Marta May

2x PhD, 3x lingual dropout, co-founder MD practice in the Arctic Circle, working on cognitive and behavioral AI

[Marta May @martiimay]

the frontier always sees the problems first

[Demis Hassabis @demishassabis]

https://t.co/PTeDiv1b6L

[Marta May @martiimay]

the communication channel between human and AI is too narrow

[John Schulman @johnschulman2]

We started Thinking Machines a year and a half ago with a couple of instincts: that people should have much more ability to customize models and do research on them, and that even as AI becomes more autonomous, there's a lot more to build to make humans and AIs work well together.

A lot has happened since then, especially the massive progress in agents, so we wanted to revisit those instincts in light of everything we've learned, argue about them, and write down what we actually believe now.

This is where we landed after a lot of debate. I'm happy with it!

[aether @bluewdrift]

we get maybe seventy summers if we’re lucky. seventy times to taste peaches at their peak. seventy Augusts to swim in water so warm it feels like absolution. seventy chances to fall in love with the same person in different ways, or different people in the same way. 3/6

[Anthropic @AnthropicAI]

New Anthropic research: A global workspace in language models.

Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with.

We found a strikingly similar divide inside Claude.

[Russell Brunson @russellbrunson]

"Entrepreneurship = the best personal development program on the planet.
Daily risk. Daily growth.
Unlimited upside."

[Marta May @martiimay]

<3 nerds

[Paul Graham @paulg]

No one knows yet what the next form factor for computing will be, and yet there will be a next form factor, and it will seem obvious in retrospect.

[Marta May @martiimay]

I don’t think emotions and relationships can be reduced to formulas,they’re too dynamic and context-dependent, even though patterns often repeat. What we can model are relatively stable traits and processes that influence behavior. Those models are imperfect, but they can still provide useful probabilistic insights

[Marta May @martiimay]

after what I’ve learned recently, I’m increasingly convinced that we should apply mathematics to every aspect of life.

wherever there is pattern, there is structure.
wherever there is structure, there is information.
and wherever there is information, there is something we can understand, predict, and improve

[Marta May @martiimay]

@JimEberhart imo we’re much closer to impressive demos than to true 95%+ autonomy.
the challenge isn’t doing a task once. It’s handling thousands of edge cases, changing environments, interruptions, mistakes, and recovery without human intervention- at a reasonable cost

[Marta May @martiimay]

@BlueTimeFairy Mathematics isn’t valuable because it removes uncertainty. It’s valuable because it helps us quantify it

[Marta May @martiimay]

@elGabrielC You should check out Neri Oxman! She does an amazing job borrowing design ideas from nature

[Marta May @martiimay]

most companies optimize for individual intelligence. but cognition is distributed.
the unit that thinks isn’t the person- it’s the system

intelligence isn’t the bottleneck anymore.
how it moves is
https://t.co/k2mltGMXM4

[Marta May @martiimay]

@elGabrielC I thought the same about the bees!

[Sam Altman @sama]

man the early days of the internet were so special

[Marta May @martiimay]

@ZeinolabedinV Norway

[Marta May @martiimay]

10:30 pm and the sun is still shining

[Marta May @martiimay]

@radbackwards the carpet looks like grass!

[Marta May @martiimay]

catching some wind

[Marta May @martiimay]

@mistastuttas I’ll send you a link to the field notes once we publish them

[Marta May @martiimay]

building this kind of thing makes me disproportionately happy