AI applied sciences, which have change into more and more built-in into our day by day lives, proceed to evolve at breakneck pace. Whereas we witness outstanding developments virtually on daily basis, vital milestones are being crossed one after the other. In our “AI Diaries” sequence, we proceed to document the evolution of this technological revolution that holds the potential to dramatically reshape life on Earth.
What Occurred within the World of AI This Week?
The occasions of the previous week point out that we’re on the verge of a serious shift within the synthetic intelligence panorama. On one hand, efforts to achieve AGI (Synthetic Common Intelligence) have accelerated; on the opposite, timelines have been set for the discharge of devoted AI {hardware}. Parallel to those strategic strikes, new AI instruments have entered our lives, as they do each week.
Listed below are the standout developments from this week on this planet of AI:
AGI Predictions: The Threshold is Close to
AGI—synthetic basic intelligence that possesses cognitive skills indistinguishable from or surpassing people—has been a frequent subject of dialogue recently. Judging by the successive statements made this week, trade leaders count on this threshold to be crossed within the very close to future.
Talking on the World Financial Discussion board in Davos this week, Elon Musk predicted that AI will surpass particular person human intelligence by the top of 2026, and exceed the collective intelligence of all humanity inside 5 years.
An identical sentiment got here from the Google camp. Shane Legg from Google DeepMind said that AGI is now on the horizon and predicted radical financial transformations in consequence. Whereas there appears to be a consensus on the timing, firms differ on tips on how to get there. Google DeepMind CEO Demis Hassabis famous this week that Giant Language Fashions (LLMs like ChatGPT) alone will not be sufficient to achieve AGI. In keeping with Hassabis, AI requires a real “world mannequin” to cross this threshold. DeepMind is taking vital steps to create this subsequent technology of AI, as mirrored in Legg’s feedback.
Nvidia Ushers in a New Period with PersonaPlex-7B-v1
Nvidia researchers have introduced PersonaPlex-7B-v1, a brand new speech mannequin focusing on pure, seamless, and real-time voice interactions. Developed with a full-duplex construction, this mannequin distinguishes itself from traditional voice assistant architectures by its skill to pay attention and communicate concurrently.
Conventional voice assistants use a multi-stage construction (ASR for speech-to-text, an LLM for response, and TTS for audio technology). This method causes latency at each stage and fails to help elementary human speech options like interruption or pure back-and-forth stream. PersonaPlex eliminates this multi-layer chain, performing each understanding and speech technology concurrently inside a single Transformer mannequin. It really works on a steady audio stream encoded by a neural codec, making it perfect for eventualities requiring pure dialogue stream and low latency.
AI {Hardware} is Coming: Apple and OpenAI Set Their Calendars
The AI race, which has to this point progressed by software program and voice assistants, seems set to leap to {hardware} by the top of 2026. OpenAI, having recruited Apple’s legendary designer Jony Ive, introduced that it’s going to unveil its first AI {hardware} led by Ive later this yr.
Concurrently, reviews have surfaced that Apple has begun creating an AI-powered wearable system. Like OpenAI, Apple is getting ready to launch AI-focused {hardware}. Nonetheless, Apple appears more likely to comply with the trade from behind on this regard, with a focused launch date of 2027.
South Korea Enacts the World’s First Complete AI Regulation
South Korea has change into one of many first international locations to implement a complete authorized framework for AI, with the Framework Act on Synthetic Intelligence coming into pressure on January 22. The regulation goals to make clear the place and the way AI is used and assigns particular tasks.
The brand new legislation focuses on “high-impact” areas equivalent to nuclear security, water manufacturing, transport infrastructure, healthcare, and credit score scoring. In these areas, totally autonomous AI programs are prohibited, and human oversight is obligatory. The legislation additionally enforces strict transparency, requiring firms to tell customers when they’re interacting with generative AI or high-impact programs.
xAI Launches the Gigawatt-Scale Colossus 2
Elon Musk’s AI firm, xAI, has activated Colossus 2, the world’s first AI coaching cluster working at a gigawatt scale. Like its predecessor, Colossus 2 will probably be used to coach Grok. Whereas Colossus 1 took 122 days to change into totally operational, the brand new system surpassed the 1 gigawatt (GW) threshold the second it went on-line. Musk said that the system will probably be upgraded to 1.5 GW in April, with an final objective of roughly 2 GW complete capability.
Apple Prepares a Huge Overhaul for Siri
Apple is getting ready to make a big transfer within the AI area, the place it has lengthy confronted criticism. In keeping with Mark Gurman from Bloomberg, the corporate will remodel Siri from a traditional voice assistant into an AI chatbot much like ChatGPT.
Reportedly, Apple will shelve the present Siri interface fully. As an alternative, it is going to provide an expertise the place customers can work together by way of each textual content and voice, akin to trendy AI chatbots. The brand new Siri is anticipated to launch on iPhone, iPad, and Mac gadgets earlier than the top of this yr.
OpenAI Quietly Trains Robots
OpenAI is making silent progress in robotics. In keeping with Enterprise Insider, the corporate established a robotics lab in San Francisco in February 2025 and has since quadrupled the unit’s measurement. Roughly 100 knowledge collectors and not less than a dozen robotics engineers are producing coaching knowledge by remotely controlling robotic arms.
The work focuses on low-cost robotic arms reasonably than full humanoids. Workers use a 3D-printed management system known as “GELLO” to map human hand actions on to the robotic arms. These robots are being educated on on a regular basis duties like placing bread in a toaster or folding laundry.
New AI Instruments Launched This Week
FlowAct-R1 (ByteDance): Permits customers to livestream utilizing digital avatars. It stands out with its response pace and avatars that change their perspective based mostly on the dialog content material.Waypoint-1: Creates interactive worlds that may be edited in real-time, reflecting person instructions into video virtually immediately.OmniTransfer: Means that you can simply switch digicam actions, actions, or results from one video to a different.Qwen3-TTS (Alibaba): An open-source mannequin delivering extremely profitable text-to-speech outcomes, together with voice cloning capabilities.VibeVoice-ASR (Microsoft): A brand new instrument able to deciphering voice recordings and producing written transcripts.
AI Shorts: Information in Transient
ChatGPT has activated age estimation options to guard younger customers, mechanically making use of content material filters to minors.Elon Musk introduced that Tesla’s Dojo3 supercomputer challenge will restart because the AI5 chip design nears completion.OpenAI is engaged on commerce options, together with buying carts and vendor add pages, aiming to show ChatGPT into an e-commerce platform.ChatGPT Atlas, OpenAI’s AI-powered browser, acquired a serious replace together with computerized switching between search instruments and tab teams.Non-game cell app spending surpassed recreation revenues for the primary time in 2025, pushed by Generative AI apps. AI in-app buy income tripled to over $5 billion.OpenAI has turned its path to the Center East for a brand new funding spherical in search of roughly $50 billion.
