Ben Moskowitz

[Ben Moskowitz @BenMoskowitz1]

@ItaiYanai Exactly! Open endedness and non-instrumental scientific pursuits have historically led to some of the most impactful discoveries. You left out PCR as well :)

[Ben Moskowitz @BenMoskowitz1]

AI is misaligned.

Language is necessarily underspecified. As the programmers lament goes:

"I hate this damn machine. I wish that they would sell it. It never does quite what I want. But only what I tell it."

How can we solve this issue?

[Ben Moskowitz @BenMoskowitz1]

@navvye @tylerarose0 @shaurya_sinha7 Bitter lesson def coming to biology. Congrats!

[Ben Moskowitz @BenMoskowitz1]

@apex_scholars Lots of ways! Probably initially reviewed by expert assessment to see if it's satisfactory to persist, then we can use things possibly similar to h-index or economic value. No metric we have initially will be perfect, but we can keep iterating on that too!

[Ben Moskowitz @BenMoskowitz1]

Why do we prespecify a hidden test for all AI benchmarks?

What if we instead we make a benchmark that let's AI produce artifacts which we can evaluate their value over time? Similar to a PhD student writing a thesis.

[Ben Moskowitz @BenMoskowitz1]

Link to OpenAI/Retro work: https://t.co/yfG8jLH6fp

[Ben Moskowitz @BenMoskowitz1]

A new hypothesis for protein engineering:

As life becomes more complex, proteins tend to evolve to work across many contexts, composing into higher level organization. That means these proteins don't optimize any singular function.

OpenAI/Retro utilized this to optimize the Yamanaka factors (specifically SOX2 and KLF4) to achieve >50x expression of reprogramming markers and hit rates of 30-50% (from 0.1% baseline).

Evolution gave us the constrained search space. We have the necessary AI models and biological tools.

Lets get this paradigm started!

[Ben Moskowitz @BenMoskowitz1]

Why is everyone obsessed with RSI on a single agent?

Every major evolutionary breakthrough came from greater organization of previously independent agents ( eukaryotic cells, multi-cellularity, human culture...).

What might this transition look like for AI?

[Ben Moskowitz @BenMoskowitz1]

@owl_posting You've been my biggest inspiration for now starting my own blogging journey! Your writing will be missed. Go and do great things!

[Ben Moskowitz @BenMoskowitz1]

The century of biology is here, but we're still not talking about surgical robotics enough!

If we had superhuman surgical robots, what procedure would you most want invented?

Read my latest article on the topic here:
https://t.co/AZV3SfPz6F

[Ben Moskowitz @BenMoskowitz1]

@lexfridman Also I'd suggest adding When Breath Becomes Air by Paul Kalanithi to your list. It's a very powerful, intense, yet beautiful work that I'm sure you'd appreciate @lexfridman.

[Ben Moskowitz @BenMoskowitz1]

@lexfridman I am going on a backpacking trip for 2 months and set the same goal of reading a book a week throughout the trip! Love this list and funny to see some overlap with my own list. Looking forward to hearing about your journey and sending you love ❤️

[Ben Moskowitz @BenMoskowitz1]

@bryan_johnson This is an awesome opportunity!!! As someone just graduating from college I am super interested in this aging research and I'm so excited to see how other people do on your blueprint program @bryan_johnson! Would love to find a way to make a career in this field.