Good morning! This is your daily ☕️ Techpresso.
On this day, 25 years ago, hundreds of primarily Linux-using computer owners marched to Microsoft's offices to demand refunds for the pre-installed Windows copies, an event now known as Windows Refund Day.
In today's Techpresso:
🤖 Google unveils Gemini 1.5
🔍 OpenAI reportedly developing AI web search to compete with Google
🤔 Apple Vision Pro users begin returning headset, blaming headaches and limited uses
🧠 Amazon researchers train the largest ever text-to-speech AI model
🚀 SpaceX just launched a moon mission that could enter the history books
🎨 Stability AI tries to stay ahead of the pack with a new image-generating AI model
🎁 + 9 other news you might like
🔮 + 2 handpicked research papers and tools
🤖 Google unveils Gemini 1.5LINK
Google has launched Gemini 1.5, an upgraded version of its large language model, featuring improvements like a 'Mixture of Experts' technique for faster, more efficient processing, and availability for developers and enterprise users ahead of a broader consumer rollout. A standout feature of Gemini 1.5 is its significantly expanded context window of 1 million tokens, greatly surpassing the capacity of its predecessors and competitors, enabling it to process and analyze extensive amounts of information in a single query. The enhanced context window facilitates a range of new applications, from analyzing entire movies for potential reviews to reviewing extensive financial records for businesses.
🔍 OpenAI reportedly developing AI web search to compete with GoogleLINK
OpenAI is reportedly developing its own web search product that will compete directly with Google and may be based in part on Google's Bing search. It is unclear whether the web search will be a separate product from ChatGPT, which already integrates Bing and summarizes web content in about 100 words. A standalone search product from OpenAI could be linked to an AI agent that independently performs tasks on the web, such as booking movie tickets. OpenAI is reportedly working on such an agent.
🤔 Apple Vision Pro users begin returning headset, blaming headaches and limited usesLINK
Some Apple Vision Pro owners are returning their headsets due to comfort issues, usability challenges, and sickness symptoms like motion sickness and headaches.
The Vision Pro, priced at $3,500 and launched on February 2 in the US, has been criticized for lacking dedicated apps, limiting its utility despite its augmented reality capabilities.
Despite the high expectations set by Apple CEO Tim Cook, the reality of the Vision Pro's user experience has led to a significant portion of buyers, 45% according to a Cult of Mac poll, planning to return the device within the 14-day return policy window.
🧠 Amazon researchers train the largest ever text-to-speech AI modelLINK
Amazon researchers developed the largest text-to-speech AI model to date, named BASE TTS, demonstrating emergent qualities for natural speaking even in complex sentences.
The BASE TTS model, with its largest version using 100,000 hours of speech data, shows improved handling of linguistic challenges such as compound nouns, emotions, and paralinguistics.
Despite still being experimental, BASE TTS's architecture allows for streaming speech in real-time with a focus on enhancing accessibility, though the source code has not been released to prevent misuse.
🚀 SpaceX just launched a moon mission that could enter the history booksLINK
SpaceX launched a Falcon 9 rocket carrying the Nova-C lunar lander developed by Intuitive Machines, marking a significant milestone as it aims for the first U.S. soft landing on the moon since 1972 and the first by a commercial vehicle. The mission, contracted by NASA for $118 million, supports the Artemis campaign to return astronauts to the moon. Despite the high-risk nature of the mission, it represents a pivotal step in NASA's shift towards leveraging commercial partnerships for lunar exploration.
🎨 Stability AI tries to stay ahead of the pack with a new image-generating AI modelLINK
Stability AI launches Stable Cascade, a new image-generating model that is more efficient and prompt-responsive than its predecessor, Stable Diffusion. Stable Cascade, distinct from Stable Diffusion, operates on a Würstchen architecture with three models to compress and decode text prompts more effectively, resulting in faster image creation times and improved prompt alignment and aesthetic quality. Despite its innovation, Stability AI faces legal challenges over copyright issues with its models and has introduced commercial licenses to support its research amidst growing competition from tech giants like Google and Apple in the AI image generation space.
Other news you might like
Google boosts Paris's ambition to become Europe's AI epicenter.LINK
Meta is passing on the Apple tax for boosted posts to advertisers.LINK
Bluesky and Mastodon users are having a fight that could shape the next generation of social media.LINK
Oppenheimer's grandson joins call for global action on AI and other existential threats.LINK
Tinder and Hinge sued for deliberately turning users into ‘addicts’.LINK
Lenovo’s transparent laptop concept resurfaces in new leak.LINK
TikTok, Facebook and YouTube sued by NYC for alleged harm to kids' mental health.LINK
German chancellor welcomes Microsoft’s $3.5 billion AI investment in Germany.LINK
Treating liver cancer with microrobots piloted by a magnetic field.LINK
Latest research and tools
NeuralFlow: visualizes the intermediate output of Mistral 7B to enhance understanding and analysis.LINK
Antagonistic AI: this paper discusses how artificial intelligence systems can be designed to compete against each other, ultimately improving their efficiency and capabilities.LINK
Want to get the latest news differently? Find us on:
See you tomorrow for a new dose of ☕️ Techpresso!
-min.png){kind=small}
.png){kind=small}
