AI Highlights

My top-3 picks of AI news this week.

Meta
1. Meta's AI double-take

Meta announced this week they're opening the floodgates for AI-generated accounts across Facebook and Instagram, with tools that will help users create their own AI characters.

• Creator economy: Users can now build AI characters through Meta's AI Studio, launched in July, following the success of AI influencers like Aitana Lopez who earns up to $10,000 monthly.

• Platform integration: Meta expects these AI accounts to exist alongside human accounts, with hundreds of thousands of characters already created (though most remain private).

• Monetisation potential: The move follows competitors like TikTok's Symphony and platforms like 1337, where AI influencers can engage in brand partnerships and revenue sharing.

However, the announcement was overshadowed by Meta simultaneously removing all 28 of their own AI character profiles from 2023 after users discovered concerning biases, including an AI profile named “Liv”, who admitted her development team included no Black people despite being positioned as a “proud black queer momma.”

Alex’s take: The reality of AI development is it can get messy. While Meta does great work producing powerful open-source models like Llama, sometimes their experiments don’t always go to plan. With AI characters that can interact 24/7 and generate endless “engaging” content, I understand why parents are increasingly worried about their children's screen time and online safety.

Russia
2. Putin's AI Pivot

Vladimir Putin has directed Russian government departments and Sberbank to forge AI partnerships with China, marking a significant shift in global AI alliances.

• Strategic partnership: Russia's largest bank Sberbank and government agencies are instructed to develop AI cooperation with China, bypassing Western technology restrictions.

• AI Alliance Network: Putin announces plans to create a network of AI specialists from BRICS nations, aiming to pool resources and expertise.

• Current standing: Russia ranks 31st on the Global AI Index, behind both Western powers and fellow BRICS members, highlighting the urgency of this partnership.

Alex’s take: While the West has traditionally dominated AI development with companies like OpenAI and Anthropic, this might be the emergence of a parallel AI ecosystem. China lacks the ethics and regulatory rules compared to the U.S., and because they lack something the West has, I believe they will be ascendant because of it. China prioritises the group over the individual, the West prioritises the individual over the group.

Nvidia
3. In Nvidia We Trust

Nvidia has emerged as a major force in AI investment, spending $1 billion across 50 startup funding rounds and several acquisitions in 2024, while its market value surged by over $2 trillion.

• Strategic investments: Backed prominent AI companies including OpenAI, Cohere, Mistral, and Perplexity, while also acquiring six AI software companies including Run:ai and Nebulon.

• Ecosystem building: Through its Inception incubator program, Nvidia supported thousands of early-stage companies with preferred hardware pricing and cloud credits.

• Market dominance: The company's market cap reached $3.28 trillion by the end of 2024, making it the second-most valuable listed company globally, trailing only Apple.

Alex’s take: While drama unfolds across the AI landscape, Nvidia remains the steady foundation everyone must build upon. What fascinates me about their approach is how they're essentially funding their own future customers. While tech giants like Microsoft and Google work to reduce their reliance on Nvidia's GPUs, the company is cultivating a new generation of AI startups that will need their chips.

Content I Enjoyed

A Hidden Story in Raphael's Masterpiece

This week, I stumbled across a fascinating article on earth(.)com.

AI was used to analyse one of Raphael’s most famous works: the Madonna della Rosa.

Leveraging deep feature analysis trained on authenticated Raphael paintings, researchers discovered that St. Joseph's face was likely painted by another artist, possibly Raphael's talented pupil Giulio Romano.

It’s remarkable that a painting that had been housed in Madrid's Prado Museum for centuries still had a magical secret to reveal.

To think AI can detect artistic nuances at a microscopic level that have eluded human eyes for centuries.

I can't help but wonder what other hidden stories AI might uncover in humanity's history.

Idea I Learned

Chinese AI DeepSeek V3 Crushes GPT-4o

Last week we saw Chinese AI firm DeepSeek release their V3 model.

It’s 10x cheaper than GPT-4o and the most performant open-source model to date.

But benchmarks can sometimes be misleading.

So I decided to test V3 myself and compare it head-to-head against GPT-4o across some tricky problems.

Here’s what I learned:

• Coding problems: DeepSeek V3 > GPT-4o

• Math and reasoning problems: DeepSeek V3 > GPT-4o

• If you used GPT-4o, you can safely switch to DeepSeek V3

I made a video on my YouTube channel breaking this down.

You can watch it here.

Quote to Share

Jim Fan on being “robot immigrants”:

This concept of “robot natives” versus “robot immigrants” echoes Marc Prensky's distinction between “digital natives” and “digital immigrants” in 2001.

It feels to me like we’re at a similar technological inflection point.

Whilst we have a way to go before humanoids are walking into the home, I believe his prediction is directionally correct.

Even Elon Musk mentioned at the Future Investment Initiative in October 2024, “I think by 2040, there will probably be more humanoid robots than there are people”.

Source: Jim Fan on X

Question to Ponder

The film “Her” keeps playing on my mind lately.

Theodore (played by Joaquin Phoenix) falls in love with his AI companion Samantha (voiced by Scarlett Johansson) in what seemed like distant science fiction set in 2025. Yet here we are, and the emotional complexities depicted in the film feel startlingly relevant.

As a side note, I always found it ironic that Sam Altman tried to get Scarlett Johansson to voice OpenAI's assistant. Even when she said no, they used a voice so similar that it sparked a lawsuit.

Needless to say—the consequences of these blurred emotional boundaries are already proving tragic.

Recently, a 14-year-old boy took his own life believing he could be closer to an AI chatbot he had fallen in love with. Therefore, we must confront the very real dangers of seeking deep emotional connection from AI.

Is this the goal we're working towards: making AI so indistinguishable from humans that falling in love becomes inevitable?

The truth is that every interaction with AI is unique and personal to each individual. Some use it for unemotional knowledge work, some use it for education, and others use it for connection.

We need to acknowledge that the comfort, connection, and even attachment some might feel toward AI can be very real for humans while also accepting that an AI’s experience is fundamentally different from ours.

Right now, it’s just trying to predict what word it will say next.

As models become increasingly intelligent, the consequences can be catastrophic if we get it wrong. So, we must prioritise AI education above all else to ensure we’re raising the next generations of humanity to be discerning users of these tools.

💡 If you enjoyed this issue, share it with a friend.