Rami's Readings #25
The latest on LLMs, Speech Generation, the Great Indian Back Office, Dubai Real Estate, Kenya, and more.
Welcome to Rami’s Readings #25 - a weekly digest of interesting articles, videos, Twitter threads from my various sources across the Internet. Expect a list of reads covering technology, business, culture, fashion, travel and more. Learn about what I do at ramisayar.com/about.
Are you enjoying this newsletter?
If yes, I have a favor to ask you. If you know someone else who would benefit, please forward or share it with them!
If you received this newsletter from a friend, subscribe here to get the next one.
🤖 AI Reads
Voicebox: Text-Guided Multilingual Universal Speech Generation at Scale
Notes: Another big release from Meta! Paper. Follows the previous MMS release.
Voicebox can be used for mono or cross-lingual zero-shot text-to-speech synthesis, noise removal, content editing, style conversion, and diverse sample generation. In particular, Voicebox outperforms the state-of-the-art zero-shot TTS model VALL-E on both intelligibility (5.9% vs 1.9% word error rates) and audio similarity (0.580 vs 0.681) while being up to 20 times faster.
MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoning & ReAct: Synergizing Reasoning and Acting in Language Models
Notes: I got reminded of these older papers in a conversation this week: MRKL and ReAct. They are worth re-reading.
InternLM/InternML-techreport - Jointly Developed by Shanghai AI Lab and SenseTime
Notes: Is this their first public open LLM? Does anyone know of previous publications?
InternLM is a multilingual large language model jointly developed by Shanghai AI Lab and SenseTime (with equal contribution), in collaboration with the Chinese University of Hong Kong, Fudan University, and Shanghai Jiaotong University.
Exploring the MIT Mathematics and EECS Curriculum Using Large Language Models
Notes: Neat!
Our results demonstrate that GPT-3.5 successfully solves a third of the entire MIT curriculum, while GPT-4, with prompt engineering, achieves a perfect solve rate on a test set excluding questions based on images.
What will GPT-2030 look like?
Notes: Read 2. Inference Speed and 4. Knowledge Sharing sections.
💼 Business Reads
Goldman’s Biggest Office Beyond New York Attests to India’s Rise
Notes: India rising! 🚀I can’t wait for my next visit.
Y Combinator leads accelerators in unicorn-creation rate
Notes: Power law everywhere! Seriously!
Cost Benefit Comments
Notes: Hidden details that matter to every day lives.
The Biden Administration is proposing major changes to cost-benefit analysis used in all regulations.
A $204 Million Mansion Is Dubai’s Most Expensive House for Sale
Notes: More ridiculous Dubai real estate. 🤪
China Faces World’s Biggest Millionaire Exodus as Economy Slows
Notes: It's surprising that the rate of outflows from the UK is being maintained - you would think the Brexit-triggered outflows are over. Also, what is happening in Australia to attract HNWI inflows at this rate?
Australia is expected to surpass the United Arab Emirates to become the top country to attract the inflows of HNWIs, with Singapore in third place. About 5,200 millionaires will move to Australia this year.
Kenya Is Poised to Become the ‘Singapore of Africa’
Notes: Tyler Cowen for Bloomberg on Kenya.
How Rich Were the Rich? An Empirically-Based Taxonomy of Pre-Industrial Bases of Wealth
Notes: Surprising to see the following countries and time periods be the most unequal: Mexico 1940, Iraq (Mesopotamia) 900, France 1866, Moscow region 1811 & United Kingdom 1867.
That is all for this week.
I wrote a disclosure for this newsletter in #7. Please consider reading it.
Reach out to me on LinkedIn or Twitter.