Ayrton Senna

[Ayrton Senna @SierraAlisa]

@friedberg everyone in agreement that AI is the culmination of humanity’s ingenuity, it stands to reason that these models will ultimately need to be open source and cannot belong to one person, company or country.

[Ayrton Senna @SierraAlisa]

@raw_works Would you mind posting total token (input/output) and cost estimations for each run? Thank you for posting... super interesting!

[Ayrton Senna @SierraAlisa]

@natolambert Is it possible that scaling laws on larger models hold true for problems that utilize longer context windows? for example a problem with 200k tokens, might not scale well with a 0.6b params but might scale well with 14b params

[Ayrton Senna @SierraAlisa]

@josancamon19 @josancamon19 love this study, thank you for sharing. Is it possible that scaling laws on larger models hold true for problems that utilize longer context windows? for example a problem with 200k tokens, might not scale well with a 0.6b params but might scale well with 14b params

[Ayrton Senna @SierraAlisa]

@Citrini7 Atlanta

[Ayrton Senna @SierraAlisa]

@Jason @TOMFORD @Zegna @dieworkwear Suitsupply

[Ayrton Senna @SierraAlisa]

@Rainmaker1973 Gattaca

[Ayrton Senna @SierraAlisa]

@crypt0x_0 Flowers for Algernon

[Ayrton Senna @SierraAlisa]

@ikoukas @TheRealAdamG Let’s see the next o-series models. My intuition is that these models cannot inference beyond the logic they have already been exposed to - ie they cannot innovate, this is why they struggle with open ensued problems imo. That is not to say most human task will not be automated..

[Ayrton Senna @SierraAlisa]

@ikoukas @TheRealAdamG Maybe.. but make an extreme hypothetical scenario: 100 param model, regardless of time inference compute won’t solve any problem. There will always be an upper limit. Also I think AIME problems are a bad benchmark - I would like to see these model tested on open ended questions.

[Ayrton Senna @SierraAlisa]

We already tried in the uk and it didn’t work.

Research conducted by LSE found that help to buy increased house prices in London by 8%, and boosted developers’ revenues by 57% in the process. The policy “led to higher new-build prices but had no discernible effect on construction volumes”, effects that are “arguably contrary to the policy’s objectives”.

https://t.co/lXmF2bor1X

[Ayrton Senna @SierraAlisa]

In the UK, research conducted by LSE found that help to buy increased house prices in London by 8%, and boosted developers’ revenues by 57% in the process. The policy “led to higher new-build prices but had no discernible effect on construction volumes”, effects that are “arguably contrary to the policy’s objectives”.

https://t.co/lXmF2bor1X

[Ayrton Senna @SierraAlisa]

@mcuban

In the UK, research conducted by LSE found that help to buy increased house prices in London by 8%, and boosted developers’ revenues by 57% in the process. The policy “led to higher new-build prices but had no discernible effect on construction volumes”, effects that are “arguably contrary to the policy’s objectives”.

https://t.co/lXmF2bor1X

[Ayrton Senna @SierraAlisa]

@EranMalach Amazing work @EranMalach! Any intuition on the Elo performance based on the size of the training dataset?

[Ayrton Senna @SierraAlisa]

@OpenAI @AnthropicAI @sama @elonmusk can we get a live fact checker for the presidential debate?

[Ayrton Senna @SierraAlisa]

@OpenAI @AnthropicAI @sama @elonmusk can we get a live fact checker for the presidential debate?

[Ayrton Senna @SierraAlisa]

@nntaleb @nntaleb I know you are not a fan of gpt, but I think this is pretty neat

[Ayrton Senna @SierraAlisa]

@chamath https://t.co/LwbgYXaWc7
All-in podcast e117

Love how 4 months ago this was a great business model

Chamath: “Adyen … how increadibly profitable this business is, and how much operating leverage they have, which means their opex is relatively constraint”

[Ayrton Senna @SierraAlisa]

@cliftonsellers I’d love to see your AI guidebook!