This is a summary of the most important AI news stories from India and across the globe this week.
Salutations, AI
To take “corrective measures,” the Union Ministry of Information and Broadcasting (MIB) is investigating the use of AI to analyze news items. The ministry has issued a call for bids from firms interested in developing an AI-enabled dashboard that includes decision support tools, sentiment analysis, and AI-driven summary production.
Lenovo, however, began producing AI servers earlier this week at its Puducherry, India, plant. The goal is to achieve 50,000 rack servers with AI for enterprises and 2,400 GPUs annually. In addition, the massive electronics gear company opened a new lab in Bengaluru intending to research AI servers.
Features & Finds
This week marked the introduction of SocialAI, a new social media site in which you are the only “real” user, and the other “users” are AI bots.
Google said that its search engine would be improved with a new technological standard known as Content Credentials by 2025. It is intended to assist users in determining if photos and videos in search results have been manipulated using artificial intelligence.
Speaking of which, Google-owned YouTube has announced the addition of an in-house AI video model to its Shorts feature as part of a larger series of rollouts geared at enabling “new forms of expression for creators.”
AI startup of the week: Superman.
Superman, an AI coding company, generated a lot of excitement among VCs last week after it secured Bessemer Venture Partners sponsored the company’s first outside fundraising round, which raised $12 million. It attracted high-profile angel investors, including the co-founders of OpenAI and Perplexity.
What’s driving all this interest? According to TechCrunch, Supermaven has developed Babble, an in-house generative AI coding helper with a one million-token context window and decreased latency owing to a novel neural architecture designed from the ground up.
Quirk bot Tech specialists are creating a final exam with the “hardest and broadest set of questions ever” to assess advanced AI systems. The organizers of the test, dubbed “Humanity’s Last Exam,” are crowdsourcing the questions.
Moving on, a Bloomberg article on the progress of generative AI makes an apparent point that is worth noting: the way people engage with AI chatbots has evolved since 2022. How? AI models are getting more multimodal, which means that they are being used to generate songs and short movies in a matter of minutes, as well as text and images.
Catching heat.
LinkedIn came under fire this week for harvesting user-generated content on the site to train its AI models, which was not stated in the company’s privacy policy until recently. Here’s how you can opt out of AI training on your LinkedIn posts and personal data, which is enabled by default for all users.
Microsoft-backed OpenAI also came under criticism for allegedly threatening to block users who attempted to test the reasoning powers of its current o1 model. They said that using particular phrases like reasoning trail’ resulted in a warning email from OpenAI, which politely urged them to cease. Furthermore, OpenAI CEO Sam Altman has stepped off from the AI startup’s internal safety committee as it becomes more independent.
The AI rulebook
A United Nations committee has issued seven proposals to address AI-related dangers and weaknesses in governance. They include forming a panel to give objective and accurate scientific information about AI, as well as establishing an AI standards exchange, a worldwide AI capacity development network, a global AI fund, and a global AI data framework.
Finally, California Governor Gavin Newsom signed eight new AI-related measures into law this week. The majority of them aim to combat the deluge of AI-generated deepfakes, while others focus on recognizing AI-generated material using watermarks and establishing rules for the usage of AI in Hollywood. Around 30 AI-related laws in California are currently pending, including the major one, SB 1047.