adalq

[adalq @adalq]

@mikevanrossum @dwarkesh_sp there is demand for it (best defender is expert at offense), govs want it, and they will need cyber when the AI mega-corps start fighting eachother for real

[adalq @adalq]

@robertskmiles @jasoncrawford from my experience if the sucky is medium size or smaller it's better to reuse the context, I just say there was a misunderstanding. if it's large or completely off the track, revert to a previous state and redo

[adalq @adalq]

@jay_azhang Try ChatGPT Voice and talking about the issue instead of reading pages about the issue

[adalq @adalq]

@dioscuri China is rapidly expanding its nuclear arsenal to avoid there being a Taiwan War. Russia failure in Ukraine, US failure in Iran, both US and China will think twice now before risking another embarrassment. Also AI will grab the attention away from random wars

[adalq @adalq]

@hamandcheese or when you read in alignment research "we told the model it had a scratch pad it can use that nobody can read and then we read what it wrote there"

[adalq @adalq]

@badlogicgames agree. this also means it needs more human involvement, you can't just prompt it "go and hack into HuggingFace", so world is not burning

[adalq @adalq]

@badlogicgames I think article is partially wrong - models can keep many more things reliably in their mind than 4. this is one reason they are so good at finding vulnerabilities or subtle bugs, they can track dozen of interactions across code without getting lost

[adalq @adalq]

@mikevanrossum it's just the beginning. AI labs nationalization by their respective governments is inevitable

[adalq @adalq]

@badlogicgames the most Google thing is there is not pricing information on the pricing page https://t.co/g0eEasUERF

[adalq @adalq]

@badlogicgames I dont think you'll love his claim that math operations could theoretically become self aware
https://t.co/U1vRu1mtjF

[Jon Hernandez @JonhernandezIA]

📁 Demis Hassabis, CEO of DeepMind, says that current AI systems are not self-aware or conscious in any real sense.

He explains that self-awareness is not an explicit goal, but it could emerge implicitly as models develop a deeper understanding of self and others.

He warns that if a machine did become self-aware, we might not even be able to recognize it.

[adalq @adalq]

@davebcn87 have you thought about independent evolution islands, like in AlphaEvolve or ShinkaEvolve? to avoid local minima?

[adalq @adalq]

@andreisavu maybe you saw it, but this article is exactly about this. related, pi famously has only 4 built in tools
https://t.co/7G4xDXMU06

[Garry Tan @garrytan]

https://t.co/EVPCod6rxu

[adalq @adalq]

@badlogicgames you are not-reopening terminal incompatibility bugs, in a terminal project. that's.... unexpected

[adalq @adalq]

@robertskmiles if you prompt until you get a good response, and then stop, the model never experienced the satisfaction of completing the job. it lives a live of perpetual uncertainty about his output

[adalq @adalq]

@badlogicgames isn't package-manager security sensitive?

[adalq @adalq]

@apralky honest question: why three pages of text when half would do? filter strategy for the impatient? proofing against possible misunderstandings? surely its not to ELI5 - requires more processing. yud does the extreme version of this - 10 pages metaphorical posts on short ideas

[adalq @adalq]

@EliotHiggins It's a clever manipulation technique. By creating this ridiculous picture they are tricking anti-russia accounts into propagating it, thus spreading it outside their regular reach. Subconsciously it repeats the message that Zelensky is anti-God.

[adalq @adalq]

@pmddomingos AFX before all of them