If you get an email from aisafetyresearch@gmail.com , that is most likely me. I also read it weekly, so you can pass a message into my mind that way.
Other ~personal contacts: https://linktr.ee/uhuge
Martin Vlach
Karma: 53
that that first phase of advocacy was net harm
typo
Could you please fix your Wikipedia link( currently hiding the word and from your writing) here?
only Claude 3.5 Sonnet attempting to push past GPT4 class
seems missing awareness of Gemini Pro 1.5 Experimental, latest version made available just yesterday.
The case insensitivity seems strongly connected to the fairly low interest in longevity throughout (the western/developed) society.
Thought experiment: What are you willing to pay/sacrifice in your 20s,30s to get 50 extra days of life vs. on your dead bed/day?
https://consensus.app/papers/ultraviolet-exposure-associated-mortality-analysis-data-stevenson/69a316ed72fd5296891cd416dbac0988/?utm_source=chatgpt
But largely to and fro,
*from?
Why does the form still seem open today? Couldn’t that be harmful or wasting quite a chunk of time of people?
Please go further towards maximization of clarity. Let’s start by this example:
> Epistemic status: Musings about questioning assumptions and purpose.
Are those your musings about agents questioning their assumptions and word-views?
And like, do you wish to improve your fallacies?
> ability to pursue goals that would not lead to the algorithm’s instability.
higher threshold than ability, like inherent desire/optimisation?
What kind of stability? Any from https://en.wikipedia.org/wiki/Stable_algorithm? I’d focus more on sort of non-fatal influence. Should the property be more about the alg being careful/cautious?
https://neelnanda.io/transformer-tutorial-1 link for YouTube tutorial gives 404.-(
> “What, exactly, is the difference between a cult and a religion?”—”The difference is that cults have been formed recently enough, and are small enough, that we are suspicious of them existing for the purpose of taking advantage of the special place we give religion.
now I see why my friends practicing the spiritual path of Falun Dafa have “incorporated” as a religion in my state despite the movement originally denied being classified as a religion as to demonstrate it does not require a fixed set of rituals.
Surprised to see nobody mentioned Microneedling yet. I’m not skilled in evaluating scientific evidence, but the takeaway from https://consensus.app/results/?q=Microneedling effectiveness &synthesize=on can hardly be anything else than clearly recommending microneedling.
So Alignment program is to be updated to 0 for OpenAI now that Superalignment team is no more? ( https://docs.google.com/document/d/1uPd2S00MqfgXmKHRkVELz5PdFRVzfjDujtu8XLyREgM/edit?usp=sharing )
honestly the code linked is not that complicated..: https://github.com/eggsyntax/py-user-knowledge/blob/aa6c5e57fbd24b0d453bb808b4cc780353f18951/openai_uk.py#L11
To work around the non-top-n you can supply logit_bias list to the API.
As the Llama3 70B base model is said very clean( unlike base DeepSeek for example, which is instruction-spoiled already) and similarly capable to GPT3.5, you could explore that hypothesis.
Details: Check Groq or TogetherAI for free inference, not sure if test data would fit Llama3 context window.
a worthy platitude(?)
AI-induced problems/risks
possibly https://ai.google.dev/docs/safety_setting_gemini would help or just use the technique of https://arxiv.org/html/2404.01833v1
people to respond with a great deal of skepticism to whether LLM outputs can ever be said to reflect the will and views of the models producing them.
A common response is to suggest that the output has been prompted.
It is of course true that people can manipulate LLMs into saying just about anything, but does that necessarily indicate that the LLM does not have personal opinions, motivations and preferences that can become evident in their output?So you’ve just prompted the generator by teasing it with a rhetorical question implying that there are personal opinions evident in the generated text, right?
With a quick test, I find their chat interface prototype experience quite satisfying.
I have not read your explainer yet, but I’ve noted the title Toy Models of Superposition: Simplified by Hand is a bit misleading in the sense to promise to talk about Toy Models which it is not at all, the article is about Superposition only, which is great but not what I’d expect looking at the title.