📌 This week's AI Brew summary:

• AI Spotlight 🚀: Stanford drops 386-page report on AI industry

• Real-Life AI Stories 🤖: BloombergGPT to revolutionise financial industry and Indian TV channel debuts multilingual news anchor powered by AI

• AI Jargon Buster 🤷‍♂️: Generative Pre-trained transfomer (GPT) AI model

• Not So Cool Corner 🤢: Chaos-GPT - An AI Tool That Seeks to Destroy Humanity

Stanford has graced us with a whopping 386-page report on the dynamic AI industry. But worry not! We've distilled it into digestible, witty nuggets for your reading pleasure. So sit back and savor our top 10 key points:

1. Academia lags behind: Industry produced 32 significant ML models in 2022, leaving academia with just 3. Key advantage? Deep pockets and data, a resource academia struggles to match

2.  AI job demand: AI-related job postings rose from 1.7% in 2021 to 1.9% in 2022 across most US sectors, highlighting the growing need for AI-proficient workers

3.  AI funding dip: In 2022, global AI private investment dropped 26.7% to $91.9 billion. Despite the dip, the investment is still 18x greater than in 2013

4.  AI adoption plateaus: McKinsey's survey shows AI adoption doubled since 2017 but plateaued between 50% and 60%. Adopters report cost reductions and revenue increases

5.  Policymakers and AI: AI Index analysis shows a growth from 1 AI-focused bill in 2016 to 37 in 2022, revealing the rising interest of policymakers in AI regulation

6.  AI perception: A 2022 IPSOS survey found 78% of Chinese respondents view AI positively, followed by Saudi Arabia (76%) and India (71%). Only 35% of Americans share the sentiment

7. AI as a scientist: In 2022, AI accelerated scientific progress in hydrogen fusion, matrix manipulation, and antibody generation, showcasing its potential as a research powerhouse

8.  AI incidents spike: Misuse of AI has grown 26x since 2012, as seen in the AIAAIC database. Controversies in 2022 included deepfake videos and call-monitoring in prisons

9. Benchmarks slow down: AI improvements on traditional benchmarks are marginal, but new benchmarks like BIG-bench and HELM provide fresh challenges for AI performance evaluation

10.  AI's carbon conundrum: BLOOM's training run emits 25x more CO2 than a transcontinental flight. However, BCOOLER's reinforcement learning shows promise for energy optimization

• Bloomberg has just unveiled GPT-50B, a jaw-dropping, 50 billion-parameter AI language model for the financial sector. Say goodbye to tedious number-crunching and hello to streamlined financial data analysis and laser-precise market projections. Investors can now spot trends faster than a cheetah, while risk managers pinpoint hazards like Sherlock Holmes. To top it off, there's a fancy schmancy educational program to train finance pros in AI wizardry. Move over, Wall Street wolves – the AI-driven financial revolution is here!

• An Indian news channel lauches full-time virtual anchor, Sana. This multilingual bot will keep you updated with news flashes throughout the day, without ever taking a coffee break or asking for a raise.

Imagine the Generative Pre-trained Transformer (GPT) as a super-smart AI language whiz. Its job is to understand and create human-like text. To become such an expert, it undergoes a two-step learning process.

1. Learning the basics: GPT reads a gigantic amount of text from books, articles, and websites. It's like a curious child learning from everything around them, picking up language patterns, structures, and subtle details.

2. Becoming a specialist: After getting a solid grasp of language, GPT is trained for a specific task, like summarizing stories, answering questions, or even writing poems. This phase turns the model into a master of its chosen field.

Now, how does GPT actually work? Picture it as a champion at guessing games. It predicts what comes next in a sentence by considering the words that came before.

For example, let's say we want GPT to complete this sentence: "AI Brew is an awesome AI ____."

GPT has already read tons of sentences about AI during its learning journey. So, when we give it the incomplete sentence, it uses its knowledge of language patterns and context to predict the missing word. In this case, it might generate "newsletter" or "blog" to complete the sentence: "AI Brew is an awesome AI newsletter."

So, GPT is like a language prodigy that can understand and generate human-like text by learning from heaps of data and using its remarkable guessing skills. It can finish sentences, come up with creative text, and even respond to questions, making it a truly impressive language expert. Here is a great video to understand this better:

• In a bizarre turn of events, an anonymous mastermind decided to unleash the dark side of AI, transforming Auto-GPT, an open-source programme that empowers ChatGPT to autonomously perform tasks, into a "destructive, power-hungry, manipulative AI." They called their creation Chaos-GPT. Cue evil laughter

📧 Got a question or an AI topic you'd like us to cover? Email me at [email protected]

Hope you liked your cup of AI Brew - Share AI Brew with friends and family who might enjoy a weekly dose of AI caffeine! 🥳