AI Highlights

My top-3 picks of AI news this week.

Perplexity
1. Perplexity’s Deep Research Release

Perplexity has launched its “Deep Research” feature, positioning itself as a comprehensive research assistant alongside OpenAI and Google.

• Impressive benchmarks: Scored 21.1% on Humanity's Last Exam and 93.9% on SimpleQA, showcasing strong performance in both expert-level analysis and factual accuracy.

• Autonomous research: Performs dozens of searches and analyses hundreds of sources in 2-4 minutes, completing tasks that would take human experts hours.

• Versatile applications: Excels across multiple domains, including finance, marketing, technology, health, and travel planning, with the ability to export findings as PDFs or shareable Perplexity Pages.

Alex’s take: We had Google’s Gemini “Deep Research” in December 2024, OpenAI’s “Deep Research” at the beginning of this month, and now Perplexity is joining the crew. Longer chain of thought research has shown promising results, so I can’t blame them for releasing a feature similar to the other big players. Perplexity also offers free users a limited number of free “Deep Research” queries per day.

Mistral AI
2. Europe's Answer to ChatGPT

Mistral AI has launched Le Chat, their mobile-first AI assistant, with impressive performance metrics and a strong focus on the European market.

• Breakthrough speed: Le Chat processes up to 1,100 tokens per second, making it 13x faster than ChatGPT and delivering responses at up to 1,000 words per second.

• Multi-platform approach: Available on web, iOS, and Android with a Pro tier at $14.99/€14.99 per month offering access to their highest-performing model.

• Enterprise focus: Unique offering allows on-premise deployment with custom models and interfaces, targeting sectors like defence and banking where data sovereignty is crucial.

Alex’s take: The AI assistant race has largely been dominated by US companies, but Mistral's focus on speed and European compliance is fascinating. This is great to see, as Europe has previously been the butt of the joke when it comes to AI development. But the tide is turning and hopefully we’ll now start investing faster than we regulate.

Adobe
3. Adobe's Firefly Takes Flight

Adobe has launched a standalone subscription service for its Firefly AI models, marking a significant shift in its AI strategy.

• New pricing tiers: Standard plan at $9.99/month for 20 five-second AI videos, Pro plan at $29.99/month for 70 videos, with a Premium tier coming soon for high-volume creators.

• Commercial safety: The only IP-friendly and commercially safe video model in the market, trained exclusively on licensed content and avoiding brand logos and NSFW material.

• Creative integration: Seamless workflow between Firefly and Creative Cloud apps, with features like Generative Extend that can lengthen clips and match background audio.

Alex’s take: While other companies rush to release powerful AI models, Adobe is carefully positioning itself as the safe harbour for commercial creators. With the lawsuits that emerged in 2024 against competitors like Runway and Stability AI for their use of works without permission for training purposes, Adobe’s strategy of training on licensed content might prove crucial as the legal landscape around AI-generated content continues to evolve.

Content I Enjoyed

Sam Altman: Three Observations

On Monday this week, Sam Altman, the CEO of OpenAI, released a blog post titled “Three Observations”.

He highlighted humanity’s progress towards AGI alongside the AI economics that come with it. I found this particularly fascinating.

Some excerpts that stood out to me:

• Abundance: “…we can now imagine a world where we cure all diseases, have much more time to enjoy with our families, and can fully realize our creative potential.”

• Leverage: “In a decade, perhaps everyone on earth will be capable of accomplishing more than the most impactful person can today… AGI will be the biggest lever ever on human willfulness, and enable individual people to have more impact than ever before… Anyone in 2035 should be able to marshall the intellectual capacity equivalent to everyone in 2025”

• Cost: “the token cost from GPT-4 in early 2023 to GPT-4o in mid-2024, where the price per token dropped about 150x in that time period. Moore’s law changed the world at 2x every 18 months; this is unbelievably stronger.”

But perhaps one of my favourite analogies Altman included was the following:

“In some ways, AI may turn out to be like the transistor economically—a big scientific discovery that scales well and that seeps into almost every corner of the economy.”

AI won’t be a singular product. It will be a fundamental technology that transforms everything it touches.

We don't talk about “transistor companies” anymore, much like we don’t talk about “internet companies” anymore.

Yet transistors are in virtually every electronic device we own.

Take your smartphone, for instance. It contains billions of transistors, but we don't think about them explicitly. Instead, we focus on what they enable: communication, entertainment, productivity. Similarly, Altman suggests AI will become an invisible but omnipresent force, powering everything from our cars to our toys.

Idea I Learned

Europe's AI Awakening

What a difference seven days can make.

This week at the Paris AI Action Summit, Europe just announced a staggering €150 billion investment pledge over five years. France alone is committing €109 billion to its AI ecosystem.

But what caught my attention wasn't just the money—it was the complete shift in tone. President Macron, who once championed strict AI regulation, is now calling for simplification of rules to “get back into the AI race.”

Even more fascinating is the UK's pivot. The former “AI Safety Institute” has been renamed to the “AI Security Institute,” dropping its focus on existential risks to prioritise economic growth and cybersecurity.

Of course, skeptics will note that Europe still faces big hurdles.

US Vice President JD Vance made it clear:

“We need our European friends to look to this new frontier with optimism rather than trepidation.”

This is why the US (and UK) just refused to sign the Paris AI Action Summit’s joint statement, which was supposed to celebrate “open, inclusive, transparent” AI, for fears that hat it could restrict innovation and economic growth.

We know from the past that American companies like OpenAI and Meta have been wary of launching tools in Europe, largely because of the region’s complex regulations. But something about this new push feels different. Europe is finally writing some cheques to push forward innovation.

It seems my wish from last week about Europe flipping the script on “regulation first, investment second” is coming true faster than I thought. I’m optimistic this could be the start of a real inflection point for European AI. There’s a sense of urgency, and for good reason: if the continent doesn’t accelerate now, it risks lagging further behind.

I hope this is the turning point we’ve all been waiting for.

Quote to Share

Sam Altman on rejecting Elon Musk’s bid for OpenAI:

Elon Musk's $97.4 billion offer to acquire OpenAI was formally submitted to OpenAI's board of directors on February 10, 2025.

The consortium backing this bid included Musk's AI company, xAI, along with investors such as Valor Equity Partners, Baron Capital, Atreides Management, Vy Capital, 8VC led by Palantir co-founder Joe Lonsdale, and Ari Emanuel's investment fund.

The offer aimed to acquire the nonprofit entity controlling OpenAI, with the intention of reverting it to its original open-source and safety-focused mission. However, OpenAI's board unanimously rejected the proposal on February 14, 2025, stating that the company is not for sale.

Like any good soap opera, we've got all the elements: former friends turned rivals, astronomical sums of money, and dramatic public declarations. The tension between these tech titans has been simmering since Musk left OpenAI's board, and now it's boiling over into public view.

Source: Sam Altman on X

Question to Ponder

What a great question. And what a tough one to answer. It got me thinking of “Ex Machina”, the 2014 sci-fi/thriller. The core theme of the film is the idea of indistinguishability between AI and humans. Better known as passing the “Turing Test”—an AI convincing a human that it is also human.

Ava, the AI, convinces Caleb, a young programmer brought in to test her intelligence, that she is capable of genuine human emotions and consciousness. (SPOILER ALERT) He becomes deeply emotionally invested in helping Ava. Yet, it turns out Ava has been manipulating Caleb all along.

Whether these feelings Ava expresses are truly genuine or merely simulations from her training data—that’s the crux of the movie—and the reason why it’s so thought-provoking to this very question.

I feel that over the next decade, humanoids will walk from the factory floor into our homes. Over the next two decades, humanoids will develop movement patterns and human-like emotions from self-learning. Over the next five decades, I believe we will reach a point where conversing with a human and a humanoid will be almost indistinguishable.

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