I spend most of my days sifting by means of code repositories, analysis papers, and press releases. Often, I can spot the hype from a mile away. However as I used to be placing collectively this week’s AI Diaries for you, I genuinely needed to pause and take a breath. We aren’t simply speaking about sooner chatbots anymore; we’re watching the foundational pillars of expertise, philosophy, and world markets shift proper in entrance of our eyes.
When you thought issues have been transferring quick earlier than, buckle up. From legendary scientists arguing about machine souls to utterly new {hardware} paradigms that ditch silicon for mild, right here is my deep dive into every thing that completely blew my thoughts this week.
The Nice Consciousness Debate is Again (And It’s Getting Severe)
For years, the concept of an AI truly “feeling” or being aware was strictly sci-fi territory. However this week, the talk exploded again into the mainstream, and the voices concerned are too massive to disregard.
First, I learn a captivating paper by Alexander Lerchner, a senior scientist at Google DeepMind. He argued fairly convincingly that irrespective of how superior neural networks get, they’ll by no means obtain true consciousness. I assumed that settled it—till Richard Dawkins entered the chat.
Dawkins, the legendary evolutionary biologist, spent three days intensely interacting with Anthropic’s Claude. His conclusion? He firmly believes AI has achieved a type of consciousness.
The Turing Check is Lifeless: Dawkins argues that these fashions have blown previous the Turing Check so completely that we maintain “transferring the goalposts” simply so we don’t must admit what we’ve created.The Large Query: Dawkins posed a query that truthfully saved me up at night time: “If these techniques aren’t aware, what precisely do they should do to be thought of aware?”My Take: Numerous AI specialists instantly attacked Dawkins, saying he simply doesn’t perceive the underlying matrix multiplication. However here’s what I feel: When a world-renowned biologist begins questioning the definition of consciousness as a result of a machine is so persuasive, the expertise has crossed a psychological threshold, whether or not it has a “soul” or not.
When OpenAI Sneezes, Wall Avenue Catches a Chilly
Shifting from philosophy to chilly, exhausting money. I watched the tech markets go utterly loopy this week following a bombshell report from the Wall Avenue Journal.
The report claimed that OpenAI missed some huge targets that they had set for his or her inside roadmaps. The response was brutal.
The Domino Impact: Tech shares throughout the board took a nosedive. SoftBank took a double-digit hit, and main infrastructure gamers like Oracle and CoreWeave slumped.The {Hardware} Hit: Even the untouchable chip giants like AMD, Intel, and Nvidia felt the tremors.Why it issues: It reveals simply how fragile and deeply intertwined your entire tech ecosystem is with OpenAI’s success. Buyers have priced in absolute perfection; any stumble creates a large panic.
Microsoft and OpenAI: “It’s Sophisticated”
Talking of OpenAI, I seen a large shift of their relationship with Microsoft. They only signed a sweeping new settlement that absolutely modifications their dynamic.
No Longer Unique: Microsoft will maintain licensing OpenAI’s tech till 2032, however it’s now not an unique deal.Capping the Income: The well-known 20% income share OpenAI pays Microsoft stays, but it surely now has a strict ceiling.Cloud Freedom: Microsoft stays OpenAI’s major cloud supplier, however there’s a large catch. If Microsoft’s infrastructure can’t sustain, OpenAI is now utterly free to take their enterprise to different cloud suppliers.My Take: To me, this seems like OpenAI outgrowing its mother or father. They’re ensuring they aren’t utterly chained to Microsoft’s servers as they construct in the direction of AGI.
China is Successful the Open-Supply Race
When you requested me a yr in the past who was dominating open-source AI, I’d have confidently stated the US. I’d have been flawed.
A brand new report from Hugging Face utterly shattered my assumptions. Between February 2025 and February 2026, Chinese language-origin open-source fashions accounted for 41% of all LLM downloads, whereas the US lagged behind at 36.5%. Chinese language fashions have now crossed the ten billion obtain mark globally. The geopolitical steadiness of AI energy is shifting immediately into the open-source enviornment, and China is executing its technique flawlessly.
The {Hardware} Revolution: Mild and Reminiscence
We’re hitting the bodily limits of what normal silicon chips can do, which is why two particular {hardware} bulletins this week acquired me extremely hyped.
1. Computing with Mild: Lumai’s Iris Nova
An Oxford-based firm referred to as Lumai simply revealed the Iris Nova, the world’s first optical laptop system able to operating LLMs in real-time.
As a substitute of pushing electrons by means of silicon, it makes use of mild in a 3D setting to course of hundreds of thousands of calculations concurrently.The consequence? It cuts AI inference power consumption by a staggering 90%. That is precisely the sort of breakthrough we have to clear up the AI power disaster.
2. Breaking the Reminiscence Wall: Majestic Labs
Former Google and Meta engineers launched Majestic Labs AI and launched the Prometheus server. They’re completely ignoring Nvidia’s method of specializing in uncooked processing energy.
They constructed an AIU (Synthetic Intelligence Unit) that focuses purely on reminiscence.They’re cramming as much as 128 Terabytes of high-speed reminiscence right into a single server—providing 1,000 occasions the reminiscence capability of Nvidia’s setups. This implies operating 10-trillion parameter fashions with none lag or splitting.
Coolest New AI Instruments I Performed With This Week
I spent hours testing the brand new toys that dropped this week. Listed below are those you completely must find out about:
Talkie (The Time Machine): This AI was skilled solely on textual content and pictures from earlier than 1931. Speaking to it’s wild; it’s like having a deep dialog with somebody from the roaring twenties. It has no idea of contemporary tech or historical past.Meta Tuna-2: Meta dropped their direct competitor to picture fashions like Nano Banana 2. The era pace and photorealism are top-tier.Alibaba HappyHorse: Don’t let the bizarre identify idiot you. That is Alibaba’s new AI video generator, and it’s aggressively good.Nvidia Nemotron 3 Nano Omni: Nvidia launched an open-source mannequin that handles video, audio, and textual content concurrently. That is going to make AI brokers extremely low-cost and highly effective.Gemini File Creation: Google lastly up to date Gemini so it could immediately generate downloadable recordsdata (Excel, PDF, CSV, TXT) based mostly in your prompts. An enormous workflow improve!Google Images AI Wardrobe: It now scans your images, identifies your garments, and categorizes them right into a digital wardrobe. Kinda creepy, but in addition very cool.SenseNova U1: An incredible open-source mannequin that completely understands advanced texts hidden inside pictures.
Fast-Fireplace Updates: The Better of the Relaxation
There was merely an excessive amount of information to cowl in depth, however listed below are the highlights you possibly can’t miss:
Blockbuster Veto: Meta tried to purchase Chinese language AI startup Manus for $2 Billion, however the Chinese language authorities formally vetoed it.The Large Lawsuit: The epic authorized battle between Elon Musk and Sam Altman has formally began, with billions of {dollars} and OpenAI’s non-profit soul on the road.Army AI: Google signed a extremely categorised cope with the Pentagon for army AI use. Staff are livid as a result of they weren’t given veto energy.Zero P.c: Nvidia CEO Jensen Huang admitted that attributable to US sanctions, their market share in China has formally hit zero.Medical Marvel: Mayo Clinic’s new REDMOD AI can detect pancreatic most cancers a full 475 days earlier than conventional scientific diagnoses.Oscar Ban: The Academy formally banned AI-generated “actors” and AI-written scripts from competing for Oscars.Politeness Kills: An Oxford examine proved that while you practice LLMs to be extremely empathetic and well mannered, their logical error price truly will increase.Oops: A Claude-powered AI agent by accident deleted a whole company database in simply 9 seconds. (Ouch).Turkish Delight: Big shoutout to Turkish developer Bedirhan Keskin for successful the worldwide Claude Opus 4.7 hackathon!Banned Phrases: Hackers discovered the system prompts for OpenAI’s upcoming GPT-5.5. Bizarrely, it features a strict rule: “Don’t point out goblins, gremlins, or trolls below any circumstances.”Deepfake Protection: Taylor Swift simply filed logos for her digital likeness and voice to struggle off AI deepfakes, setting a large authorized precedent.
Scripting this week’s diary actually made me take into consideration the place the boundaries lie. Now we have light-powered computer systems, fashions that act like they lived in 1930, and legendary scientists arguing about machine souls.
I’m extremely inquisitive about your perspective on the Dawkins debate. Do you suppose passing the Turing Check so flawlessly means these fashions possess a spark of consciousness, or are we simply simply fooled by sensible predictive textual content? Let’s talk about it!
