I was busy still with Christmas vacations so I missed these news. Wow, Library Genesis is mentioned.
Jan 9, 2025 Meta Secretly Trained Its AI on a Notorious Piracy Database, > Newly Unredacted Court Docs Reveal
Meta just lost a major fight in its ongoing legal battle with a group of authors suing the company for copyright infringement over how it trained its artificial intelligence models. Against the company’s wishes, a court unredacted information alleging that Meta used Library Genesis (LibGen), a notorious so-called shadow library of pirated books that originated in Russia, to help train its generative AI language models.
The Boss said torrents are cool, but only for Big Tech. If you’re alone at home, never torrent! Copyright laws only apply to (poor) individuals, corporations are exempt!
Before these documents were made public, Meta previously disclosed in a research paper that it had trained its Llama large language model on portions of Books3, a dataset of around 196,000 books scraped from the internet. It had not previously publicly indicated, however, that it had torrented data directly from LibGen.
These newly unredacted documents reveal exchanges between Meta employees unearthed in the discovery process, like a Meta engineer telling a colleague that they hesitated to access LibGen data because “torrenting from a [Meta-owned] corporate laptop doesn’t feel right ”. They also allege that internal discussions about using LibGen data were escalated to Meta CEO Mark Zuckerberg (referred to as “MZ” in the memo handed over during discovery) and that Meta’s AI team was “approved to use” the pirated material.
Very interesting article produced by a Microsoft research team (Tools for thought) to be presented on the Conference on Human Factors in Computing Systems next April.
The rise of Generative AI (GenAI) in knowledge workflows raises questions about its impact on critical thinking skills and practices. We survey 319 knowledge workers to investigate 1) when and how they perceive the enaction of critical thinking when using GenAI, and 2) when and why GenAI affects their effort to do so. Participants shared 936 first-hand examples of using GenAI in work tasks.
Quantitatively, when considering both task- and user-specific factors, a user’s task-specific self-confidence and confidence in GenAI are predictive of whether critical thinking is enacted and the effort of doing so in GenAI-assisted tasks. Specifically, higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking.
Qualitatively, GenAI shifts the nature of critical thinking toward information verification, response integration, and task stewardship.
Our insights reveal new design challenges and opportunities for developing GenAI tools for knowledge work.
So, in the year of the Lord 2025, confidence in AI results only reveals low self-confidence. I’m a dumbhead, anyway I’ve learned that high self-confidence is important in the corporate world.
For good or bad, that’s science when people is studied.
The purpose of this research is to make a better AI. An AI that fills our gaps in skills, not one that makes us dumber. So, a bit of bias and minor errors are not that critical. It’s not about finding the truth but guiding the development of AI.
What matters today is that the people selling AI like a miracle cure are just a bunch of carnival barkers.
Well, the study was performed by Microsoft and as a public, for-profit-organization, its objective is to make more profit. My point is that if corporations are left to their own devices to regulate and amend AI to make it “better” for the society, it will end up being “better” for the bottom-line.
MedeaFleecestealer’s link above is a case in point. Without regulation, we’ll likely end up dumber.
“Vance just dumped water all over that. [It] was like, ‘Yeah, that’s cute. But guess what? You know you’re actually not the ones who are making the calls here. It’s us.’” said McBride.
AI is a bit like sci-fi were robots and the “future” are just tools to explore the present and the human condition. From this perspective, AI could be used as a mirror to know and understand ourselves. AI has flaws, but in spite of them or even due to them, it’s an interesting research tool.
Different things. Some are as simple as asking questions about German Grammar and getting an answer e.g.:
Key Takeaways and Common Mistakes to Avoid
Feminine Nominativ and Akkusativ Look the Same: The biggest initial hurdle is that die and eine are used for both Nominativ and Akkusativ feminine nouns. You must rely on the verb and the sentence structure to determine the case.
Learn Verb-Case Combinations: Certain verbs always take a specific case. For example:
helfen (to help) + Dativ
danken (to thank) + Dativ
gefallen (to please/like) + Dativ
sehen (to see) + Akkusativ
haben (to have) + Akkusativ
brauchen (to need) + Akkusativ
lieben (to love)+ AkkusativeMemorizing these common verb-case pairings is crucial . A good German dictionary will tell you which case a verb takes.
Prepositions are Key: Prepositions (words like “in,” “on,” “to,” “with,” “for”) often govern a specific case. Some prepositions always take the Dativ, some always take the Akkusativ, and some can take either depending on whether they indicate location (Dativ) or direction/movement (Akkusativ).
Dativ Prepositions (some common ones):aus (out of, from), bei (at, near), mit (with), nach (after, to), seit (since, for), von (from, of), zu (to)
Akkusativ Prepositions (some common ones):durch (through), für (for), gegen (against), ohne (without), um (around)
Two-Way Prepositions (Dativ for location, Akkusativ for direction):an (at, on), auf (on), hinter (behind), in (in, into), neben (beside), über (over, above), unter (under, below), vor (in front of), zwischen (between)
Practice, Practice, Practice: The best way to master cases is through consistent practice. Read German texts, listen to German audio, and try to construct your own sentences. Don’t be afraid to make mistakes – that’s how you learn!
Word Order is (Somewhat) Flexible: While German word order is generally more fixed than English, the cases allow for some flexibility. For example, you could say:
Ich gebe der Frau das Buch. (I give the woman the book.)
Der Frau gebe ich das Buch. (To the woman, I give the book.) – Emphasizes the recipient.
Das Buch gebe ich der Frau. (The book, I give to the woman.) – Emphasizes the book.The cases clearly show who is doing what to whom, even if the word order changes.
This is a foundational concept, so take your time, review it often, and practice regularly. Good luck with your German learning!
Another way is to practice speaking and listening - I set up a chatbot and I speak and that gets transcribed into German and the chatbot then responds and speaks back in German. This way, I can role-play certain scenarios or practice discussing certain topics.
Afterwards I ask the LLM to review the conversation and comment on mistakes and what items should be worked on etc.
One difficulty with LLMs is that you can never be 100% sure whether what it says is correct, but when I’m only trying to get to B1, the AI is likely ‘good enough’.
So, I was enquiring on Lucy Calkins and her disastrous “theory” on teaching reading. Turns out she was a major factor influencing the teaching style in 8 states (NY, NJ, MA, CT, IL, MN, CA, and OR) (with catastrophic consequences). So I asked the free version of ChatGpt
“in a map with the lower 48 states in their correct locations and positions, highlight those 8 states”
and this is the “geographically accurate map” I got:
off topic of AI, but this way of teaching cases (not specific to AI, but actually explained like that in English books) must be very difficult. As a native speaker of a language with cases I had to learn how to name them in given sentences, though as every native speaker I used them correctly before going to school. The method we were taught was to pretend you can’t read, you didn’t hear, the word and ask for it. The question word you use to ask for it tells you the case. It works for me quite naturally in German as well.
”. They also allege that internal discussions about using LibGen data were escalated to Meta CEO Mark Zuckerberg (referred to as “MZ” in the memo handed over during discovery) and that Meta’s AI team was “approved to use” the pirated material.
