In the fast-paced, Tech News today, technology has ceased to be a mere tool; it has become the very essence of survival.
From the very first wake in the morning to tilting eyes-close to a dream, technology is threaded into every aspect of life.
Artificial Intelligence (AI) has become the forerunner of this technological revolution.
We see constant innovations in software and hardware, shaping our daily interactions and future possibilities.”
From the very concept of science fiction, AI is now practices which revolutionizes industries, disrupts markets, and redefines the essence of human capability-with great speed.
In this blog post, we aim to treat the state of affairs regarding AI now and its expected effects on society; however, it will also lead to a discussion on the challenges and opportunities that lie ahead.
Apple is set to launch port-less iPhone 17 Air
Apple is facing a federal lawsuit over alleged false advertising of its Apple Intelligence features.
The class-action suit, filed in San Jose, claims Apple misled customers by promising Al features that have not been delivered.
Plaintiffs argue that Apple created expectations that “transformative” Al capabilities would be available at launch, but the actual features were either limited or absent.
The lawsuit adds to Apple’s challenges in rolling out Apple Intelligence, with reports suggesting CEO Tim Cook has lost confidence in Al chief John Giannandrea’s execution.
Apple is set to launch the iPhone 17 Air this fall, featuring a thinner design like the MacBook Air.
Engineers faced significant challenges in reducing thickness while maintaining battery life, reportedly requiring a “Herculean effort.”
Apple considered making the iPhone 17 Air completely port-free, relying solely on wireless charging and cloud-based data syncing.
The company ultimately decided against this move, partly due to European regulations mandating USB-C compatibility.
Google to shift Android development in-house
Google will shift all Android development to its internal branch, moving away from the public Android Open-Source Project (AOSP) as a primary development area after 16 years.
AOSP, licensed under Apache 2.0, allows free use, modification, and distribution of Android,
But Google also maintains a private internal branch for partners with Google Mobile Services (GMS) licenses.
Maintaining both branches has led to inefficiencies, with features and APIs in the internal branch often outpacing the public AOSP version.
Google cited frequent merge conflicts, such as mismatched accessibility lists and manual syncing issues, as reasons for the transition.
Researchers develop Al that detect cancer with 99.26% accuracy
Researchers have developed an Al model named ECgMLP that detects endometrial cancer with an accuracy of 99.26%, significantly surpassing existing methods which range from 78.91% to 80.93%.
The model analyzes microscopic images of tissue, enhancing image quality and identifying critical areas to detect early signs of cancer.
The development team includes scientists from Charles Darwin University, Daffodil International University, the University of Calgary, and Australian Catholic University.
The ECgMLP model utilizes advanced techniques such as ablation studies and self-attention mechanisms, making it computationally efficient and robust across various histopathology datasets.
Al is starting to kill Google’s ‘ten blue links’
Al-powered search is rapidly gaining traction, replacing traditional search engine methods like Google’s ten blue links.
Adobe’s research, analyzing over 1 trillion visits to U.S. retail sites, found that Al search referral surged 1,300% during the 2024 holiday season, with a 1,950% jump on Cyber Monday.
Users referred via Al search engage more:
they stay on sites 8% longer, browse 12% more, and are 23% less likely to bounce compared to traditional search referral.
Adobe’s survey of 5,000 consumers found that 39% use Al search for shopping, 55% for research & 47% for product recommendation.
While Perplexity and Google incorporate ads, OpenAl avoids them for now, seeing ads-plus-Al as “uniquely unsettling.”
Pruna open source its Al optimization framework
Pruna Al, a European startup focused on Al model compression, is open sourcing its optimization framework.
The framework integrates caching, pruning, quantization, and distillation to improve Al efficiency, standardizing saving, loading, and evaluating compressed models.
Unlike existing tools that focus on single methods, Pruna Al aggregates multiple efficiency technique into easy-to-use system
A compression agent feature will soon allow users to set performance goals (e.g., “increase speed without losing more than 2% accuracy”), automating the optimization a process.
Pruna Al charges on an hourly basis, like cloud GPU rentals, and claims its compression significantly reduces inference costs.
DeepSeek release their new Al model
Chinese Al startup DeepSeek launched an upgraded version of its large language model, DeepSeek-V3-0324.
The model shows enhanced reasoning and coding abilities compared to previous versions.
DeepSeek-V3-0324 is accessible through the Hugging Face platform.
DeepSeek has been expanding quickly, positioning itself as a cost-effective alternative to Western Al models.
The company launched DeepSeek-V3 in December and DeepSeek R1 in January.
China Construction Bank has already deployed a financial model based on DeepSeek R1 earlier this year.
Researchers discover new Al scaling method
Google and UC Berkeley researchers propose “inference-time search” as a potential new Al scaling law.
The technique involves generating multiple answers in parallel and selecting the best one through self-verification.
The method reportedly improves older models’ performance.
Researchers claim that Google’s Gemini 1.5 Pro, using this approach, outperforms OpenAl’s ol-preview on math and science benchmarks.
Experts remain skeptical. They argue that inference-time search is only effective when the best answer is easy to verify and may not work for more general language-based tasks.
Alibaba chairman warns of a ‘bubble’ in Al data center
Alibaba Chairman Joe Tsai warned of a potential bubble in data center construction, arguing that Al infrastructure investments may be outpacing actual demand.
He pointed out that many projects are being built without clear customers, raising concerns about speculative spending.
Big tech firms like Microsoft, Amazon, Alphabet, and Meta have pledged massive investments in Al infrastructure, with commitments exceeding $100 billion each.
Wall Street analysts have started questioning whether Al spending is excessive, especially after DeepSeek demonstrated that competitive Al models can be built at much lower costs.