January 2023
Welcome to 2023. I believe that humanity has never before begun a year with so much hope and promise. This month’s newsletter includes links on longevity, protein development and UBI.
I will be using this newsletter to keep myself and my subscribers up to date with the AI revolution. Thank you for being patient and for subscribing.
First thing’s first ; it took ChatGPT two months to reach 100 million users1. A feat that took the telephone 75 years2. Since then, it has passed the USMLE3, a final exam of a typical MBA course at Wharton4, and graduated law school5.
AI is also competing against humans in artistic endeavours and winning. We are in the exponentially exponential phase of AI, Moore’s law is in full effect and it is going to get very interesting very quickly.
I believe that we are eyeing the Singularity/AGI very narrowly from this year on, and the point is no longer in the far future. Dr. Kurzweil still estimates it to be around 20456, but other experts are now revising it to be highly likely by 20307. The distinctions between AI and human work are already blurry enough to the point that I had trouble classifying and organizing this newsletter. (AI has entered every field of human endeavor and cannot be separated from it, but I have still classified AI dominant work under the AI heading).
By the way, ChatGPT speaks romanized Urdu.
1 Artificial Intelligence
Moore's Law for Everything, by Sam Altman. This is one of the few pieces I’ve read that entertain a positive outlook on the future that awaits us post the AI revolution. Interesting new insights on how to achieve universal basic income (UBI).
AI 2030: What to expect, and five areas for investment, from our head of AI, by Andrew Moore, head of Google Cloud.
MetaAI’s CICERO, a diplomacy agent. This is the first AI power to perform at a human level in the strategy game Diplomacy. CICERO showed its abilities by playing with humans on an online version of the game, scoring more than double the average of human players and ranking in the top 10%. This achievement is due to the combination of strategic reasoning and natural language processing, allowing CICERO to reason, strategize, communicate, reach agreements, form alliances, and coordinate plans.
An A.I.-Generated Picture Won an Art Prize. Artists Aren’t Happy, the preview image for this newsletter.
Proteins
These three papers were published within a day of each other, and are all from MetaAI. Knowing the structures of proteins can help us understand the function of proteins and how they interact with other molecules. This information can then be used for various applications, such as developing new drugs, understanding and treating diseases, improving crop yields, and advancing our understanding of biology and biochemistry.
A high-level programming language for generative protein design, scientists have discovered a new way to design proteins, using AI. They created a computer program that can create new proteins by putting together smaller pieces like a puzzle. The program uses a special language to make sure the proteins have certain properties, and can make a lot of different combinations to create proteins with different shapes and abilities. This new method allows for more control and creativity in protein design compared to previous methods.
Evolutionary-scale prediction of atomic level protein structure with a language model, researchers used AI to predict the structures of over 600 million proteins from microorganisms that were not characterized before. The team fed the AI with sequences of known proteins and the AI learned to fill in the missing parts of the protein sequences. The process took just two weeks and the structures and code are freely available for use. The AI was trained with an understanding of protein sequences and combined that with information about known protein structures and sequences to generate predictions. The researchers tested their model on DNA from various sources and found many potential proteins from single-cell organisms that were previously unknown to science.
Language models generalize beyond natural proteins, researchers explored the potential of using language models for protein design. They found that language models, even though only trained on sequences, were capable of generating de novo proteins with successful experiment outcomes. The generated proteins showed a diverse range of topologies and secondary structures, and many of the designs were not found in natural protein families. This demonstrates the ability of language models to learn the grammar linking sequence and structure, extending beyond natural proteins.
2 BioScience
Papers and patents are becoming less disruptive over time, a team measured the disruptiveness of a paper by how obsolete it made previous work in the field, and found that it has decreased by 90% between 1945 and 2010.
They attribute this trend to scientists relying on a narrower set of existing knowledge which benefits individual careers but slows down scientific progress as a whole. The study also showed a shift in language used in scientific papers, with a decrease in usage of disruptive words like "produce" and an increase in words like "improve" or "enhance".
Biohacking
Time-restricted eating with or without low carbohydrate diet reduces visceral fat and improves metabolic syndrome: A randomized trial, three month study in China shows restricting your eating to eight hours a day in addition to low carbohydrate intake lowers subcutaneous and abdominal fat. Also lowers fasting blood glucose and uric acid levels.
Tirzepatide Once Weekly for the Treatment of Obesity, upto 21% loss of body fat in 72 weeks. This level of success reaches that of bariatric surgery!
These scientists used CRISPR to put an alligator gene into catfish, edited in proteins from alligators to make catfish (dietary item for Americans) immune to bacterial infections. Survival rate of the edited fish rose to “between two- and five-fold higher,”.
Long-term trends in human body size track regional variation in subsistence transitions and growth acceleration linked to dairying, dairy consumption correlates with increased body size. The study shows evidence of increased stature and body mass in regions such as Central and Northern Europe where the domestication of foreign crops was difficult and there is strong evidence for natural selection for lactase persistence in relation to dairying. This provides support for the "Lactase Growth Hypothesis."
Lactase is an enzyme present in the human small intestine responsible for the digestion of whole milk. A deficiency of the enzyme results in lactose intolerance, which presents with abdominal pain, diarrhea, nausea and bloating upon consumption of dairy products. Lactose intolerance affects 70-100% of East Asian communities, but only 5% of European communities8.
Longevity
Loss of epigenetic information as a cause of mammalian aging, decades of work from David Sinclair’s team. Here is a short video and Twitter thread explaining the paper.
To explain it simply, think of epigenetic information as the instructions for how a cell should function. Loss of this information is like losing the instructions, which can result in the cell not functioning as it should. And, according to this paper, the team has proven that this is why we age.
The team mimicked aging by introducing breaks in the DNA of young mice and observed signs of aging within weeks, including grey fur, reduced activity, and increased frailty. They then started gene therapy by using three genes that guide cells to reprogram themselves and restart the epigenetic changes. These genes can turn back the clock on adult cells to their embryonic state. Using three of these factors turned back the clock about 57% and made the mice youthful again.
The team is likening it to how we reboot softwares on our computer to make it work properly again.Selfish, Reckless, Satanic: Life Extension in Movies, article inspired by the new Avatar (go watch it in IMAX if you can). The author expresses regret that media portrayal is negatively affecting public opinion about life extension. The author is also insists longevity researchers and proponents are not villains.
I personally believe the pursuit of life extension is a frivolous pursuit. Death is a gift the immortals envy9. Advances in health science are already increasing lifespan and raising quality of life. A focus on halting aging feels superficial and superflous. Aging is a natural part of life and brings with it a certain level of humility and perspective.
Neuroscience
What is brain fog? - the paper doesn’t tell us anything new, it’s just interesting they “scraped” data from Reddit using Python for just over a week. No humans were recruited or surveyed for this study.
“The strengths of Reddit as a source of data are, first: comments and posts are located in the public domain, rather than contained within private groups as is often the case on Facebook; and that relatively long posts (up to 40000 characters) are allowed, supporting richer description than on Twitter, where individual posts are restricted to 280 characters. Reddit is a rich source of anonymous but candid and detailed qualitative data, and there are many examples of similar methodologies.” Social media is a knowledge pool for anyone with intelligent means to extract from.Identification of shared and differentiating genetic architecture for autism spectrum disorder, attention-deficit hyperactivity disorder and case subgroups, seven genetic variants are shared by both ADHD and autism.
Investigating cognitive neuroscience theories of human intelligence: A connectome-based predictive modeling approach, researchers found that considering the whole-brain connectivity was most effective in predicting general intelligence, rather than just looking at individual, localized brain regions.
Predictors of aversion to happiness: New Insights from a multi-national study, strongest predictors : unhappy childhood, perfectionism, belief in black magic and karma, and loneliness.
The connectome of an insect brain, “Some structural features, including multilayer shortcuts and nested recurrent loops, resembled powerful machine learning architectures. The identified brain architecture provides a basis for future experimental and theoretical studies of neural circuits.”…omg
3 Culture
Grow the Puzzle Around You, essay describing the qualities one needs to be successful by one of Y-Combinator’s cofounders Jessica Levingston.
On Guam there is no birdsong, you cannot imagine the trauma of a silent island, “in Guam, the god of death is being outdone by the god of war.”
The Guardian is publishing a series of essays named ‘Before it is lost’ from the Pacific Islands.Vietnam’s Red Napoleon, a good write up on Võ Nguyên Giáp, a history teacher turned military commander.
Titanic’s Greatest Unsolved Mystery Involves a Conga Line, P.C.P., and an Unidentified Chowder
Escaping Darwin’s shadow: how Alfred Russel Wallace inspires Indigenous researchers, the article discusses how Alfred Russel Wallace, who independently discovered the theory of evolution, used local knowledge in the Amazon to create his work on species ranges. Today, Indigenous scientists in the region are leading research by utilizing cross-cultural tools that were inspired by Wallace.
It's a little thing, but recommend you always have an emoji you use for the same Headings like this: https://bensbites.substack.com/p/new-entries-aws-and-bain
A good emoji finder: https://emojidb.org/science-emojis