India is choosing to not regulate AI

Plus Anthropic wants to outbuild OpenAI

Author

The Alignment
April 07, 2023

India, soon to be the world's most populated country is choosing not to regulate AI. OpenAI believes that AI regulation will be important, but it will be equally important to continue deploying solution for real users to use. Also, Anthropic founded by ex-OpenAI folks wants to outbuild OpenAI. 

Let's dive right in ! 

 As always if you enjoy reading our posts be sure to spread the word ! 

India will not regulate the growth of AI within the country. The government has identified the AI sector to be a significant strategic area for the country and wants to nurture the growth that it promises. This move comes at a time when calls are being made for tighter regulation in other parts of the world. 

India’s Ministry of Electronics and IT has assessed the risks and ethical concerns and will implement policies and  infrastructure to minimise harm. However, the goal will not be to regulate AI’s “growth”. The government believes that AI will be a cornerstone for entrepreneurship and business development with the immense opportunity to build personalised citizen centric services. 

The lack of AI regulation means that the local startup and tech scene should feel encouraged to experiment with AI. International companies may also be attracted to the region for similar reasons. 

India is one of the largest economies in the world. Its population will soon overtake China. With a young population and highly skilled tech sector, a lack of regulation towards AI could be the dawn of new technological and business growth in the country.

Anthropic’s founding team has a strong presence of ex-OpenAI folks who worked on the initial GPT models. This is probably why they believe they have what it takes to seriously challenge the market dominance that OpenAI has established over the past 6 months. 

Anthropic is putting its bets on its Frontier model. A model that can self-teach  and align itself with human intentions. To build this, Anthropic believes it will require 10^25 floating point operations, several orders of magnitude larger than the biggest models today (although we don’t quite know how many FLOPs are running on GPT-4).  This will easily require tens of thousands of GPUs, which in turn is going to cost a lot of money. 

This Frontier model will be the successor to Claude, Anthropic’s very own competitor to ChatGPT. But this model will be “much less likely to produce harmful outputs,” “easier to converse with” and “more steerable.”

All of this will come at a cost. $5 billion to be precise. That’s right, Anthropic is aiming to raise $5 billion over the next 2 years, that’s a lot of money but that’s what you need to compete with OpenAI.

OpenAI has not publicly commented on the Pause AI petition that has been making waves. But now they have released a post outlining their approach to safety. Their main point is that AI alignment is not a task that can be achieved in the lab. In closed environments it’s impossible to see unforeseen use cases and real world experimentation is the only solution. 

For every “GPT” release that happens there is an entire army of red teamers (both internal and external people) at OpenAI that are flagging system vulnerabilities. The company is particularly focusing on child safety and privacy data, which are directly integrated in OpenAI's models through their training data. 

Build, test, iterate, deploy that is the OpenAI mantra, and so far it seems to be working. 

Around the industry - 

Product Corner - 

There are way too many cool products to showcase just one !

  • Wisdom AI : Unlock Your Content With Conversational Chat Powered By GPT-4

  • AI Backdrop : Hyper realistic backgrounds for any picture 

  • June AI : Product analytics and sql in simple english

  • Quadratic : Spreadsheet with Python, SQL and Formulas built in

  • Banter AI : Call up your favourite celebrities and talk with realistic AI

Apollo: a ChatGPT-powered app for real-time knowledge

Talk to it all day long through your headphones and have it answer all of your questions.