Episode #41 - AI Weekly: by Aruna

Welcome back to "AI Weekly" by Aruna - Episode 41 of my AI Newsletter!

I'm Aruna Pattam, your guide through the intricate world of artificial intelligence.

Now, let's delve into the standout developments in AI and Generative AI from the past week, drawing invaluable lessons along the way.

1. AI in Research: Can It Truly Drive Scientific Discovery?

As AI evolves, we’re starting to ask: can it take on scientific research all by itself? And if so, can it do it in a way that’s clear, accurate, and trustworthy?

A new platform called “data-to-paper” is testing this idea.

Here’s how it works: with just some initial data, data-to-paper guides AI through every step of research—raising questions, designing studies, analysing results, and even writing full research papers. In many cases, the platform produced papers almost as accurate as those reviewed by humans. For simpler studies, it handled things on its own, but for more complex research, human help was still needed to keep things on track.

This approach shows AI’s potential to speed up scientific discovery while keeping it transparent and verifiable.

Could this be the future of research?

Read more here:

2. Transforming 5G with AI: SoftBank and NVIDIA's Vision for AI-Driven Services

Imagine turning cell towers from cost centers into revenue-generating hubs. SoftBank and Nvidia are teaming up to make this possible, unlocking a future where AI-driven services are integral to telecom networks.

At the Nvidia AI Summit in Japan, SoftBank showcased its groundbreaking AI-powered 5G network, leveraging Nvidia’s technology to transform network infrastructure into profit centers.

Traditionally, telecom networks use only about a third of their capacity, but with AI-RAN (Radio Access Network), operators can monetise the remaining capacity by deploying AI-driven services, from smart city solutions to real-time data analysis.

Nvidia and SoftBank estimate that for every $1 invested in AI-RAN, operators could earn around $5 in AI revenue, achieving a 219% return on investment.

Click here for more details!?

3. OpenAI Faces a Slowdown in Model Improvement

As OpenAI prepares for its next big release, "Orion," the company is encountering something unexpected: a slowdown in performance gains.

While Orion is reportedly better than existing models, early tests suggest that the leap in capability isn’t as significant as past upgrades, like the jump from GPT-3 to GPT-4. In some areas, such as coding, improvements may even be limited.

In response, OpenAI has formed a “foundations” team to explore new strategies to continue improving their models. These efforts include using synthetic data generated by AI and refining models during post-training.

This slowdown in AI improvement could mark a turning point in how we approach model development.

Could new strategies like synthetic data be the key to pushing AI forward?

Click the link to know more:

4. Saudi Arabia’s $100 Billion AI Vision – Project Transcendence

Saudi Arabia is making waves with its $100 billion AI initiative, “Project Transcendence,” aiming to become a global AI powerhouse and redefine its economic future.

Through heavy investments in AI infrastructure, talent, and partnerships with tech giants like Google, Saudi Arabia plans to transform itself into a major player in the AI landscape. This bold step aligns with Vision 2030, the Kingdom’s strategic plan to diversify beyond oil.

Project Transcendence isn’t just about AI—it’s creating an ecosystem, from data centers to AI startups, that fosters growth and innovation. By prioritising Arabic-language AI models and advanced research centers, the Kingdom aims to be among the world’s top 15 AI leaders by 2030.

5. Amazon's $110M AI Grant:

Amazon Web Services (AWS) is launching a $110 million initiative, “Build on Trainium,” to support AI research, offering grants and access to its AI chips.

Build on Trainium will award up to $11 million in credits to partner universities and individual grants up to $500,000 for researchers worldwide.

The goal?

Provide AI resources where they’re scarce.

However, some question whether this could steer academic research towards commercial interests, as corporate-funded projects often focus on business-oriented outcomes over critical ethical or societal studies.

While AWS claims no “lock-in” to its platform, recipients must publish their work openly. This move reflects a broader industry trend, where AI’s largest models and breakthroughs increasingly come from corporate labs due to resource gaps.

Click here to read the full story!

6. Should Elon Musk Be the AI Adviser to the White House?

A recent petition is urging President-elect Trump to appoint Elon Musk as a special adviser focused on AI. This raises a big question about the role of industry leaders in shaping government policy on emerging technologies.

Americans for Responsible Innovation (ARI), an AI advocacy group, believes Musk’s unique experience makes him the perfect candidate. Musk has been a vocal advocate for safe AI development, balancing rapid advancements with the need for ethical standards.

From launching xAI to championing California’s AI safety bill, his commitment to responsible AI is well known. However, some argue that Musk’s extensive business ventures, including Tesla and SpaceX, may spread him too thin for a role in government.

The call for Musk’s involvement highlights a growing trend of industry influence in government policy on AI.

Read the link for more details.

7. Coles Launches Australia’s First AI-Powered Smart Trolley

Coles has just unveiled an AI-powered shopping trolley, aiming to transform in-store experiences and bring new convenience to Australian shoppers.

Planned launch at Melbourne’s Coles Richmond Traders store in January next year, the “smart trolley” enables customers to search for items, weigh produce, scan as they go, and pay directly through an in-built EFTPOS machine—no checkout lines needed.

Created in partnership with Instacart, this AI trolley features an interactive screen that helps shoppers track their spending, locate discounts, and even win rewards through Coles’ loyalty program.

While many embrace the innovation, some are concerned about data privacy, questioning how personal information is stored and used. Despite this, Coles sees the smart trolley as a step towards a more engaging, budget-friendly shopping experience.

8. AI Doppelgangers: The Next Step in Social Media or a Step Too Far?

Meta is testing a new AI feature that lets popular influencers create “digital doppelgangers” to chat and engage with their followers on Instagram.

But is this the future of social media, or a step too far?

These AI clones mimic influencers' voices, appearances, and even personalities, allowing them to "speak" as if they were the real person.

While this may save influencers from repetitive interactions, it raises questions about authenticity.

Is social media shifting from connecting with real people to engaging with carefully crafted AI personas?

Meta’s goal is to streamline the influencer-fan interaction, but early feedback suggests that these AI versions feel more like impersonations than genuine connections.

AI doppelgangers open a door to new ways of automating social media, but at what cost to real, authentic interactions?

9. The Race to Redefine AI Benchmarks – Are Current Tests Enough?

As AI advances rapidly, traditional benchmarks may no longer be enough to evaluate new model capabilities. Tech giants like OpenAI, Microsoft, Meta, and Anthropic are now creating internal benchmarks to assess AI in complex, real-world tasks.

Current public benchmarks like Hellaswag and MMLU assess models with multiple-choice questions, but researchers argue that these tests don’t capture the complexity required for reasoning and multi-step tasks. New tests, like SWE-bench Verified, ask AI models to solve real-world coding problems, but results vary – OpenAI's latest model achieves 41.4% accuracy, while Anthropic's Claude 3.5 Sonnet hits 49%.

This rush to develop new benchmarks highlights the challenge: as models improve, they often "learn" existing tests, making it hard to measure true capability. Companies are now seeking innovative benchmarks that test reasoning and planning, but without industry-wide standards, comparing models remains difficult.

As the AI landscape evolves, redefining benchmarks is crucial to understanding each model’s capabilities.

10. Elon Musk’s AI Chatbot Grok Expands to Free Users on X

Elon Musk’s AI chatbot, Grok, initially exclusive to paying users on X, is now testing free access in regions like New Zealand, marking a significant shift in strategy.

Grok, developed by Musk’s AI company xAI, offers advanced features like image generation and understanding, powered by Black Forest Labs’ FLUX.1 model. Until now, these capabilities were available only to Premium and Premium+ users.

With this expansion, free users can explore Grok with limited queries – 10 every two hours for Grok-2 and 20 for Grok-2 mini, plus three image analyses daily. Requirements include a linked phone number and an account that’s at least seven days old.

By opening up Grok to a broader audience, xAI aims to gather more user feedback and accelerate improvements, positioning Grok to better compete with popular models like ChatGPT, Claude, and Gemini.

That wraps up our newsletter for this week.

Feel free to reach out anytime.

Have a great day, and I look forward to our next one in a week!

Thanks for your support˙