petteri retweeted
1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers.
We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy
petteri retweeted
Sufficiently advanced agentic coding is essentially machine learning: the engineer sets up the optimization goal as well as some constraints on the search space (the spec and its tests), then an optimization process (coding agents) iterates until the goal is reached.
The result is a blackbox model (the generated codebase): an artifact that performs the task, that you deploy without ever inspecting its internal logic, just as we ignore individual weights in a neural network.
This implies that all classic issues encountered in ML will soon become problems for agentic coding: overfitting to the spec, Clever Hans shortcuts that don't generalize outside the tests, data leakage, concept drift, etc.
I would also ask: what will be the Keras of agentic coding? What will be the optimal set of high-level abstractions that allow humans to steer codebase 'training' with minimal cognitive overhead?
petteri retweeted
Turns out Musk's starlink satellites are raining out of orbit at a rate of 120 per month and each one produces 30 kilograms of aluminum oxide which eats away at the ozone layer and these oxides have increased 8 fold in the atmosphere since 2016👍
petteri retweeted
A lot of people simply don't like thinking, and are more than happy to delegate any thinking to a tool or a process, usually with poor results. AI will magnify this effect
This Post is from an account that no longer exists. Learn more
petteri retweeted
Every. Single. Week. There's. A. Breakthrough.
Text to video by Meta: makeavideo.studio/
petteri retweeted
Wrote some notes about prompt injection attacks against GPT-3 simonwillison.net/2022/Sep/1…
petteri retweeted
To the brain, reading computer code is not the same as reading language news.mit.edu/2020/brain-read…
petteri retweeted
Over the years, I've tried to keep a list of criteria for deciding on whether to take on a project.
Here are the 9 questions I ask myself now
petteri retweeted
Meet Thatcher Effect.
When faces are upside down, we can’t understand them as well.
We rely on separate parts instead of them as a whole.
The eyes look like eyes and the mouth looks like a mouth, so we don’t notice anything off until she is flipped and the truth is revealed.
petteri retweeted
please, tech people. PLEASE. if you dont have a background working in healthcare/medicine, stop the hackathons and just give money.
the potential to do harm is too high, and designing for medicine is a thing. health communication/tooling is a specialty, not a weekend project.
petteri retweeted
Ubers M3DB, a distributed timeseries database m3db.io/
petteri retweeted
scene from werner herzog's "programming in python" -
"i see the lie in front of me - `import time`, and i am appalled - how can a machine offer such a promise, such a lie, the ability to import time as if it were a simple commodity. once again, the vile snake has bitten me."
