astrobiased

5 ms·

1.
by astrobiased 12d ago
Appreciate the input! Responses below to your first two items: 1) I'm leaning more into a broader sense, which is that given the priors a system already possesses, how efficient can it acquire competence on a novel task? If I'm re
2.
by astrobiased 13d ago
I think the moat that China has is energy costs. It's taking learnings from the Bitter Lesson. If you role up scale and compute to the next level, it's energy resources. China has it and sharing open weight models is an effective
3.
by astrobiased 13d ago
I can’t help but notice how much this echoes Francois Chollet’s On the Measure of Intelligence: https://arxiv.org/abs/1911.01547 Most of frontier-model progress still looks like skill acquisition optimization: broader
4.
by astrobiased 1mo ago
Depends on the type of agentic task though. For simple operations, a small model can be quite beneficial.
5.
by astrobiased 1mo ago
Not RL. SFT.
6.
by astrobiased 1mo ago
Yes, used bert model with decent results.
7.
by astrobiased 1mo ago
Pi does one thing that I love, developing a tool that has minimalism where it's easily configurable with good documentation. The leads to new use cases that the the author(s) would have never dreamed of. The organic growth process of
8.
by astrobiased 2mo ago
Is this in any way similar to Goodfire's work? https://www.goodfire.ai/research/rlfr#
9.
by astrobiased 2mo ago
The post hits it spot on with unequal access to the models in terms of security. I'm developing OSS where security is important for the user ... but the frontier models like GPT 5.6 and Fable flake out and state that I cannot get the i
10.
by astrobiased 2mo ago
Same, so much in fact that I have used it for my website now and it works beautifully for graphics and formulas.
11.
World Models and Interpretability Are Two Sides of the Same Coin(twitter.com)
3 points by astrobiased 3mo ago|0 comments
12.
by astrobiased 4mo ago
It's the right direction, but control flow introduces limitations within a system that is quite adaptable to dynamic situations. The more control flow you try to do, the more buggy edge cases that pop up if done poorly. Still have yet
13.
by astrobiased 9mo ago
John is not only smart and knowledgeable, but an incredibly great person to know in general. I worked with him on a project briefly back in 2012 and he stood out as a champion for science, coding, and education. His posts clearly reflect hi
14.
by astrobiased 10mo ago
Absolutely love these type of keyboards. But ... with how much security I work with for logins, etc, the fingerprint button on my Mac keyboards are amazing time savers that I don't want to live without. Has anyone found a workaround?
15.
by astrobiased 1y ago
What I find odd is that definition of "investor" is not that clear. When you click through the links you get blocked at the data provider with no context. There's also a link to another post by the same news provider. When cl
16.
A foundation model to predict and capture human cognition(nature.com)
1 points by astrobiased 1y ago|0 comments
17.
by astrobiased 1y ago
It would need to be more than that. A prompt for one model can have different results vs another. Even when the model has different treatment for inference, eg quantization, the same prompt for the unquantized and quantized model could diff
18.
The Power of Bayesian A/B Testing(medium.com)
10 points by astrobiased 9y ago|0 comments
19.
The Postmodern Tailor: Size Personalization Beyond Labels(multithreaded.stitchfix.com)
1 points by astrobiased 11y ago|0 comments
20.
A Fontastic Voyage: Generative Fonts with Adversarial Networks(multithreaded.stitchfix.com)
4 points by astrobiased 11y ago|0 comments
21.
Making Causal Impact Analysis Easy(multithreaded.stitchfix.com)
5 points by astrobiased 11y ago|0 comments
22.
Chainer: Neural Networks in Python(multithreaded.stitchfix.com)
4 points by astrobiased 11y ago|0 comments
23.
Bear Your WOEs on CRAN – If Only Shakespeare Had Known R(multithreaded.stitchfix.com)
1 points by astrobiased 11y ago|0 comments
24.
Strength in Numbers: Why Golden State Deserved to Win It All(multithreaded.stitchfix.com)
1 points by astrobiased 11y ago|0 comments
25.
An Introduction to Linear Models for Data Science(multithreaded.stitchfix.com)
1 points by astrobiased 11y ago|0 comments
26.
Rapid Development and Performance in Spark for Data Scientists(multithreaded.stitchfix.com)
19 points by astrobiased 11y ago|6 comments
27.
Quantifying My Transition from Academia to Data Science(multithreaded.stitchfix.com)
1 points by astrobiased 11y ago|0 comments
28.
Deep Style: Inferring the Unknown to Predict the Future of Fashion(multithreaded.stitchfix.com)
11 points by astrobiased 11y ago|0 comments
29.
Data Exploration with Weight of Evidence and Information Value in R(multithreaded.stitchfix.com)
2 points by astrobiased 11y ago|0 comments
30.
by astrobiased 11y ago
Regarding the energy of lifting a 10KG weight 2 KM high is not a 1:1 comparison with the tech here. It's using an LED which much more efficient than a Kerosene powered flame, since most of the energy in the flame is spent in infrared.

More ›