Nikhil Pahwa’s Post

We've just rolled out MediaNama's first AI deployment: Below each MediaNama article, you’ll see a tab or button called “Explainer.” Clicking on this tab generates a concise explainer in Q&A format for every story we publish. We’re calling this product Tabber AI, and work on version 2 will begin shortly. We’ll roll out other AI-based use cases shortly, and in case there’s a use case that you would prefer, do let me know. Tabber AI is in live beta, and we’re testing to see how it takes a large volume of usage and would appreciate feedback (I’m at nikhil@medianama.com), in case you face any issues with it. We chose not to integrate a chatbot because our understanding is that for a vast majority of users, looking at a chatbot is like looking at a blank page and wondering what to write. The ability to draft instructions and get high-quality outputs also varies. More details in the comments below.

So Kamya and I discussed AI outputs and the complicated aspect of regulating jokes generated on religious figures by chatbots.

In today's Times of India, I write about why the Indian startup ecosystem should celebrate its success like Naukri, MakeMyTrip and Indiamart and not mourn its failures because failure is a part of a founders journey and companies fail, founders don't. Aprameya and Mayank have not failed. I also warn about the impact of the removal of safe harbor protections for social media which would have impacted potential partnerships and acquisitions. cc folks (or their co's) mentioned in the article Sanjeev Bikhchandani Dinesh Agarwal Pratyush Prasanna Pravin Jadhav Kulin Shah Deep Kalra Aprameya Radhakrishna Mayank Bidawatka Beerud Sheth Vijay Shekhar Sharma Kavin Bharti Mittal Sachin Bansal Binny Bansal Aloke Bajpai Deepinder Goyal Rajeev Chandrasekhar

MediaNama's 16 years of coverage of the tech industry is public. In 2008, we had multiple Indian language search engines, including Raftaar which was Hindi only. Google's rollout of Indian language products was led by Rahul Roy-Chowdhury, who is currently CEO of Grammarly. Link to the rollout story in the comments.

I couldn't disagree with this more. The difference between cable TV devices and a mobile phone, which works as a device to access the internet is in the role of a user. With cable TV devices we are mere consumers of content, and scope of interactivity is the usage of buttons to perform a limited set of actions. With mobile devices, we are creators: we consume, we share, we interact, we play, at scale in a manner that cable never was. We might be consuming 100x more than we create, but there are also a billion times more creators creating than there were with cable. Also consider games, a fully interactive medium that is actually responsive to what we do in the game: the content changes as we change it. This was never possible with cable TV. We are, also creating actions and not just content. We are sending emails, messages, emojis, and asking tools to do our work for us. We are interacting with friends and family across the world. How someone can equate an interactive medium with a pure consumption medium is beyond me.

An honor to be speaking at the Closing Plenary of CPDP LatAm, especially about a topic I feel strongly about: AI and privacy. In you're there (or in Rio and want to meet), do ping. I'm in Rio for the week.

So, #ShameOnMetaAI was trending on Twitter because MetaAI was used to generate jokes on Hindu gods including Lord Ram & Lord Vishnu. Some tweets say it declined to crack jokes on Prophet Mohammed, but others have screenshots with jokes on him. We tried this as well, and there's a clear trend of inconsistency when it comes to the output from Meta AI when it comes to these jokes. Sometimes it cracks jokes, at others it doesn't. If you ask it to crack disrespectful jokes, it doesn't. You need to keep a few things in mind: 1. AI outputs are trying to please you. You want a joke & ask for it, it will try its best to create one. It's meant to give you an output you want, not an output that is offensive. It assumes if you're asking for a joke, you want one (and won't offend because it's a joke). 2. Most importantly, an output is not just based only on the system: the AI Model or its training. It's impacted by the user as well. Someone can provide problematic context and it can lead to a problematic output. These things are still being fixed but *no guarantees*. 3. The output is going to be inconsistent and factually wrong because it doesn't pull a joke or any info from a database. It tries to create/generate a probabilistic output, afresh. You probably won't get the same joke twice, because it's generating on the fly. 4. It might even also not be able to distinguish btw Ram & Lord Ram at times, or Mohammad and Prophet Mohammad. These things are tricky for a chatbot because it's trying to estimate a desired outcome... An outcome that you desire. At times it will be more creative, less accurate. All probability based. It's also difficult for an AI co to instruct what a chatbot should & shouldn't say or do. It doesn't have social/religious/local context. Sometimes that instruction to not make jokes will work. 5. Sometimes it can be overridden. People have jailbroken restrictions using prompts. Chatbots never be 100% accurate or reliable. 6. So, whose responsibility is this? Who's liability is it for an output? Is it the developer? Is it the user? We don't know because we don't know who's responsible? The model, the user, the person publishing the screenshot. We will have regulation eventually and these issues will impact AI and its deployment in India. If you'd like to discuss these issues, we're hosting a round table on it on the 11th in Delhi. Link in the comments. And some slides explaining these points from my AI workshop attached.

MediaNama is hosting a roundtable discussion on "Governing the AI Ecosystem" on July 11, 2024, in Delhi. We will look at aspects of artificial intelligence (AI) need regulation, the approach India should adopt for the regulation, and how India can develop a conducive environment for the growth of AI development and deployment.  So far, the Indian government’s attempts to regulate AI have come in the form of two advisories that were issued earlier this year, which left a lot to be desired with it mandating under-tested or unreliable AI models to seek government approval before being available for public use in India. While a second advisory addressed some of the problems with the first, the fact remains that a more nuanced approach is needed that addresses legitimate concerns around liability, ownership of AI-generated content, and the potential harms emerging from AI while still fostering an environment conducive to AI development. MediaNama’s roundtable discussion will bring together stakeholders to help shape a balanced regulatory approach that protects against risks while unlocking AI's transformative potential. Those building AI tools and products are especially encouraged to apply to attend and contribute their expertise to this discussion. Do join us for the discussion: this is a physical event with limited space and a curated audience. Registration link in the comments.