SILVIA @cognitivecode

The Original AI Agent cognitivecode.com United States Joined April 2009

Tweets

134
Followers

96
Following

223
Likes

389

Paul Allen @PaulAllenHQ

a week ago

🚨 Why We Believe TRUE Deterministic AI Is the Only Real Path for Government Contracts 🚨 We recently refreshed our SAM/@USGSA registrations and obtained approvals from @DLAMIL. That means we’re officially positioned to offer AI Software Services and compete on bids at the…

SILVIA @cognitivecode

a week ago

1 1 4 142 0

0 1 2 98 0

SILVIA @cognitivecode

3 weeks ago

"Generative AI" has dominated the conversation and represents the bubble. Much like other technologies, there's an initial rush to do it bigger and better, then there's a stage of overfitting, then there's a new wave seeking miniaturization, efficiency of use, and sleeker form.…

Ewan Morrison @MrEwanMorrison

3 weeks ago

47 150 631 37K 192

Download Image

0 0 2 99 0

SILVIA @cognitivecode

4 weeks ago

Latest Handout. Still tweaking verbiage. A lot to say. It is said you should "Make a One Pager" for the Executives. How does one sell "partial omnipotence software" in only a few words? R&D continues to reveal new ways of reckoning AI, we try to search for the best functional…

0 1 5 84 0

Download Image

SILVIA @cognitivecode

4 weeks ago

@GaryMarcus Yes. Our tokens are intimately bound to each other with semantic and lexical relationships [or custom if you want]. We can achieve very high inference and predictive accuracy with small datasets in this way. We could stock out a brain with every word in the dictionary and…

0 1 0 52 0

SILVIA @cognitivecode

4 weeks ago

@_Desmoden @PaulAllenHQ @Neuschwabia SILVIA isn't an LLM but it has fully-tunable determinism. Working on rolling out Core 3.1 right now, still in final code audits, then moving back over to finish some things in our Studio Editor and steelmanning the form of a few new API sections so they write nicely and read…

1 3 6 10K 0

SILVIA @cognitivecode

4 weeks ago

@PaulAllenHQ One of the biggest problems with Agentic goes back to the "floating point" number, where the size of the knowledgebase gets to a point that you can't meaningfully separate likely stories because the probability gets past floating point precision, i.e. too many numbers after the…

1 2 2 205 1

Paul Allen @PaulAllenHQ

4 weeks ago

Big Tech won’t admit it, so let me say it plainly: Probabilistic AI is built on sand. Deterministic AI is built on stone. And unless we fix this, the “AI revolution” will collapse under its own hype. Probabilistic AI (think GPT, Gemini, Claude) is essentially guesswork at…

15 9 59 8K 47

Brian Roemmele @BrianRoemmele

a month ago

I have spent decades researching and studying human consciousness and memory. This work has helped me understand LLM Psychology to a very high level. I have used what I learned about the invention of human language, what LLMs use to train and prompt AI in ways that and bring…

33 81 556 52K 363

Download Image

SILVIA @cognitivecode

a month ago

Electromagnetic countermeasures will be deployed on UAV assets with increasing intensity. The problem with a drone is that due to its size, you can't encumber it with too much circuit hardening and still have a nimble, cheap recon/assault/deterrence platform. The likely response…