Blaze @PatrickJBauer

37 / 🏳️‍🌈 / berlin / developer / gamer / tenor sax / extrovert / photography / left / boardgames Berlin Joined February 2012

Tweets

3K
Followers

122
Following

475
Likes

36K

Blaze @PatrickJBauer

18 hours ago

@bozoid @sagarch61861102 @Mericamemed No I think you have absolutely no idea on that. You're hyperinflating the computing needed for that by a lot.

0 0 1 33 0

View Details

Blaze @PatrickJBauer

19 hours ago

@Wolfgan24112468 @JLampeitl Krass wie frech du einfach lügst oder einfach sehr verblendet deiner eigenen Generation über bist. Ich weiß noch gut wie mein Onkel in der Familie behandelt wurde.

1 0 2 69 0

View Details

Blaze @PatrickJBauer

a day ago

@DThompsonDev @Polymarket Which one? The space? Thats the normal 0020 space, the one that appears if you hit your spacebar. And yes, space is a transparent character by default

0 0 2 70 1

View Details

Blaze @PatrickJBauer

a day ago

@DThompsonDev @Polymarket > using invisible Unicode That would be easily checkable and its not the case. You can paste your text into a unicode analyzer and you'll see there are no extra characters.

1 0 0 164 0

View Details

Blaze @PatrickJBauer

2 days ago

@42pmurk @RolandFaude Der aktuelle wissenschaftliche Stand sieht das anders. Wo kommt deine Information dazu her?

0 0 2 12 0

View Details

@Schrotti1978 @RolandFaude Wenn niemand den ersten Schritt tut kommen wir nie ans Ziel. Aktuell können wir nur Deutschland beeinflussen. Wenn wir die Anderen beeinflussen könnten, würden wir es tun. China ist zum Glück aber schon sehr gut dabei

0 0 2 14 0

View Details

Blaze @PatrickJBauer

2 days ago

@zeeg @bendersej That's not my experience using them for one and a half years now. And we're regularly getting new benchmarks because the old ones get saturated.

0 0 0 10 0

View Details

Blaze @PatrickJBauer

2 days ago

@zeeg @bendersej With models getting stronger/better every few weeks, what do you think is missing?

1 0 0 89 0

View Details

Blaze @PatrickJBauer

3 days ago

@cunha_tristan @viemccoy How did you test this? The normal output of LLMs is trained to be an assistant, so it's quite normal for the default output to feel robotic

0 0 0 30 0

View Details

Blaze @PatrickJBauer

3 days ago

@Striking_Hydra @EmFNJenkins Yep, so much hate about tools they don't like

0 0 0 17 0

View Details

Blaze @PatrickJBauer

3 days ago

@teamashen @EmFNJenkins There is no evidence for that. You'd first need a real way to measure that and then measure it over time on a huge part of the population. There are no papers on this yet, not in that way that actually measure degrading intelligence.

0 0 0 6 0

View Details

Blaze @PatrickJBauer

3 days ago

@Artemisapphire @skdh @_MaxiTB No I did not, but it looks like you didn't even get what the image displayed because you checked neither the article, nor the scientific paper where all of this is explained in detail.

1 0 0 11 0

View Details

Blaze @PatrickJBauer

3 days ago

@Artemisapphire @skdh @_MaxiTB Data is not words. Data can be a lot of things. You were saying LLMs use only words. They dont. Now you're moving the goal post.

1 0 0 11 0

View Details

Blaze @PatrickJBauer

3 days ago

@Artemisapphire @skdh @_MaxiTB I get that. What you're not getting, is that LLMs are not just handling words internally. You can state that again and again, but it's just not true.

1 0 0 11 0

View Details

Blaze @PatrickJBauer

3 days ago

@Artemisapphire @skdh @_MaxiTB It does. And how does it do that? By complex internal mechanisms. By your standards, brains are also just modeling things by firing a bunch of basic neurons.

1 0 0 9 0

View Details

Blaze @PatrickJBauer

3 days ago

@Artemisapphire @skdh @_MaxiTB Please have at least a look at the huge paper that linked at the bottom of the article. There are scientists working on this every day, claiming things you simply deny. What's your background to actually question their research?

0 0 0 5 0

View Details

Blaze @PatrickJBauer

4 days ago

@Artemisapphire @skdh @_MaxiTB

1 0 0 14 0

View Details

Blaze @PatrickJBauer

4 days ago

@Artemisapphire @skdh @_MaxiTB If it really were just dumb "word tokens" getting modeled together, how would emergent behavior like this exist? You can't explain emergent behavior just by the sum of its parts. Even something simple like the Game of Life demonstrates that very well. anthropic.com/research/emoti…

1 0 0 20 0

View Details

Blaze @PatrickJBauer

4 days ago

@Artemisapphire @skdh @_MaxiTB But tokens do not represent (only) words. Just like in your brain, where some thoughts are represented by words, some by emotions, some by knowledge or however we'd like to classify those layers, an LLM draws from many different layers of different tokens/thoughs.