Was es am 20.06.2023 nicht in den Newsletter geschafft hat
Aber dennoch vielleicht spannend für dich ist
🟥Wie KI der Ukraine auf dem Schlachtfeld hilft
🟥Was ChatGPT alles schon gelesen hat
🟥Die Grammys skizzieren neue Regeln für den Einsatz von KI
🟥Die Bedrohung der menschlichen Existenz durch KI ist eine "absurde" Ablenkung von den tatsächlichen Risiken
🟥Was wird mit GPT-{n} passieren, wenn LLMs einen Großteil der online gefundenen Sprache beisteuern?
🟥🟩KI-Algorithmen finden Medikamente, die die Alterung bekämpfen könnten
🟥KI-Startup Cohere, jetzt mit über $2,1 Mrd. bewertet, erhält $270 Mio.
🟥Wird die KI das neue McKinsey?
This was an impressive paper with some big implications:
GPT-4 was given 4,550 novel questions representing the entire “MIT Mathematics and EECS undergraduate curriculum, including problem sets, midterms, and final exams.”
These questions were not in the training data.
Out of the box, GPT-4 scored 90%. With good prompts, it scored 100%. Good prompts included the technique of asking the AI “to identify experts in the field, generate answers as if the experts wrote them, and combine the experts’ answers by collaborative decision-making.” It is absolutely nuts that this approach works.
Last week, our LinkedIn post was viewed over 4 million times. Our secret?
Crafting an irresistible hook. Here’s how we employed the power of Zapier & Excel to make it happen:
We populated column #1 with viral Twitter/LinkedIn posts.
In column #2, we jotted down bullet points for new posts.
Finally, we sent a "Zap" to ChatGPT, instructing it to generate hooks in column #3 for our fresh ideas using the successful formats from column #1.
The result: an endless list of engaging hooks inspired by the most successful content on the internet.