Is the Small Language Model the new approach to solving longstanding AI challenges?

Welcome AI Brewers,

Have you heard about that new gpt2? It popped up out of nowhere, outclassed all other LLM models. Then it disappeared, leaving no clue about where it came from. It's wild—being in the AI field is just like living in a Hollywood movie.

In today’s edition:

• gpt2 chatbot, the mystery model has excellent performances in mathematical and logical puzzles, coding, and reasoning.

• Core Weave, Motional, Traceable AI and Blaize raised funding from the esteemed pool of investors.

• While the world is gushing towards building large language model, here is a company solving the fundamental problems of AI by building small language model

• Perplexity AI Founder talks about current landscape of search engine and how Google’s business model is more aligned with advertiser interest than user interest.

• Chart: AI 50 - Companies of the Future

• Bonus: Industry Report

Highlight of the Week

Continuing the earlier discussion we had on mysterious gpt2 chatbot that surpassed GPT-4 in many areas.

Speculation has been that the gpt2-chatbot is either GPT-5 being stealthily tested on benchmarks or a modified version of GPT-2 fine-tuned on modern data.

Deal Flow

• NVIDIA backed company Core Weave triples its valuation to $19 Billion in five months - Link

• Hyundai bought a majority stake in the self driving startup Motional Link

• Traceable AI raised $60 Million funding in Series B funding, valued at $450 Million Link

• AI Based Computing Solution Blaize raised $106 Million from investors like Mercedes Benz Link

Startup Spotlight

The future is in making language model smaller and not the other way.

This company inspired from the brain of a sea elegans worm to develop liquid neural networks, challenging traditional AI approaches.

This company uses liquid neural networks with 19 neurons and around 1,000 parameters challenging the traditional parameter-heavy systems.

Backed by investors like Naval Ravikant, Liquid AI, a MIT spinoff also secured $37.5 Million to build general-purpose AI systems using a new type of AI model called liquid neural network.

Founder’s Corner

The fundamental problem with search today is that it wastes a lot of time by providing links that are not ordered to save time but to make Google more money - Aravind Srinivas

Pick Aravind’s brain to learn more about AI search, the search industry in general and challenges it pose and how are they solving

VC Insights

Today, we have Pat Grady, Partner at Sequoia Capital talk about the opportunities in the AI universe. Pat further shared some fascinating insights around AI, some of them are below:

• The opportunity set for AI is massive and the landscape is wide open, with the potential for 40 or 50 new billion-dollar companies in the future.

• The rapid growth of generative AI is evident in its $3 billion in revenue, surpassing the SAS Market in just its first year out the gate.

• 2024 is the year we see AI applications take us from co-pilots to agents that can take the human out of the loop entirely, feeling more like a coworker than a tool.

Chart for the Week

Sequoia Capital shared their 2024 edition of the AI 50

Industry Report

• 1800 ML Practitioners and Leaders talks about AI adoption in the midst of Gen AI hype AI Readiness Report 2024

• AI Spending in the first quarter of 2024 Business Spending Benchmarks

Thanks for reading.

See you soon!

Arindam

p.s. If you liked this newsletter, share it with your friends and colleagues brew ai