Hey there, 

I hope you liked our first edition of AI Emergence - a newsletter covering AI while it is Emerging :)

If you are new here, we are building Human Intelligence about Artificial Intelligence while AI is Emerging!

What would be the format? Every week, we will break the newsletter into the following sections:

• The Input - All about recent developments in AI

• The Tools - Interesting finds and launches

• The Algorithm - Resources for learning

• The Output - Our reflection 

• Question to ponder before we meet next!

Table of Contents

• The Input

  • Launch of ChatGPT store by OpenAI

  • OpenAI response to NYT claim

  • Midjourney V6 faces heat on IP infringement and ch …

  • Bezos and Nvidia back: Perplexity AI in bold move …

  • Nvidia gaining ground in ongoing AI GPU Market Rac …

  • Other Highlights

• The Tools

• The Algorithm

• The Output

  • Question to ponder before we meet next!

The Input

Launch of ChatGPT store by OpenAI

OpenAI launched the GPT Store on January 10, a platform for custom ChatGPT versions by partners and the community. It offers various Generative Pre-trained Transformer models, especially for ChatGPT Plus, Team, and Enterprise subscribers. The store includes a leaderboard, categories, and a revenue program for creators based on user engagement with their models. (source)

OpenAI response to NYT claim

OpenAI disputed The New York Times's claims of unauthorized article use for chatbot training, asserting the fairness of using public data and highlighting their commitment to ethical AI development and media collaboration. They also addressed concerns about AI 'regurgitating' content and questioned the transparency of the Times's allegations. (source)

Midjourney V6 faces heat on IP infringement and changes its ToS

Midjourney's V6 alpha model, capable of generating images resembling copyrighted content, has faced copyright infringement concerns. In response, Midjourney made a clear statement: they will actively pursue and seek compensation from anyone using their V6 model to infringe on intellectual property. They have called out that they are committed to protecting copyrighted material and ensuring their technology is not misused. (source)

Bezos and Nvidia back: Perplexity AI in bold move to rival Google search

Perplexity AI, an AI search engine, recently received $73.6 million in funding from Jeff Bezos and Nvidia. It aims to challenge Google's search dominance with unique features like conversational search and an ad-free user experience, now valued at $520 million. (source)

Nvidia gaining ground in ongoing AI GPU Market Race with new launches

Nvidia has launched three new AI chips: RTX 4060 Super, RTX 4070 Ti Super, and RTX 4080 Super, ranging from $599 to $999. These GPUs are designed with extra tensor cores to efficiently run generative AI applications. Additionally, Nvidia provides tools for developers to train and optimize Large Language Models (LLMs) on these GPUs, enabling both local and cloud-based operations of generative AI models. This move strengthens Nvidia's position in the consumer AI hardware market. (source)

Other Highlights

• Alphabet's Isomorphic, developed by Google DeepMind, has secured two major pharmaceutical deals with Eli Lilly and Novartis, worth almost $3 billion. (source)

• Intel strengthens its automotive presence at CES by acquiring Silicon Mobility SAS for EV energy management and unveiling AI-powered vehicle SoCs for enhanced monitoring. (source)

• Walmart CEO Doug McMillon unveiled AI-driven enhancements for customer experience at CES, featuring AI search tools and an AI-powered system at Sam's Club, set to eliminate receipt check lines and expand to 600 clubs from an initial 10-location pilot. (source)

The Tools

• Google's Bard Advanced, an AI chatbot integrated with Gemini Ultra for enhanced capabilities, offers a three-month free trial. It features the "Motoko" tool for custom bot creation and a "Gallery" for exploring various topics. (source)

• Apple plans to introduce new AI tools in iOS 18. This update aims to boost device functionality and user experience, highlighting Apple's commitment to evolving AI technology. (source)

• Project Mockingbird, initiated by McAfee, is a cybersecurity measure against AI audio scams. It utilizes advanced AI to identify and block fake audio, aiming to enhance online privacy and security. This initiative is particularly focused on tackling issues such as deepfake cyberbullying, misinformation, and fraud, thereby promoting safer internet usage. (source)

• Deloitte's new AI chatbot, 'PairD,' is designed to boost productivity among its 75,000 employees in Europe and the Middle East. It automates tasks like email drafting, presentation creation, and coding. While Deloitte advises caution in using PairD's outputs, the initiative also extends to the disability charity Scope, showcasing Deloitte's focus on inclusivity and responsible AI implementation. (source)

The Algorithm

• I am planning to take this new course on langchain.js built by Deeplearning.ai and Langchain. According to Deeplearning.ai- “In this course, you’ll expand your tool kits with LangChain.js, a popular JavaScript framework for building applications with large language models (LLMs), and grasp useful concepts for creating context-aware applications.”

• In the recent episode of 'Leading With Data', I spoke with Khalifeh Al Jadda, who transitioned from Glycoinformatics to heading a recommendation engine team at Home Depot. He highlighted the significance of LLMs in conversational search and stressed the role of data science leaders in educating business executives.

• Test your AI knowledge with our new quiz series, featuring 10 AI-related questions! It's perfect for everyone, from experienced data professionals to beginners and puzzle enthusiasts.

• AI and next-gen cloud tech are supercharging scientific discovery, especially in chemistry and materials science, through a Microsoft-PNNL partnership. (source)

• Article from Ethan Mollick on the impact of AI - A very good summary of the impact AI has created over the last 12 months (source)

The Output

Wow! The week feels full of action - again!

Thoughts on overfitting problems of LLMs

In the good old Machine Learning days (sounds like a different era!), we used to check the models for overfitting before finalizing the model. That makes sure that your model is not outputting the training dataset closely. For example, in decision trees, your final nodes should not contain thin data.
However, in the new LLM-driven world - that call sits with the model creators as you are largely a user. I feel that models like mid-journey and OpenAI are overfitting in the niche prompts now. This is also leading to the copyright claims.
It would be interesting to see how these models handle these scenarios. I expect some form of “overfitting correction” to come in the future models. Let’s keep an eye out for them.

Question to ponder before we meet next!

In the last edition, I asked how you feel about AI assisting the judicial systems and it was great to hear views from all of you replying to this email. 
I think that the use of AI to filter straightforward use cases should be a great enabler for humans to focus on more complicated cases. For complicated cases, AI currently does not have the empathy and the understanding required to make decisions - so they continue to sit with humans.