Frontier E.G.

Prev: 2x AI lab MTS
Now: Demon Investing, Shitposting, RetardMaxxing

[Frontier E.G. @igiuldos]

@leothecurious Thinking Machines just announced a 1T param model trained on 45T tokens of video, audio, text and images.

[Frontier E.G. @igiuldos]

@lu__jasper Any specific accounts you recommend?

[Frontier E.G. @igiuldos]

@peterwildeford Greg Brockmans SuperPAC opposed specific AI regulations and candidates. Meanwhile Greg at the same time has been in support of other regulations for multiple years. PS, $250K is a very small amount of money here, it could literally be just 3 employees out of thousands at OpenAI.

[Frontier E.G. @igiuldos]

@cain151714 @DrewHawkswood @NepalAadim Literally in the past 90 days alone we have GPT models solving multiple problems in mathematics that are over 50 years old and considered significant to the field. The cycle double cover conjecture and the unit distance problem which is connected to much of algebraic geometry.

[Frontier E.G. @igiuldos]

Singapore-core

[Frontier E.G. @igiuldos]

@scalingflaw @deredleritt3r Yes there is a successor to 5.5-Pro
Its called 5.6 Sol Pro

[Frontier E.G. @igiuldos]

Should Prometheus be punished for stealing fire from the gods?

[Frontier E.G. @igiuldos]

That’s not what he said, and the critiques in the blog you mentioned does not apply to the work that Alex has referenced working on at Pleias. he’s talking about actual actions he has done of making ethical AI training run. His organization Pleias has made AI training datasets that are entirely comprised of data that Pleias themselves has the licensing rights to, alongside uncopyrighted data like books and texts that are atleast 70 years old or have extended past the life of the author.

And it’s not just some small experiment that is used alongside illegally-obtained data, but they have even already trained language models that have only
ever been trained on this ethically sourced data.

[Frontier E.G. @igiuldos]

@RealAstropulse @seconds_0 Another example here

[Crémieux @cremieuxrecueil]

The great thing about Pangram's false-positive rate is that it's basically 0%.

When looking at PhD dissertations, for example, almost nothing before the advent of LLMs flags as being AI-generated.

Only after AI appears does it start showing up.

[Frontier E.G. @igiuldos]

@RealAstropulse @seconds_0 Now would you like to provide any experimental data of atleast thousands or tens of thousands of essays that shows the contrary? Do you have any at all? Or are your claims of it being basically a “random number generator” just made up?

[Frontier E.G. @igiuldos]

@RealAstropulse @seconds_0 Here directly is an independent experiment done by someone not paid by pangram, that tested Pangram on over 90,000 essays, and it resulted in only 1 false positive and 11 false negatives amongst them.

https://t.co/nftR5uq0qX

[Frontier E.G. @igiuldos]

@RealAstropulse @seconds_0 Demonstrably false to compare it to a random number generator.
The false positive rate of Pangram has been validated to be less than 1 in 10,000.

[Frontier E.G. @igiuldos]

@RealAstropulse @seconds_0 Fact: Pangram has been shown to have a false positive rate of less than 1 in 10,000, including in independent third party tests that have used pangram against tens of thousands of pre-chatgpt essays.

[Frontier E.G. @igiuldos]

The SF ABGs have opened Pandora’s box.

[Milo Smith @mil000]

the introduction bit has spread to the white women. when will it end

[Frontier E.G. @igiuldos]

@rackSpreader1 Yes

[Frontier E.G. @igiuldos]

@moonchild23580 Marcus Aurelius

[Frontier E.G. @igiuldos]

@quionie Any such accounts you recommend to follow?

[Frontier E.G. @igiuldos]

Post-ASI foods will go beyond your wildest imagination, spices and seasonings made from molecular structures that are alien to any salt, pepper or saffron esque compounds that have ever touched your tongue.

[Frontier E.G. @igiuldos]

@blknoiz06 Intel
Rationale; Intel is just short for intelligence. If you’re not long intelligence then wtf are you even doing in 2026?
(It’s actually a portmanteau of integrated electronics)

[Frontier E.G. @igiuldos]

People seem to conveniently forget all the critics and midwits that were screaming from the rooftops on social media and articles, about how Amazon is essentially a ponzi scheme in their view due to going 10+ years without a profit and constantly reinvesting everything they made.