Good morning everyone,
A lot as happened in the spam of two weeks in the AI world, and I’ve tried my best to gather all the relevant news.
This week, we cover POTUS visit to the UK and wave of investment promises that followed ; Oracle rise from the shadow and some new research from OpenAI and Anthropic looking at the usage and adoption of AI.
A lot of promises for the UK after Trump visit
Donald Trump’s state visit to the UK wasn’t just pomp. He came with an entourage of American AI leader including Sam Altman (OpenAI) or Tim Cook (Apple), dangling promises of billions in funding if the UK eases regulation and leans closer to Washington than Brussels. The deal talk centers on infrastructure (chips, data centers) and applied AI in finance — very fit for the London ecosystem. For Europe, the risk is clear: Trump wants to split the bloc’s AI strategy by pulling the UK deeper into the U.S. orbit. Whether Britain gets real investment or just photo-ops remains to be seen.
Among the biggest promises:
Google started the show by pledging about $7 billions of investment over the next 2 years, a big figure soon made small by Microsoft pledge of $30 billions.
Palantir, already in talk with the UK Ministry of defense over a $750 millions contract, pledged $1.5 billions of investment. Perhaps a good moment to take a closer look at what Palantir actually does if you haven’t done so yet.
This full-week long show was another testament to the growing relationship between Tech companies and POTUS. A relationship that seems mutually profiting and should keep on growing notably with the announce from Sen. Ted Cruz to create a regulation “sandbox” that would allow AI companies to experiment with minimal federal oversight.
Others news from the US:
The UK is not the only country profiting for good deals, a recent investigation from the NYTimes shows how the U.A.E. is acquiring advanced computer chips through investment in a cryptocurrency startup linked to the Trump Family.
China
On the other side of the globe, the US main rival in the race toward AI dominance is also making move this week.
As Beijing and Washington step up negotiations over TikTok and trade, China started the week by prohibiting Chinese tech companies to buy Nvidia chips, just before making a move in the other direction and dropping it’s competition investigation into Google.
Jack Ma, founder of Alibaba, returns to the company leadership with the objective to boost tech development just as they release a new open-source AI agent to rival OpenAI’s deep search.
Market
You probably would not have missed it, for a brief moment, Larry Ellison became the world the world’s richest person. This came after the company he founded, Oracle, announced a $300 billions deal with OpenAI’s Stargate project, but more importantly, revenue growth that no analyst thought possible. Find out what happened with this great video from Bloomberg.
Is OpenAI shaping itself for an IPO? That is the assumption behind their transformation toward a Public Benefit Corporation that according to them and Microsoft, is getting closer to happening.
With more than $42bn is invested in just five companies - Amazon, Apple, Meta, Microsoft and Nvidia - the Swiss National Bank (SNB) turned itself into a major Silicon Valley investor, a “whale”.
Product
Not a massive fan of shoping? Lucky you, you probably won’t have to do it yourself soon. Google is paving the way for AI Agent to shop in your stead with its new Agent Payments Protocol (AP2). Mastercard, PayPal, American Express, and Worldpay are already on board
Microsoft launch new Copilot AI Agent feature in Teams. Facilitator agents will now sit in on Teams meetings, creating agendas, taking notes, and answering questions. Agents can also suggest time allotments for different meeting topics — letting participants know if they’re running over — and create documents and tasks.
Zoom, one Teams’ main competitor, also release a new update to their AI Companion, an agent that help you navigate, summarize and manage knowledge around meetings.
Notion (note taking app) introduce new… wait for it… agent. They will be able to do everything a human can do in Notion: Create docs, build databases, search across tools, and execute multi-step workflows.
At the Made on Youtube event, the company announced a few new AI tools for its creator, among which two that might be interesting for our marketing team: Ask Studio, an AI chatbot that creators can use to ask analytics questions about how their content is performing ; and a new A/B testing for title and thumbnail feature.
Industry adoption
Monday.com (project management software), Nokia (hardware) and Amazon (retail) are upping their AI game with one goal in mind: more top-line.
A good news for commercial insurance: the shipping industry is increasing turning to AI to reduce rising number of cargo fires.
But perhaps AI adoption is more than the usual new technology adoption. In a new report, Contrary research explored the different way to integrate Vertical AI into industry, one takeaway: “The next generation of CEOs will look less like pure technologists and more like capital allocators. Their job is to treat AI not as a feature but as a labor class, and to deploy it with the same discipline that the most successful serial acquirers applied to capital.”
Startup - (a lot of) money is moving
Micro1, competitor to Scale AI whom founder was recetly pouched by Zuckerberg in one of the most expensive move of all time, raises funds at $500 millions valuation.
Scale AI sign a $100M-cap deal with the Department of
DefenseWar to deploy its AI on Top Secret networks.Groq (not to be mistaken for Grok, Musk’s AI company) valued at $6.9 billions, the company is trying to compete against NVIDIA in the chip race.
Still pursuing the future of online search, Perplexity raises $200m at $20B valuation.
Life Sciences, which uses AI to develop novel drugs and materials, raised $235 millions.
Replit, developing a tool to faciliate app making with AI, closes $250 million in Funding.
Black Forest Labs, an image generation AI company, signed $140 million deal with Meta.
Databricks, a platform for massive data analytics and AI development, confirms $100B valuation based on conservative future annual revenu prediction of $4B.
I could have shorten the list but I wanted to share the scale of the investment going on in a week of time. Considering this and the amount that are raised, one could, and perhaps, should ask themselves if this yet another Bubble or things are here to stay?
Ethic and society
In its annual trade report, the WTO head warns “AI could deepen inequalities rather than reducing them”.
Teen relationship and access to AI concern is growing, to face the problem, Italy has enacted a new AI law focusing on privacy and child access, while OpenAI launches Teen Safety features - ChatGPT will now guess how old you are base on your interaction and might prohib you access.
In an interesting move, Albania’s PM is considering the appointment of AI to his ministry.
Research
What are people using ChatGPT for? A new publication from OpenAI looks exactly into that. Top 3: Practical guidance (28.3% of the prompt); Writing (28.1%) and Seeking Information (21.3%).
Anthropic (creator of Claude) released it’s Economic Index Report highlights the rapid adoption of AI, with 40% of US employees using it at work, reflecting its utility and ease of use. It introduces the Anthropic AI Usage Index to analyze geographic patterns of AI adoption, revealing that higher-income countries show greater usage.
Can AI help us fulfil one of human long dream: understanding and talking with animals? Some researcher are trying to find out.
Deepmind (Google) and OpenAI get Gold Medal at the coding Olympics - a first.
ThinkingMachine - a startup you should keep an eye on, started by Mira Murati former OpenAI CTO - is looking into defeating nondetermism in LLM inference. Nondeterminism in large language model (LLM) inference arises from floating-point non-associativity and batch size variations, complicating reproducibility. To achieve deterministic results, it's crucial to implement batch-invariant kernels for operations like RMSNorm, matrix multiplication, and attention. By addressing these issues, researchers can ensure consistent outputs across multiple requests, enhancing the reliability of LLM applications.
Ai requires data but also chip, always more powerfull and with smaller component. Johns Hopkins researchers have developed a new process using metal-organic materials to create microchips with circuits smaller than 10 nanometers, which are invisible to the naked eye.
Are AI glasses the needed interface for better AI?
While the rest for AI might look like a chatbot race for now, many are dubious that this model of chat through an app or webpage is actually the best interface. Since the start of the AI race, we’ve seen company like Humane or Rabbit trying their best at developing new hardware to allow easier human-machine interaction. More recently, OpenAI announced a partnership with Johny Ive, one of the designer whom have shaped Apple product at the time of Steve Jobs. Of course, there is the futuristic neuralsystem like Musk’s backed company Neuralink: plug your brain in, the rest will be automatic. But what before that? In a more realistic future it seems that many big tech are turning toward AR/VR goggles. Apple has released last year its AR headset and since send a lot of people are gossiping that their new iOs update, always more transparent and round, is actually ment to get consumer to adapt to a more augmented reality like word. Where you see (transparent) and icon are bubble floating in the air. But for now, one company is taking the lead: Meta. With the release of it’s new AR glasses in partnership with Rayban. Something to follow for certain, we might be working with gogle in a couple years.
A tool you should know
This week I would like to share my experience with NotebookLM from Google. I cannot recommend a better tool to learn about a new concept, product, market etc. Gather the material you want to study (article, website, youtube video etc.) drop it into the tool, then you can easily get a report, video or podcast on the topic of the materials. I’ve tried it by giving it a dozen of article and a PhD thesis on EV vehicule then asked plenty of questions and ask it to generate a podcast for me to learn about the articles. I got an amazing 15 minutes podcast that covers all the important topics in the documents. Quite fantastic!