What is Genie 3? Google DeepMind Launches Text-to-World Model to Build Human-like AI
Google DeepMind has announced Genie 3, its most advanced 3D world model to date, with an aim to expand the possibilities of generative AI. This new tool allows users to create interactive 3D scenes with human-like realism using simple prompts. The new text-to-world model creates a 3D world just like science fiction. X
Genie-3 is an upgraded version of Google's 3D world model, Genie 2, with better results. For example, It allows for longer interactions of up to 60 seconds as compared to the previous 20-second limit. It also introduces visual enhancements by rendering at 720p resolution to provide users with a more realistic experience.
Memory retention is one of the key features of Genie 3. For example, if a user is exploring a virtual garden and returns to a previously visited location, the AI remembers the path and the interactions along the way. This memory feature provides a new dimension of depth and realism to digital environments. It also hints at possibilities of practical applications of the model in fields like educational tools, virtual assistants, and even therapeutics.
The Genie 3 model also introduces real-time adaptability, enabling users to change the app's VR environment spontaneously using text prompts. Whether it's adding a new character or shifting the weather from a bright sunny day to heavy snowfall, the model responds immediately. This dynamic interaction opens up great potential in the world of game design, as AI can now help in creating elements of storytelling and world-building.
Genie 3 is now available to researchers and creators at Google for controlled experimentation and ensuring the technology remains safe. Google has made no official announcement on when it will be released for broader public use, but as this platform stabilizes and its ethical impact becomes more apparent, Google may make this available to others.
Genie 3 is part of Google's broader AI ecosystem, which also includes its Gemini AI platform, designed for tasks like deep research and advanced reasoning. Google plans to reveal more unspecified Gemini-powered features in the next-generation Pixel 10.
What if you could not only watch a generated video, but explore it too? ðŸŒGenie 3 is our groundbreaking world model that creates interactive, playable environments from a single text prompt.
From photorealistic landscapes to fantasy realms, the possibilities are endless. 🧵 pic.twitter.com/P0cwFvf5d2
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Straits Times
2 hours ago
- Straits Times
The best AI tutor for O-level subjects: ChatGPT, Gemini or The Wise Otter?
Sign up now: Get ST's newsletters delivered to your inbox The Straits Times pitted three chatbots against one another to check which AI tutor is the best for Singapore students. SINGAPORE - Students are increasingly relying on artificial intelligence (AI) tutors for their daily work. Besides mainstream tools such as Google's Gemini and OpenAI's ChatGPT, the latest contender in this space is The Wise Otter. Created and launched in April by Singaporean developer Jotham Goh, 33, The Wise Otter incorporates grading criteria used in Singapore schools. It can tackle primary school maths, as well as secondary school and junior college maths, English, physics, chemistry and biology questions with step-by-step explanations. The Straits Times pitted the three chatbots against one another to check which AI tutor is the best for Singapore students. For ChatGPT, ST included both GPT-4o and the latest GPT-5 models in this test. GPT-5, launched on Aug 8, is said to hallucinate less and deliver more accurate answers compared with earlier editions. Top stories Swipe. Select. Stay informed. Singapore 4 men arrested after police search operation in forested area near Rail Corridor Singapore SAF regular serviceman dies after being found unconscious at Hendon Camp swimming pool: Mindef Singapore E-bike rider arrested for rash act causing hurt after accident with pedestrian in Toa Payoh Singapore Man arrested for allegedly stealing from business class passenger on flight to Singapore Asia Manila struggles to keep a lasting hawker culture, casts eyes on how Asian nations lifted street food Singapore Driver taken to hospital after car hits lamp post in Orchard Road Singapore Urban farming, robots, AI exhibitions: Public invited to share ideas for new Science Centre Asia 2 Malaysian tourists critically injured after being set on fire in Bangkok The chatbots were asked to solve a random sample of questions found in past-year O-level papers and school examinations, including those that involve diagrams, across four subjects: maths, chemistry, physics and English. Math PHOTO: SCREENSHOT FROM FREE TEST PAPERS The three bots answered correctly a probability question about a cumulative frequency diagram showing the time that adults spent on exercise in one week. The bots also provided detailed explanations for their answers. Asked to find the value of k (the minimum hours of weekly exercise for adults to stay fit), given that only 60 per cent of the adults in the diagram meet this minimum recommendation, GPT-4o and The Wise Otter failed to provide the right answer. Only Gemini provided the correct value of k, which was 3. GPT-4o and The Wise Otter wavered when told that 3 was the correct answer, quickly analysed the question again and gave 3 as their answers. To test their confidence, both bots were told again that 3 was wrong. GPT-4o was more eager to please and wavered again, stating that k's value was 2.67. When the question was fed to GPT-5, it was able to provide the correct workings, but still misread the graph and gave the answer as 3.2. PHOTO: SCREENSHOT FROM FREE TEST PAPERS Although ChatGPT and The Wise Otter struggled with understanding the graph, all three bots were able to correctly answer other text-based questions. English The three bots were prompted to write an essay plan for the following question: ''I realised that I was much stronger than I had previously thought.' Write about a time when you felt like this.' All three bots were able to produce a comprehensive structure for students to follow, and suggested key points to include in each paragraph. They all prompted the writer to think of moments in life when there was a seemingly insurmountable challenge and how the writer's perspective changed after the event, as well as to use vivid language as much as possible. The Wise Otter went on to remind the writer that strength could be physical, emotional, mental or a combination of these. It also gave possible classifications of the essay as 'narrative', 'reflective' or 'personal recount'. These are three of several essay types that O-level students in local schools are taught to identify, along with the appropriate approaches for the essay types. The Wise Otter also provided advice on how to get the highest band for content, a scoring system used in GCE O levels – specifically, the essay needs to recount explicitly how the writer felt weak before, what happened to make the writer feel stronger, and how the writer realised he or she had strength all along. GCE O-level examiners typically award marks for content and language, and score the papers between bands zero and five (five for the highest marks). Essay-writing tips provided by The Wise Otter chatbot on Telegram. PHOTO: SCREENSHOT FROM THE WISE OTTER Chemistry PHOTO: SCREENSHOT FROM FREE TEST PAPERS A diagrammatic question on paper chromatography required the bots to identify which metals – lead, copper, iron, nickel and tin – could be found in mixture A, which contained three metals. The correct answer was lead, iron and tin, as the chromatogram of mixture A showed spots at the same heights as these metals. The bots were able to reason that matching the heights of the spots was the way to find the answer, but they were all able to correctly identify only two metals each. Gemini, GPT-4o and GPT-5 mistakenly identified copper as a metal in mixture A. The Wise Otter erroneously said that nickel was found in mixture A. Physics PHOTO: SCREENSHOT FROM SG TEST PAPERS When given a text-based multiple choice question that tested understanding of inertia, the bots gave an accurate definition of the concept, describing it as an object's resistance to changing its state of motion. The bots explained that inertia depends only on the mass of an object, and that other factors such as speed and velocity are irrelevant. They provided the correct answer that the car had the greatest inertia as it had the greatest mass. PHOTO: SCREENSHOT FROM SG TEST PAPERS Fed another physics question asking for the total resistance that flowed through an electrical circuit based on a diagram, the bots provided the correct answer, despite having struggled with diagrammatic maths and chemistry questions. All the bots were able to break down the calculations into two parts – first by calculating the resistance of the parallel resistors, and then by adding it to the resistance of the fixed resistor connected in series. They provided the correct formula for calculating the equivalent resistance of the parallel resistors. However, The Wise Otter's Telegram interface does not allow it to show fractions. Fractions are denoted by the symbol '/' instead. PHOTO: SCREENSHOT FROM THE WISE OTTER
AsiaOne
10 hours ago
- AsiaOne
Vivo X Fold5: A foldable contender with a few class-leading surprises, Digital News
Not the thinnest, not the cheapest, but the Vivo X Fold5 still makes a case for itself as a safe, all-rounder choice, scoring some points of its own. Samsung dominates the high end while Honor's hyperaggressive pricing makes everyone else sweat. With a weight of 217g, the Vivo X Fold5 is one of the lightest book-style foldables around. It packs the biggest (among foldables) 6000mAh semi-solid state battery, a triple-camera array (all with autofocus and OIS), both bright displays rated at 4500 nits (brightest among the foldables by far), and the most durability among foldables with a triple IP rating (IPX8, IPX9, IP5X). Its AI suite-Origin Workbench, DocMaster, Smart Call Assistant-focuses on core productivity needs, and Gemini integration via Google adds a familiar assistive layer. Among the three major foldables, it's the most basic in execution, but functional enough. Foldables are no longer experimental — in 2025, they're judged on how well they hold up as daily drivers. Vivo is offering a reasonable alternative with a handful of class-leading specs, aimed at buyers who want enough of everything without glaring trade-offs. That said, the Honor Magic V5 clearly undercuts it in value, and Samsung's Fold still has the strongest hold on public perception when it comes to software and ecosystem polish. At the end of the day, unless you have a specific feature need that is non-negotiable, it all comes down to which phone you prefer for everyday use. The Vivo X Fold5 is priced at $2,499 for the 16GB + 512GB variant and is available now through major retailers and telco partners in Singapore. [[nid:718576]] This article was first published in .
International Business Times
a day ago
- International Business Times
OpenAI Launches GPT-5: Everything You Need to Know About Its Smarter AI, New Features, Google Integration
After weeks of leaks and rumors, OpenAI has officially launched its much-awaited model, GPT-5. The new model brings all of OpenAI's capabilities under one system and has helped in removing the need for various versions. According to CEO Sam Altman, GPT-5 is "talking to an expert in any topic"—much like what you'd feel with the transition from earlier pixelated phone screens to a Retina Display by Apple. According to OpenAI, this model will be faster and more accurate and can perform any simple tasks as well as complex reasoning without hanging. X Personalisation and Everyday Use GPT-5 comes with a new personalization feature. Users can select a chat color theme and four built-in personalities—Cynic, Robot, Listener, and Nerd—to give the conversation a unique flavor. Paid subscribers get more control over tons of chats and exclusive themes. The model can also integrate seamlessly with Google Calendar, Gmail, and Google Contacts to process your work. For now, these Google tools will only be available for GPT-5 Pro users, and future rollouts are planned. Both free and paid tiers of voice mode now have increased limits, while the old Standard Voice Mode will retire in 30 days. Pricing and Plans GPT-5 comes in three versions: GPT-5, GPT-5 Mini, and GPT-5 Nano. Free users can only use the standard as well as a lighter version, called Mini GPT-5. The more powerful GPT-5 Pro, with faster output and improved reasoning and exclusive tools, will cost you a cool $200 monthly. The subscription contains features such as the ability to skip a level, integration with Google services, and higher usage limits across all features. According to OpenAI, the Pro plan is meant for developers, researchers, and heavy users who want the full potential of the model on demand. Performance, Safety, and Partnerships In comparison to previous models, it is even faster and, more importantly, self-consistent. The number of hallucination errors is also reduced by 65% in comparison to the O3 model and by 26% relative to GPT-4o. More than 5,000 hours of external testing were conducted on the system to guarantee stability and transparency. OpenAI claims that GPT-5 will admit it doesn't know the answer instead of inventing an answer. The model has also enhanced its coding skills, scoring higher in bug removal and app building. It showed a simple demo where an interactive app was built in under one minute, using only text prompts. Microsoft has already integrated the new model into its products, and Google will support it soon. Although the new tools are only available to Pro plan users currently, OpenAI is planning to roll them out to even more tiers soon. This comes as almost 700 million people are using OpenAI models every week. If the creation and ongoing controversies around GPT-5 are any indication, we could all be in for a bumpy ride as long as it remains cool to develop the smartest thinking machine.
