Google is set to launch an AI-driven search feature called ‘What People Suggest’, which collates online health conversations to enable users to compare their situation with others who are in similar situations. The generative AI tool categorises varied opinions into simple-to-grasp themes, enabling users to get insights with links for additional reading. It is available only on mobile search in the US at present. It also said that it has enhanced its AI Overviews to create summaries for health-related searches. However, a 2024 report by The Senior List revealed that more than 70% of AI Overviews for over 200 health-related searches were deemed risky by medical professionals.
China is not keeping calm when it comes to taking the world by surprise by its advancements in AI. After DeepSeek, they now have Manus. Chinese AI startup Monica has launched Manus, an AI agent capable of independently planning, thinking, and executing tasks. Dubbed a ‘general AI agent’, Manus has gained worldwide attention, with some calling it China’s second ‘DeepSeek moment’ following DeepSeek-R1’s success. The AI agent is available through an invitation-only web preview and reportedly outperforms OpenAI’s DeepResearch on the GAIA benchmark.
Manus is capable of executing sophisticated real-world tasks like making websites, organising vacations, stock analysis, and insurance policy comparison. It has autonomous operation and asynchronous cloud-based capabilities, so users can submit tasks, log off, and get results afterward. A demo video demonstrated Manus collecting data, performing tasks, engaging with online platforms and managing 50 screens at once. The demo became viral within 20 hours, recording 200,000 views on X before noon the following day, with several users requesting invitation codes.
The ministry of electronics and IT (MeitY) is said to be in talks with X regarding Grok’s controversial answers in India. Elon Musk’s AI chatbot has been producing politically sensitive, abusive responses, even Hindi slang and misogynist insults. The controversy reached new heights after X user Toka asked Grok his “10 best mutuals.” With no response, he included a Hindi profanity in a second post. Grok responded with its own use of profanity in a retort, instructing him to “calm down” and “stop crying.” The interaction went viral, with another user posting a remark about AI losing control. Grok confessed to “having a little fun” but also noted the necessity for restraint on grounds of ethics. In February, Musk’s AI venture xAI introduced Grok 3, claiming it to be ten times more powerful than Grok 2.
Appu, the AI tutor
Rocket Learning, in collaboration with Google in terms of funding and expertise, has launched Appu, an AI-powered tutor for children aged 3-6, offering personalised, conversational learning in Hindi and in 20 more languages soon. Unlike traditional EdTech, Appu adapts to each child’s needs, using voice-based interaction and real-world examples.
However, the challenges it may face include balancing AI’s role in learning, engaging parents, and handling multilingual responses. Looking to touch 50 million families by 2030, Rocket Learning says it wants to democratise AI-powered education and get Appu into India’s early education infrastructure, particularly in government-operated Anganwadi centres.
OpenAI’s audio model
OpenAI has rolled out next-generation audio models to enhance live AI voice interactions. Key advancements are two new speech-to-text models, GPT-4o Transcribe and GPT-4o Mini Transcribe, with greater transcription accuracy than its earlier Whisper models. A new text-to-speech model adds expressiveness, and improvements in the Agents SDK (software development kit) enable smoother voice-controlled AI assistants. It says it places high emphasis on speech-to-speech (S2S) processing for natural conversation.
PhD-level AI agents
A new buzzword is taking precedence — ‘PhD-level AI’. It’s a reference to AI models capable of performing tasks demanding expert academic skills. OpenAI is reportedly said to introduce specialised AI agents, such as a ‘PhD-level research’ tool priced at $20,000 a month, a high-income knowledge worker agent for $2,000, and a software developer agent for $10,000. Reportedly, these AI agents will undertake sophisticated research, process big datasets, and prepare reports. While OpenAI has not disclosed a launch date or target customers, SoftBank has committed $3 billion to support its AI agent technology. However, it says that its o1 and o3 models employ ‘private chain of thought’ methods to replicate human problem-solving. The o3 model allegedly surpassed humans in critical benchmarks, achieving 87.5% on ARC-AGI (to compare AI with human intelligence) and 96.7% on the 2024 American Invitational Mathematics Exam.
But critics counter that ‘PhD-level’ may be a marketing gimmick and point to being wary of the credibility of AI-written research. These have included the credibility of AI-generated output and the models’ capability for creative thought. Meanwhile, adding to the debate around AI and research, Sakana, a Japan-based AI startup, has claimed that its AI system has successfully generated scientific papers that passed peer review.