Pinned Tweet
Software Architecture — A Simple Criterion
Ask an AI coding agent not to break your architecture and it will agree, and break it anyway. But with the Locality Criterion, enforcing architectural decisions is simpler than ever.
sapience.org/software-archit… #SoftwareArchitecture
🤖 Made with AI
14
IMMORTAL PSYCHOPATHS
In Optimal Timing for Superintelligence, Bostrom argues that superintelligence would save lives. But life extension will reach the richest and most powerful first, among them the most psychopathic. is this what we want?
tinyurl.com/4m4x6m63
3
Sapience.org retweeted
Chomsky was wrong.
Researchers from UC Berkeley and Google DeepMind cracked one of the biggest mysteries in AI.
For years, scientists have been finding impossible geometric shapes hidden inside the brains of LLMs.
Calendar months arrange themselves into literal circles.
Historical years bend into smooth, rippled timelines.
Cities organize themselves into actual 2D maps of latitude and longitude.
Everyone assumed this proved massive transformers were spontaneously inventing complex "world models" out of thin air.
They were wrong.
This paper proves you don't need a massive LLM for any of this to happen.
Even basic, decade-old word embeddings like word2vec produce the exact same circles, timelines, and maps.
The geometry isn't coming from the AI's architecture.
It’s coming from a hidden mathematical property inside human language itself.
Researchers call it translation symmetry.
In natural text, how often any two months appear near each other depends almost entirely on the time gap between them.
January and February share the same statistical relationship as July and August.
When a neural network trains on data with that kind of symmetry, the math forces it to learn Fourier representations, literal sine and cosine waves.
The circles? Those are the long waves.
The strange "ripples" researchers found on AI timelines? Just higher harmonics.
Here is the craziest part.
Rresearchers tried to break it.
They deleted every single sentence in the training data where two months appeared together. No "January" next to "February." No "March" next to "April."
And the circle formed anyway.
Why? Because the geometry is a collective phenomenon across the entire language.
Even if two months never touch in a sentence, they are both pulled into orbit by thousands of other words tied to the same seasons, weather, and events.
The rest of human language forces the shape into existence.
The Technological Singularity: A Unified Definition
Acceleration, discontinuity, and superintelligence are common to all accounts. A useful definition must add impenetrability, inevitability, and proximity.
sapience.org/technological-s… #ArtificialIntelligence
17
WHAT IS INTELLIGENCE, PART I: INTELLIGENT AT WHAT?
The EU's AI Act is in force, yet the Commission is still explaining what an "AI system" is. Your board duties are determined by a term everyone uses and no-one defines.
sapience.org/what-is-intelli…
1
7
HOW DO YOU DECOMPOSE YOUR HARNESS HIERARCHICALLY?
We're looking to *methodically* decompose hierarchically our harness, eg 15:37 of youtube.com/watch?v=M8HuXu_b…
Project-level --> Team-level --> Enterprise-level
10
Python: A programming language so impoverished that it is only suitable for primary school students and programmers with a lobotomy, is apparently the no. 1 tool for AI programming. How did it happen; and why does it persist?
3
Software Architecture — A Simple Criterion
Ask an AI coding agent not to break your architecture and it will agree, and break it anyway.
There is a simple solution. Enforcing architectural decisions is simpler than ever.
sapience.org/software-archit…
#SoftwareArchitecture
7
The Point of No Return: Singularity Hypotheses Retrospective
I edited the volume and chose no sides. That was the right call in 2013. Thirteen years on, neutrality is no longer tenable — it is a dangerous evasion.
sapience.org/point-of-no-ret…
8
Sapience.org retweeted
To those of you who claiming that smarter-than-human AI is decades away, what *specific tasks* are you willing to bet me that AI won't be able to do within 5 years?
Sapience.org retweeted
Colorization APIs are becoming widespread; AI-colorized historical photos are circulated without caveat. But is AI colorization providing an accurate image of the past? To find out, I digitally desaturated these color photos by Sergey Prokudin-Gorsky, taken between 1909 and 1915.
Sapience.org retweeted
In 2018, the UN Secretary-General, Antonio Guterres, said 'The prospect of machines with the discretion and power to take human life is morally repugnant'.
In 2021, this is more than a mere 'prospect'.
Algorithms are now being given the decision over who lives and who dies.
Sapience.org retweeted
The deadlines for our @VitalikButerin Fellowship in AI Existential Safety are fast approaching!
- Applications for our PhD Fellowships are due Oct 29;
- Applications for our Postdoctoral Fellowships are due Nov 5!
Apply at grants.futureoflife.org
This quoted post is unavailable.
Sapience.org retweeted
Prediction: By lowering the barrier to entry to create software, Codex (and future models like it) will make it faster and more worthwhile for people to learn to code.
The "no code" movement had a huge breakthrough: AI-generated code. The easier it is to write software, the more important it is to teach CS to every student and democratize the power to create technology that helps solve world problems. By @hadip & @gdb fal.cn/3hnFD
Sapience.org retweeted
Today, August 12, at 6:30 pm EDT, tune in to hear @GCRInstitute's @SethBaum discuss a variety of issues in AI governance. Brought to you by @eanewyorkcity. Register via Meetup. buff.ly/3s1AJIz
Sapience.org retweeted
In another turn of events, Mark Zuckerberg also respects his own privacy, by using a chat app that has end-to-end encryption and isn't owned by @facebook
This is the number associated with his account from the recent facebook leak.
Does software just keep getting worse? A minor programming error in an airliner ticketing data nearly brought down an aeroplane full of passengers.
theguardian.com/world/2021/a… theguardian.com/world/2021/a…
1
Does anybody else feel that much of the software developed today is as poorly crafted as it's ever been? That #SoftwareEngineering hasn't improved iin practice for decades? That code written today isn't better tested, easier to change, or more resilient to errors than 70 yrs ago?
3
#Ethereum market cap now bigger than Wells Fargo, Phillip Morris, Siemens, Bank of Cinna, HSBC, Volkswagen, Goldman Zacks, Uber, Lockheed Martin, and GlaxoSmithKline, for example
Ethereum entered the top 100 assets by market 🚀
Source: 8marketcap.com/
1
Sapience.org retweeted
We’ve developed two neural networks which have learned by associating text and images. CLIP maps images into categories described in text, and DALL-E creates new images, like this, from text.
A step toward systems with deeper understanding of the world. openai.com/multimodal/




















