Dominik Lukes
@techczech
Followers
2K
Following
4K
Media
2K
Statuses
15K
Schemas and propositions about language models of all kinds: https://t.co/GU07uzaA4F
UK
Joined April 2009
I'm sure others have had this hot take before but I'm calling it now. The one unexpected consequence of AI will be the death of (most) software. Instead of thinking about an app and searching if someone made it, you just describe it to an LLM and there it is!
2
0
3
Fully agree. This complete delusional insistence on human accountability makes no sense. It is as if those insisting on it have never actually lived in a world where decisions are being made by humans!
1
0
1
Even GPT-5 Thinking is finding the leap year and the Unix timestamp tedious...
0
0
1
This is really important. We always should see the prompt. We need context transparency.
By popular demand, we're rolling out another fix today: You can now view the customization prompt used for your studio outputs! 🥳 Whether it's audio or video overviews, flashcards, or quizzes, just tap the three dots by the output to revisit your creative genius.
0
1
2
Hey everyone, remember the last time wresting away the "means of production" from the "profit-maximising interests" led to loads of democratization all over the place!
0
0
0
From my 2007 slides on Why (not) to do discourse analysis - very much applies outside of education research, as well.
1
0
0
Complaining about the "ownership class" and charging $25 for a PDF seems a bit of a paradox to me.
0
0
1
The political bias of LLMs is most reminiscent of that defining YA fiction.
0
0
0
I cannot stress this enough. The people most loudly complaining about inequality are the ones causing most of it. In the way they talk, judge others, and exclude others through defaults they unthinkingly but irrevocably optimised to fit only those most like them.
I think the hyper-focus on the very small number of billionaires (who to be clear I do think should aim to donate most of their wealth) has mostly harmed people's ability to think about the real problems of economic inequality. Many problems are caused by middle and upper middle
0
0
2
People who say only LLMs ever hallucinate should read the Checklist Manifesto.
0
0
0
Felt tip marker to avoid surgery on the wrong body part is the original context engineering.
0
0
1
For all its flaws, Wikipedia is the crowning achievement of the human web An AI-assisted fork of Wikipedia could be interesting, but only if it embraced transparency, had access to scholarly work & if the AI was used to help humans in making information better, not override them
At the center of everything is Wikipedia. 🔎Wikipedia articles appear in 67%-84% of all search engine results & most info boxes 🔎Wikipedia generates 43M clicks to external websites a month 🔎Wikipedia is a major component of AI training data, including The Pile training set
52
43
506
If you are writing a paper about the importance of human creativity and accuracy in the face of these unreliable Large Language Models, start by making sure you are not producing a cliche-ridden bowl of word salad.
0
0
1
If you write something about Large Language Models but use the term GenAI, you are almost certainly wrong because you did not take enough time to understand either.
0
0
0
This is why reading the news is bad for you. I stopped doing it long ago and you should, too. The idea that the "traditional" media is somehow superior to the noise of the internet is laughable.
0
0
2
Let's replace "critical thinking" with "hermeneutic thinking". Don't try to uncover a conspiracy, try to uncover meaning. Instead of treating every claim as a potential lie, treat it as text — something to interpret, not expose.
0
0
1
Ok algorithm, why on earth do you think I have any interest in this? Can you please track me better? I don't care about privacy if I it lets me avoid seeing football results on my screen without asking for them.
0
0
0
The only successful learners are self-learners. Teachers don't matter unless they act as a resource for somebody learning. If I'm in the middle of learning, I will get enormous value out of a boring lecture. If I'm just skimming the surface, no motivational tricks will help.
0
0
2