NEWS YOU CAN’T MISS

Saudi Arabia Backs China's Bid to Rival OpenAI

Saudi Arabia's Prosperity7 Ventures, a subsidiary of Saudi Aramco, has made a $400m investment in Zhipu AI, a Chinese generative AI startup, marking the first major foreign investment in China's growing generative AI sector.

Founded in 2019, Beijing-based Zhipu AI is a prominent generative AI startup in China, aiming to rival global players like OpenAI and Google. It has released a chatbot and a visual language foundation model, with backing from major firms like Alibaba and Tencent.

Zhipu AI, alongside competitors such as Moonshot AI, 01.ai and MiniMax, has previously relied on government funding and local cloud service providers for growth. However, Prosperity7 Ventures' investment hints at a potential shift in funding for China's AI industry, especially as US restrictions limit access to international capital.

The investment also reflects Saudi Arabia's strategic intent to cultivate a technological ecosystem capable of rivaling Silicon Valley's dominance in the AI arena.

The bottom line: The investment underscores the shifting dynamics in global AI funding amid heightened US-China tensions. Watch this space.

Will California Lead the US in AI Regulation?

California boasts 32 of Forbes' top 50 global AI companies, including players like OpenAI, Anthropic, Meta, xAI, Google, and Microsoft. Now it’s coming up with regulations for large AI models like ChatGPT.

A new bill in California, SB 1047, aims to regulate the development of advanced AI models, and has recently cleared the Senate and is headed to the Assembly. If passed, Governor Newsom could sign it into law by September.

Key measures include establishing oversight for AI tools to prevent bias, protecting workers from being replaced by AI-generated clones and regulating large AI systems capable of causing harm. The state is also cracking down on deepfakes, particularly those related to child sexual abuse and elections.

A few other key provisions:

• Developers or large AI models trained with massive data sets must implement safety measures before training.

• Developers must obtain certifications from the "Frontier Model Division" from the Department of Technology confirming the model’s safety.

• California Attorney General can pursue civil actions against violators with penalties of up to 30% of the model's development cost.

Several critics are arguing that the regulations could create barriers for startups and smaller companies, potentially stifling innovation and giving larger tech giants an advantage.

Bottom line: California's AI regulation bill could set a national precedent, but balancing innovation and concerns is crucial.

AI's New Frontier: Smaller, Cheaper, and Powerful

Forget what you know about AI needing massive computing power and eye-watering budgets. Tech giants like Microsoft, Meta, and Google are shaking things up by releasing "small" language models (SLMs) that prove that size isn't everything.

These leaner models, with fewer parameters (the variables that determine an AI's capabilities), are challenging the notion that bigger is always better. Meta's new 8 billion-parameter model, for instance, claims performance comparable to OpenAI's much larger GPT-4. Microsoft's "Phi-3-small" model is said to outperform GPT-3.5.

Why the shift? Businesses have been hesitant to adopt AI due to the high costs and hefty computing requirements of large models. Smaller LLMs offer a solution: they're cheaper to train, consume less energy, and can be easily customized. Plus, they can be run locally on a device, keeping sensitive data safe within company walls.

This doesn't mean large models are going extinct. OpenAI, the creator of ChatGPT, remains focused on building even bigger models to achieve human-level intelligence.

But the rise of small LLMs opens up a world of possibilities. They're already being embedded in smartphones, like Google's "Gemini Nano" model powering AI features on the Pixel. Imagine having a personal AI assistant in your pocket, capable of everything from drafting emails to summarizing news articles.

The AI landscape is evolving rapidly, and these "small but mighty" models are poised to play a big role in making AI more accessible and affordable for everyone.

So, keep an eye on these underdogs – they might just surprise you.

WOMEN IN AI

Meet Anima Anandkumar, Bren Professor of Computing at Caltech

Celebrating this week's woman in AI: Meet Anima Anandkumar, Bren Professor of Computing at Caltech and former Senior Director of AI Research at NVIDIA – the $2 trillion AI chip powerhouse.

Anandkumar's groundbreaking research has transformed the AI industry with innovative algorithms like tensor algorithms, enabling efficient and scalable machine learning.

A little bit about her background: Anandkumar graduated in Electrical Engineering from IIT in 2004 and earned her Ph.D. in the same field from Cornell University.

In 2017, she joined Caltech as a Professor, and in 2018, she became Director of AI Research at NVIDIA for 5 years.

Her recent accolades and awards:  

• 2023 Guggenheim Fellow

• 2023 Schmidt Sciences AI 2050 Senior Fellow

• 2022 ACM Gordon-Bell special prize for COVID-19 research

While many women (including myself) work on AI applications, a select few trailblazers like Anandkumar are pushing the very frontiers of AI research, constantly redefining what's possible.

Case in point: to accelerate scientific applications, Anandkumar has recently proposed an alternative to large language models, showcasing "neural operators" that integrate physical laws for more accurate and faster scientific simulations and predictions. Applications range from weather forecasting to medical device design.

At Nvidia, Anandkumar's team used neural operators to make a digital twin of the earth to predict climate change.

Now that's impactful, cutting-edge and downright cool.

ENTERPRISE AI CASE STUDY

University of Michigan Sets New Bar with Generative AI Innovation

Industry: Education

In a move to address privacy, accessibility, and affordability concerns, the University of Michigan has developed its own closed generative AI tools under the leadership of CIO Dr. Ravi Pendse for use by students, faculty, and staff.

The University has already seen significant adoption, with an average of 15,000 users per day, and are being used in a variety of applications, from student advising to clinical settings (with over 2500 use cases identified).

The offerings featured U-M GPT, a ChatGPT tool available only to the university community, and U-M Maizey, a customizable AI platform that could be trained on university documents to function as a personal assistant or tutor.

I had a great chat with Dr. Pendse when he was visiting Singapore a few weeks ago, and I believe the university's approach serves as a model for other institutions, and highlights the importance of considering ethical and social implications in AI development.

Read the full case study to learn more about the University of Michigan's innovative approach to generative AI.

PRO TIPS

Most successful AI leaders I know …

Most successful AI leaders I know:

• Don't fit the "Silicon Valley tech CEO" mold

• Don't have Ivy League degrees or perfect GPAs

• Haven't been coding since they were five

They are simply people who see the transformative power of AI in business and are driven to explore its possibilities.

If you want to apply AI to your business,

And are willing to put in the work,

You can do it, too.

Take one small step today.

For example: talk to a colleague about AI trends over coffee, read an article on AI startups in your field, or sign up for a short course on business and AI.