thunderforge,” an initiative integrating AI into military operations. The program, involving partners like Microsoft and Anduril, aims to automate workflows, conduct warfare simulations, and enhance decision-making using advanced AI models.

🎓 OpenAI launches $50M grant program to fund Academic Research

OpenAI announced NextGenAI, a new academic consortium backed by $50 million to support AI research and education at 15 leading institutions, including Harvard, MIT, and Oxford University. The program offers research grants, compute resources, and API access to help students, educators, and researchers develop high-impact AI applications. Partner institutions will work on projects ranging from accelerating rare disease diagnosis to digitizing historical texts and public domain materials.

🔍 Google begins testing AI-only search results

Google is taking a bold step by testing a version of its search engine that relies entirely on AI-generated results. Dubbed "AI Mode," this experimental feature eliminates traditional organic search results, instead offering users conversational, AI-driven answers. While Google maintains that helping users discover online content remains a priority, AI Mode requires users to refine their queries or ask follow-up questions for more precise information. Currently, the feature is exclusive to Google One AI Premium subscribers.

📊 Crunchbase now uses AI to predict future IPOs

Crunchbase, the leading database for startup information, is now leveraging AI to forecast future IPOs, funding rounds, and acquisitions. According to The Wall Street Journal, the platform’s new prediction engine analyzes 17 years of data to generate highly accurate forecasts. Internal tests have shown the AI’s predictions to be 95% reliable, offering valuable insights for investors and entrepreneurs alike.

📱 The AI Phone powered by Perplexity AI is set to launch this year

Deutsche Telekom and Perplexity AI are collaborating on an AI-powered smartphone running the custom Magenta AI operating system. The device will integrate services like Perplexity Assistant, Google Cloud AI, ElevenLabs, and Picsart, supporting voice, text, and camera inputs for tasks such as booking flights and making reservations. Slated for a European debut in late 2025, the phone is expected to cost under $1,000, with a global rollout planned for 2026.

TRENDING TOOLS

• Reach by Artificial Societies > Test content in a simulation of your own LinkedIn audience

• Gong > Analyzes sales conversations to predict revenue and boost team performance.

• Data Science Agent > Google’s new tool for automating data analysis powered by Gemini 2.0

• ExplainGithub > Turn hours of code reading into minutes of understanding.

• Chat Thing > Build AI agents using your business data from Notion, websites, files, and more.

• MGX > The first AI dev team.

• Quadratic > A spreadsheet that chats with you, writes code, and connects to databases in your browser.

• Findaway Voices by Spotify > Create AI audiobooks and publish them directly on Spotify.

THINK PIECES / RESOURCES

• The government knows A.G.I. is coming.

• People are using Super Mario to benchmark AI now.

• A Bear Case: My predictions regarding AI progress.

• The questions ChatGPT shouldn’t answer.

• How are AI systems trained? A complete training guide.

• LLMs are unlikely to be the path to AGI.

• Al has forever altered innovation - here's how.

• How AI could shift the dynamic range of human empathy.

CONTENT CORNER

1/ These AI models might be releasing in 90 days.

2/ Man and machine are merging. Australian company Cortical Labs has unveiled the world's first biological computer that combines human brain cells with silicon hardware. 

3/ Meet Sesame: The most human AI voice assistant yet.

4/ How to solve your hard problem most efficiently using AI.

5/ AI Mode expands on AI Overviews with more advanced reasoning, thinking and multimodal capabilities.

THAT’S ALL FOR TODAY

That’s all for today’s issue, folks.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ Like what you see? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.