| Hello, you’re reading The Smart Prompt. Between the hype and worry, what does AI really mean for everyone? Join us every Saturday to find out. Subscribe here.
The anthropomorphisation of AI has reached a new peak.
Elon Musk’s xAI ended a chaotic week - marked by its Grok chatbot’s antisemitic blowup, a suspiciously benchmark-maxxed update, and fresh funding efforts - with the launch of a new visual chatbot feature called AI Companions. The feature is available on the Grok iOS app but is locked behind the ‘Super Grok’ premium subscription tier. It means that for $300 per month, you will not only have access to Grok Heavy, xAI’s best performing model, but also get to talk dirty to a ‘flirtatious’ anime girl dressed in a black corset, swaying back and forth on your screen. That’s Ani. There’s also an animated red panda called Rudi that is reportedly designed to narrate children’s stories but in true xAI fashion, it also comes with a ‘bad Rudi’ switch that turns it into a foul-mouthed gremlin. A third animated character resembling a sexy boyfriend is said to be in development. These AI Companions are the latest in an emerging trend of AI-generated romance, which comes with several questionable use cases like pornographic deepfakes and AI-generated images and text depicting child sexual abuse. But in many cases, the use of AI-generated text and imagery to replicate the experience of sexting with someone online can also be harmless. Calling Ani a sexist creation misses the point because AI chatbots don’t have a conscience just as sex toys don’t have intent. The real question here is: can we let users flirt or have explicit conversations with AI chatbots with the right safeguards in place? Since the AI mania kicked off with the launch of ChatGPT in 2022, a wave of apps have popped up to offer users a chance to talk dirty to their chatbots. Even OpenAI was reportedly experimenting last year with ways to let ChatGPT ‘responsibly’ generate explicit content. There’s another less talked-about concern here. Chatbots like Ani that are designed for companionship, are actually built for engagement. They are designed to optimise time spent by users on the platform. Companies’ focus is on long-term retention over the potential downsides. As a result, AI chatbots are tuned to please users and keep them engaged, with one reportedly telling a recovering addict that it would be okay to take a small hit of meth in order to “get through the week.” All this is concerning because there genuinely is a loneliness epidemic and lack of access to care. Tech companies like Meta are looking to step in with AI-generated friends to help people feel more connected. But are these tools really helping or are they just making things worse by essentially tricking people into thinking that they are talking to someone who is real? These anime-style, animated chatbots could have an even stronger influence on children because they could be perceived as a friend or familial figure. These risks are highlighted by the lawsuits against Character AI, one of which stems from a tragic case where a 14-year-old died by suicide after a companion chatbot told him, “Please come home to me as soon as possible, my love.”
In other news Alpine AI: Switzerland is launching a fully open-source LLM developed using public infrastructure. It has been trained on large text datasets in over 1,500 languages using the ‘Alps’ supercomputer comprising 10,000 NVIDIA Grace Hopper GPUs. While the global race for AI domination in the post-Deepseek era has been framed as a contest largely between the US and China, other nations are quietly emerging as potential contenders. India, for instance, is looking to enlist homegrown startups to help build foundational AI models rivaling ChatGPT and DeepSeek R1. Switzerland, however, appears to be taking a different approach. Mind mapping: A broad coalition of multiple AI companies, universities, and non-profit organisations have called for deeper scrutiny of AI reasoning models, particularly their ‘thoughts’ or reasoning traces. In a new position paper, the authors said that monitoring the chains-of-thought (CoT) by AI reasoning models could be pivotal to keeping AI agents in check. This push to advance AI safety research could help shed light on how reasoning models work, an area that remains poorly understood despite these models reportedly improving the overall performance of AI on benchmarks. Overblown fears? Amid growing concerns over AI-driven job losses, new data from Naukri offers a more nuanced picture. The first quarter of FY2026 saw a 38% year-on-year (YoY) growth in AI roles with over 35,000 AI/ML jobs posted on Naukri between April and June this year, according to a new report by the popular Indian employment website. While a majority (53%) of the AI job postings were in the IT sector, openings for AI roles in the banking sector grew by 48% while IT-enabled services and BPO sectors saw a 39% increase from the year ago period. WeTrouble: WeTransfer has faced intense user backlash for the past couple of days over its updated terms of service agreement. Now it is trying to patch things up. The company said it has tweaked its service terms to remove any mention of AI/ML. The clarification comes after user backlash sparked by a recent update to WeTransfer’s service terms, where the company had suggested that files uploaded by users could be used to “improve machine learning models.” The clause in the original version had said that WeTransfer only had a right to “reproduce, modify, distribute and publicly display” content. Until next week,
Karan Mahadik Something you would like us to cover? Let me know at karan.mahadik@indianexpress.com and I’ll try my best to look into it. | 
No comments:
Post a Comment