Huawei Sound In October last year, Huawei released the first Devale full-frequency tuned smart speaker-Huawei Sound at the Mate 40 conference. The smaller size, more cost-effective price and the blessing of Divale's core technology have…
Recent News
Apple plans to launch a foldable iPhone in 2023 with an 8-inch flexible OLED screen
Analyst Ming-Chi Kuo said in a report to investors seen by MacRumors today that Apple is working hard to launch an 8-inch foldable iPhone with a QHD flexible OLED screen in 2023. Guo predicts that…
Stepping down the technical threshold, the whole house intelligently opens the new track
Xiaodu's whole house intelligent ecological products cover all scenes of life, and Xiaodu's assistant intelligent interactive system empowers the smart upgrade of the home improvement industry and shares AI capabilities. Simply put, Xiaodu whole house…
Report: MediaTek occupies a dominant position in the global smart speaker market
According to the latest research report released by Strategy Analytics Smart Speakers and Screen Services, of the 151 million smart speakers and smart screens shipped worldwide in 2020, nearly 50% are using MediaTek's application processors.…
Small image, big picture: AI revolutionizes microscope technology
Recently, researchers have proposed that the introduction of deep learning technology into cell imaging and analysis can transform chaotic biological problems into solvable calculations. Carpenter's laboratory is dedicated to using software to accelerate drug discovery,…
Closed, lack of code review, the "old" engineer who just graduated with a master's degree exposes the cruel status of ML
To be honest, I myself don't know much about machine learning, it's purely driven by curiosity. I got my undergraduate degree from a little-known university, and my master's school is ranked in the top 20…
From "geometric deep learning" to see the unification of deep learning
Don't be surprised, don't doubt. A 23-year-old guy from the University of Erlangen gave the answer more than a hundred years ago. He created the "Erlangen Project" solely by himself, thus making a groundbreaking work…
Only 2 pictures are needed, and AI can generate a complete movement process
Specifically, this work is to stylize the video based on key frames. Different from the previous method, this style transfer is carried out in a random order, and there is no need to wait for…
Qualcomm AI Research Institute proposes to map deep neural networks to quantum computers through quantum field theory
Scholars from Qualcomm AI Research Institute proposed to directly map deep neural networks to optical quantum computers through quantum field theory. This research discovery paved the way for the development of new quantum neural network…
Personalized automatic speech recognition, Apple uses federated learning to create a new system for evaluation and adjustment
In recent years, many federal learning publications have influenced researchers to expand the system to better support federal learning. The study restricts the participation of selected equipment, and this design choice has the potential to…
Ye Qing back! Super Mario level generated by GAN, which can never be cleared
However, if you think this is a masterpiece of "Made in Mario", then you guessed wrong~ This is made by AI using GAN! The use of GAN to generate red and white game screens is…
Artificial neural network changes pulse neural network every second, new technology is expected to open a new era of edge AI computing
Part of the reason is that the power consumption of the giant SNN of the human brain is only 20 W. The output of a pulsed neuron is a series of shaped pulses. Therefore, their…
Create a "Google Brain"? Samy Bengio joins Apple: will lead the new AI research department
At present, Apple has not yet commented on Bengio's position in the company. Samy Bengio, who had just left Google on April 28, made his choice due to the involvement of Timnit Gebru's "expulsion" incident.…
Microsoft Opens Source Code of AI Security Risk Assessment Tool Counterfit
Recently, Microsoft has open-sourced Counterfit, an AI security risk assessment tool designed to help developers test the security of AI and machine learning systems. The company stated that Counterfit can enable organizations to conduct assessments…
MLP regression, no need for convolution, self-attention, pure multi-layer perceptron visual architecture is comparable to CNN, ViT
Technically speaking, the ViT model continues the long-standing trend of removing hand-built features and inductive bias from the model, and further relies on learning based on raw data. These two types of layers are executed…