A woman with severe paralysis has regained the ability to speak using an avatar, thanks to brain-computer-interface (BCI) technology. The breakthrough offers hope to individuals who have lost communication due to strokes and conditions like ALS. Previous speech synthesizers relying on eye tracking or facial movements proved frustrating and slow, preventing natural conversation. The new technology employs small electrodes implanted on the brain’s surface to detect electrical activity in the speech- and facial movement-controlling areas.
These signals are then translated into speech and facial expressions for an avatar. The aim is to restore a natural form of communication. The test patient, Ann, has been paralyzed since a brainstem stroke 18 years ago and communicates by selecting letters at 14 words per minute. The AI algorithm was trained to translate her brain signals into intelligible sentences, allowing her to control an avatar with a voice that sounds like hers.
This piece discusses the potential consequences of using AI instead of human writers, journalists, and creatives. The author argues that while AI may not be capable of completely replacing human creativity, the genuine concern lies in the motivations behind using AI. They question whether those in power care about the quality of AI-generated content, especially if their goal is to avoid paying real writers and authors. Additionally, the author raises concerns about scammers using AI to create fake books and sell them to unsuspecting consumers and the potential difficulties in weeding out AI-generated submissions in the publishing and media industries.
The author suggests that rather than dismissing the idea of AI replacing humans based on its current capabilities, it is essential to ask questions about who is investing in AI, how they plan to use it, and their long-term goals. Ultimately, the author warns that if AI is implemented without considering its quality or impact, there may be significant losses in the creative industry.
Princeton University professor Arvind Narayanan and his Ph.D. student Sayash Kapoor have written a book exploring generative AI’s risks and potential harm. The duo, who have been studying AI since 2019, believe that the progress and usefulness of generative AI have been “spiraling out of control” but could be improved if more people used their collective power to demand change. Narayanan states there is significant hype around generative AI and a need for more responsible tech.
He calls for usage transparency around generative AI and believes there needs to be more technologists in government to ensure better enforcement of existing laws. The professors view some of the hype around generative AI as misleading and state that numbers and claims about the technology mean virtually nothing without proper evaluation and testing. Despite their concerns, they are optimistic that the issues associated with AI can be dealt with effectively with increased concern and awareness.
South Korean internet giant Naver Corp. has introduced CLOVA X, a Korean language chatbot similar to ChatGPT, capable of engaging in conversations and summarizing text. Naver’s CLOVA X leverages the company’s Korean language and culture expertise to provide a tailored experience for Korean-language users. The chatbot and the HyperCLOVA X AI model that powers it are designed to support Naver’s various AI services.
Additionally, Naver unveiled CUE, a generative AI search service powered by HyperCLOVA X. The company has invested heavily in AI, spending $754 million over five years, and is expanding its data center infrastructure. Naver plans to develop localized AI applications for countries in the Middle East and target non-English speaking regions such as Japan and Southeast Asia. The beta testing phase for CLOVA X began on August 24, with the CUE search tool set to launch in September. Naver’s CEO Choi Soo-yeon expressed the company’s readiness to face the new opportunities presented by generative AI.
Researchers have discovered methods to overcome the trade-off between accuracy and beauty in digital images. Traditionally, enhancing blurry or low-quality images has been limited because blowing up an image too much results in visible pixelation and a loss of detail. However, researchers have started incorporating AI algorithms, specifically generative adversarial networks (GANs), into image-enhancing tools to produce more detailed pictures. While GANs create visually appealing images, they often introduce inaccuracies and distortions.
Researchers have plotted the performance of various image-enhancement algorithms on a graph of distortion versus perceptual quality and found a trade-off between the two. Some algorithms produce high visual quality but have high distortion, while others are accurate but do not look visually appealing. To overcome this trade-off, researchers have proposed incorporating multiple interpretations of the original image or combining data from multiple images. The challenge remains to balance enhanced image quality and accuracy in various applications, from entertainment to research and medicine.
Researchers at MIT have developed a new AI technique that allows robots to manipulate objects using their entire bodies rather than just their fingertips. The technique, known as “smoothing,” summarises complex movements into more minor decisions, enabling an algorithm to identify an effective plan for the robot quickly. This method could create more miniature, mobile robots to manipulate objects using their entire arms or bodies, potentially reducing energy consumption and costs. It may also be helpful for robots in space exploration missions where they need to adapt quickly to new environments using onboard computers.
Reinforcement learning, a machine-learning technique where robots learn through trial and error, takes a black-box approach and can be computationally inefficient for contact-rich manipulation planning. The MIT team found that smoothing removes many of the unimportant decisions and allows reinforcement learning algorithms to perform effectively and efficiently. However, the technique cannot currently handle highly dynamic motions, and the researchers plan to enhance the technique to tackle these challenges.
Chemists are building a map of every possible molecule, which could revolutionize the discovery of new compounds for various applications. The challenge lies in determining the viability and properties of such many molecules. AI could be used to overcome this challenge. The idea is to create a map where molecules with similar properties are placed next to each other. For instance, compounds that react similarly to heat or other chemicals would be grouped together. This map could significantly speed up the discovery process for drugs, materials, and other substances. AI has the potential to assist chemists in analyzing and predicting the properties of molecules, ultimately leading to more efficient and effective research and development. By harnessing the power of AI, chemists hope to uncover new compounds that could have a wide range of applications in various industries.