The Happiness Equation

Hey everyone,

A special welcome this week to our new subscribers—great to have you here!

This is your Sunday Space, where I serve up the best ideas, tools and resources I’ve found each week as we explore the technology shaping the future.

If you find something thought-provoking, forward it to a friend.

IDEAS
The Happiness Equation

Jensen Huang is known for taking NVIDIA from a sub-$500B market cap to the world’s most valuable company in less than a year and a half.

He’s less well-known for giving life advice. However, his equation for living a happy life is beautifully simple:

Happiness is equal to your expectations less your reality.

The more I think about it, the more powerful it becomes.

If I expect myself to be as successful as Elon Musk, nothing short of starting 7+ successful tech companies and becoming the richest person on Earth will make me happy.

I know it’s exaggerated, but I lived with expectations like this for a long time.

On the other hand, if I only expect myself to provide for my family and do my best, I have a much higher chance of exceeding those expectations.

Reality is often disappointing, making it all the more important to manage our expectations each and every day.

Happiness equals expectations, less reality.

(Click here to share this idea on X/Twitter.)

RESEARCH
A Look Inside ChatGPT

OpenAI recently released a paper detailing a method for discovering “concepts” learned by GPT-4 to better understand their inner workings.

Interestingly, the paper followed accusations that OpenAI is taking unnecessary risks. Even more interesting is that it was authored by the superalignment team (led by Ilya Sutskever) most of whom have now left the company.

The technique, “Scaling and Evaluating Sparse Autoencoders,” identifies specific concepts inside GPT-4 from being probed by a smaller model.

Researchers can then extract millions of activity patterns for further exploration. For example, a feature of GPT-4’s training references “phrases relating to things (especially humans) being flawed.”

This study is fascinating because we really don’t understand how these models work as well as we should.

Unlike a car, where engineers can directly design, assess and fix the chassis based on the specs ensuring safety and performance, a neural network is a black box.

This takes us one step closer to directly influencing AI safety and alignment of frontier models. All I hope is that OpenAI continues the work that Ilya started.

AI Word of the Day

Superalignment

Superalignment is the challenge of ensuring superintelligent AI systems act according to human values. It involves developing methods to train and test advanced AI to reliably follow human goals even as they become smarter than their creators.

This concept is crucial in AI safety. It aims to mitigate risks and create a future where superintelligent AI benefits humanity without unintended consequences.

INSIGHTS
1 Article

Alibaba’s LLM, Qwen, ranked as top open-source model

Alibaba's Qwen-2-72B-Instruct model has topped Hugging Face's new open weights LLM leaderboard, beating Llama-3-70B and Mixtral 8x22B by more than 10%. Evaluated using benchmarks like MMLU-Pro, Qwen demonstrated proficiency in handling complex inputs and advanced mathematics.

Luca’s take: It’s great to see models evaluated on refreshed benchmarks that are more challenging than traditional ones. Additionally, this new ranking is instructive for people underestimating China's AI capabilities. When we think of the AI arms race in a geopolitical context, right now, it seems it’ll be China vs the US, which is reflected in these rankings.

1 Post

1 Video

THOUGHTS
Quote I’m Pondering

"Your time is limited, so don’t waste it living someone else’s life.

Don’t be trapped by dogma—which is living with the results of other people’s thinking. Don’t let the noise of others’ opinions drown out your own inner voice. 

And most important, have the courage to follow your heart and intuition. They somehow already know what you truly want to become. Everything else is secondary."

— Steve Jobs, at the 2005 Stanford Commencement Address

(Click here to share this on X/Twitter.)

Was this email forwarded to you? If you liked it, sign up here.

If you loved it, forward it along and share the love.

Thanks for reading,

— Luca

P.S. Who’s gonna tell him…