The Artificial Intelligence Podcast

A recent study funded by the NIH reveals that AI models analyzing social media language can accurately predict the severity of depression in white Americans, but not in Black Americans. By analyzing Facebook posts, researchers found that words and phrases associated with depression were much more indicative of depression severity in white individuals than in Black individuals. The study, conducted by researchers from the University of Pennsylvania and NIDA, underscores the necessity of diverse datasets when developing AI models to prevent perpetuating healthcare disparities. The study challenges assumptions about language and depression among different racial groups, highlighting the importance of race in mental health expressions.

San Jose, located in Silicon Valley, is conducting a unique experiment to train artificial intelligence (AI) to identify homeless encampments. Companies have been invited to mount cameras on municipal vehicles to collect footage of streets and public spaces, which is then fed into computer vision software to train algorithms to detect tents and lived-in vehicles. The city hopes that this technology will help address complaints about homeless encampments more efficiently. While there are potential benefits, concerns about privacy and potential misuse have also been raised. San Jose's pilot project could influence how other cities adopt similar detection systems. The ongoing debate surrounding AI surveillance of homelessness and its implications for privacy and human rights is expected to intensify.

The World Health Organization (WHO) and the UN Special Programme on Human Reproduction (HRP) have released a technical brief on the use of artificial intelligence (AI) in sexual and reproductive health and rights (SRHR). The brief highlights the opportunities and risks that AI presents in this field, including AI's potential to make sexual and reproductive services more accessible. However, there are concerns over data breaches, bias, unequal access, and misinformation. The brief suggests actions such as revisiting data protection regulations, ensuring diversity in training data and development teams, and addressing misinformation. Ethical and inclusive development and regulation of AI in SRHR are crucial.

A study by AI productivity platform Plus Docs has revealed the states in the US that depend the most and least on AI-based searches for work tasks. Iowa, Wisconsin, and Mississippi were found to rely the least on AI, with average monthly searches per 100,000 residents ranging from 27.2 to 28.1, significantly below the national average of 38.1. On the other hand, Georgia emerged as the state that uses AI the most, with an average of 52.1 searches per 100,000 residents, followed closely by New York and California. The study attributed these variations to different industries and comfort levels in adopting AI technology.

Port Ellen, a distillery in Scotland, is leveraging AI technology known as SmokeDNAi to produce limited edition whisky. The distillery has introduced two rare bottles called Port Ellen Gemini, each priced at $50,000. SmokeDNAi allows the distillery to delve into and understand the aging process of whisky in casks. The parent company, Diageo, has allocated $44 million from a $230 million investment into whisky tourism projects to further explore whisky maturation using SmokeDNAi. This AI technology is used to analyze flavor profiles and mouthfeel of whisky distilled in different casks, aiming to enhance understanding of aging and control over flavors. The reopening of Port Ellen distillery after 40 years has marked the introduction of SmokeDNAi, contributing to maximizing production, flavor, and sales of whisky.

Tech firm Nvidia is making strides in the field of artificial intelligence (AI) development with its latest project, Project GR00T. This project aims to advance humanoid robots using generative AI technologies. Nvidia showcased its robot at the recent GTC conference, along with other Star Wars-themed robots. The GR00T system, also known as Generalist Robot 00 Technology, is designed to improve robots' skills and tasks in various environments. One key component is the Jetson Thor computer system, capable of performing complex tasks and interacting with people and machines naturally. Nvidia's pivot to AI development has resulted in increased revenue, and its advancements in robotics present potential opportunities in industries facing labor shortages.

Apple CEO Tim Cook emphasized the importance of artificial intelligence (AI) in combating climate change at the China Development Forum. Cook stressed how AI can help businesses reduce their carbon footprints and achieve carbon neutrality. He praised China's vibrancy and dynamism during his discussions with China Premier Li Qiang and other international CEOs. Cook also highlighted Apple's commitment to partnerships with suppliers who share their commitment to innovation and environmental protection. The emphasis on environmental initiatives during Cook's trip may be a strategic move to navigate potential geopolitical sensitivities between the US and China. Apple aims to achieve carbon neutrality and eliminate plastic from its packaging by 2030 and 2025, respectively, showcasing its dedication to sustainability.

A bipartisan bill has been introduced in the US House of Representatives that aims to regulate the use of artificial intelligence (AI) by requiring the identification and labeling of AI-generated online media such as images, videos, and audio. The legislation seeks to address concerns over deepfakes, which are AI-generated media that can replicate real content. AI developers would be required to embed digital watermarks or metadata in their AI-generated content, while online platforms like TikTok and Facebook would be obligated to label such content as AI-generated. The bill also includes provisions for civil lawsuits against violators. The legislation reflects a growing recognition of the need for accountability and trust in AI development.

Researchers at The University of Texas at Austin have developed a method called "machine unlearning" to remove copyright-protected and violent content from generative AI models. This approach allows for the active blocking and removal of undesirable content without starting from scratch. Traditionally, the process required manually removing data and retraining the model. The new method is significant for ensuring responsible use and making generative AI models commercially viable. The research focuses on image-based generative AI models and incorporates human teams in content moderation and removal. This research opens up possibilities for addressing copyrighted and violent content in generative AI models.

Biotech companies, such as Generate Biomedicines and Montai, are utilizing artificial intelligence (AI) to enhance the drug development process. Generatehas trained their AI to design new proteins that do not exist in nature, expanding the scope of disease-treating proteins. They have already generated approximately 5 million of these proteins and are focusing on autoimmune conditions, cancer, and infectious diseases. The company's AI-assisted drugs for asthma and a COVID-19 monoclonal antibody treatment are entering the first phase of clinical trials. Similarly, Montai uses AI to analyze natural molecules and identify substances that could target specific pathways in the body. Over 125,000 potential substances have been identified for developing treatments for chronic illnesses. However, the success of these AI-assisted treatments will depend on clinical trials. While AI in drug development offers promise, it still requires significant human work and faces challenges such as data gaps and limitations.