AI Highlights

My top-3 picks of AI news this week.

NVIDIA
1. NVIDIA Drops The Mic

At CES 2025, NVIDIA CEO Jensen Huang unveiled a comprehensive suite of innovations spanning personal computing, robotics, autonomous vehicles, and AI development.

• Project DIGITS: A $3,000 personal AI supercomputer that can run models up to 200B parameters (405B with dual units), bringing data centre-class computing to desktops.

• GeForce RTX 50 Series: Built on Blackwell architecture with 92 billion transistors, delivering 3,352 TOPS and featuring DLSS 4 with Multi Frame Generation for up to 8x performance boost.

• NVIDIA Cosmos: A new world foundation model platform for physical AI, powering robotics and autonomous vehicles with advanced simulation capabilities and detailed virtual environments.

Alex’s take: Jensen's vision has consistently been 5-10 years ahead of the market. What's most impressive is how he's steered NVIDIA from a gaming company to the backbone of AI infrastructure. From inventing the GPU in 1999 to this week’s announcements, Jensen articulates complex technological breakthroughs with remarkable clarity. It might have something to do with the energy he exudes from wearing his black leather (or snakeskin) jacket.

xAI
2. Grok breaks free

Elon’s xAI is expanding access to its Grok chatbot beyond X (formerly Twitter) with a standalone iOS app and upcoming web platform.

• Broader accessibility: Previously limited to X's paying subscribers, Grok is now available through a dedicated iOS app in multiple countries, including the US, Australia, and India.

• Real-time capabilities: The chatbot can access current data from the web and X platform, while offering features like text rewriting, summarisation, Q&A, and image generation.

• Unrestricted creation: Unlike competitors, Grok's image generator allows the use of public figures and copyrighted material in image creation, focusing on photorealistic rendering.

Alex’s take: Menlo Ventures released a great report at the end of 2024 highlighting the shifting market share for LLM providers. From 2023 to 2024, OpenAI dropped from 50% to 34%, whilst Anthropic doubled from 12% to 24%. It reminded me of when Facebook came to market in ‘06/07 against the current incumbent, Myspace. I believe OpenAI’s market share will continue to be eroded this year, with competitors like xAI, Google and Claude competing for the top spot.

Meta
3. Meta's in the Bad Books

Meta faces serious allegations this week regarding their AI training practices, with internal communications revealing potentially controversial decisions at the highest level:

• Executive Approval: Mark Zuckerberg reportedly approved the use of LibGen, a dataset of pirated books, despite internal warnings from Meta's AI team about regulatory risks.

• Legal Challenge: High-profile authors including Ta-Nehisi Coates and Sarah Silverman are pursuing legal action, claiming copyright infringement in training Meta's Llama model.

• Timing Crisis: This revelation comes at a critical moment when major tech companies, including Meta, are racing to train AI models while facing an increasingly limited pool of legitimate training data.

Alex’s take: The pressure to train competitive AI models is clearly pushing companies to make controversial decisions. What's particularly interesting is how this matches Elon Musk's recent claim that human-created training data has already been exhausted. Perhaps this explains (though doesn't justify) why Meta might risk using questionable datasets—they're desperately searching for untapped sources of training data.

Content I Enjoyed

Reflections By Sam Altman

This week, Sam Altman released a new blog post titled “Reflections”.

He says he is confident OpenAI knows how to build AGI and that Superintelligence is next.

Sam Altman has previously defined AGI as “A median human that you can hire as a co-worker.” This includes the meta-skill of learning to figure things out.

What’s more, I’m seeing many OpenAI employees talk openly about AGI. Will we have it sooner than expected, or are they driving the hype train?

Altman’s timeline is particularly interesting. By 2025, he expects the first AI agents to “join the workforce.” That's not far off at all. I see myself using agents (AI systems that can reason through problems and take action for you) daily, especially for repetitive and monotonous tasks like research and coding, when all I want is the final output.

As a concluding note, OpenAI started as an open-source non-profit. It has now turned into a closed-source for-profit. I believe a for-profit entity controlling superintelligence would destroy the safeguards that must keep advanced AI aligned with public safety.

After all, Artificial Superintelligence (ASI) is inevitable. How we build it and what we do with it will be all that matters.

Idea I Learned

How I Built My SaaS Tool With AI

Some of you may have found me through my content on LinkedIn.

I love posting the latest AI updates and insights and sharing what I learn along the way.

Needless to say, content creation can take A LOT of my time.

So I wanted to make a software tool that uses AI to streamline my writing process.

It would have to ingest all of my learnings and secrets from the past two years of writing online for what makes a high-performing LinkedIn post.

I wanted to create something that both I and my followers could use to build their brand on LinkedIn.

V0 has been integral to this process. It’s a tool that allowed me to use plain English to build the user interface (UI) for my new software product, GrowHub.

I come from a non-technical background, so being able to leverage AI tools for development has been a game-changer.

→ If you want to learn how to get started with v0, check out my YouTube video.

→ If you want to start building your personal brand on LinkedIn, check out GrowHub.

Quote to Share

Logan Kilpatrick on Google’s DeepMind consolidation:

Google's advantage has been quietly compounding over time. We're only now seeing the emergence of this come to light.

While OpenAI captures headlines with consumer success (75% of their revenue comes from paying consumers), both Google and Microsoft have been deeply embedded in enterprise infrastructure all this time, and that is why they have the advantage.

The strategic consolidation under DeepMind is just the latest sign that Sundar is cooking something special. By bringing Gemini API and AI Studio under the DeepMind umbrella, Google is leveraging their enterprise foundation—built through years of Workspace and Cloud service integration—to position itself for AI dominance.

Source: Logan Kilpatrick on X

Question to Ponder

If we don't introduce novel ideas, there's a risk that generative AI could reinforce echo chambers by amplifying existing biases and preferences.

However, when we blend unique (and very human) ideas with generative AI to elevate them, we can create unmatched content and experiences.

I’ve seen it all over X and LinkedIn: Some people generate content that adds little value (all of it generated by AI). But others bring fresh perspectives to our timelines, using these tools to create something marvellous.

It's up to us as individuals to use generative AI to amplify our creativity—not just replace it—and break free from these bubbles.

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