I'm going to the Codex hackathon Vienna, and our team is looking for one more. Are you in?
📅 Sat, April 18
Winner team gets $15,000 in OpenAI API credits
8
Check out my talk on AI agents for CMS at Neos Con today at 15:00
youtube.com/watch?v=RhV84mjR…
25
If you could have a Neos theme based on your wishes, how would it look like?
rolandschtz.typeform.com/to/…
The survey goes till this Saturday!
27
Building a long-term-memory “brain” for my personal assistants.
My first attempt with LangChain and Pinecone wasn't satisfying. Has anyone tried Mem for this?
1
1
54
Roland Schütz retweeted
Die Wiener Zeitung hat rund eine Mio. Web-Seiten auf tagblatt-wienerzeitung.at/ verschoben.
Aber statt die alten URLs umzuleiten, gibt's jetzt überall Broken Links (u.a. über 800 Wikipedia-Quellen) und man wird aufgefordert, Links zu löschen.
🙈🤡💩 #insideibiza #wzdaily #umlautö
Thesis: Real-time AI video generation undermines our current verification methods of trust.
Antithesis: They were already compromised; internet scams became too large a section, and personal trust will become more important.
6
Thesis: Fake AI-generated photos pose a significant risk of misinformation.
Antithesis: This was already possible long ago with Photoshop, and everyone on Instagram uses these techniques. Please just become more aware of it, because it's cheaper/faster.
3
Thesis: AI causes a new problem of misinformation.
Antithesis: Most of Twitter was already run by bots; fake news existed long before. We have effective filtering mechanisms in place, like spam filters.
5
What if you had someone who researches your competitors and gives you a detailed outline before you start writing an article?
Just released, AI Sidekick does the boring SEO work for you. It analyzes your competition and crafts a comprehensive outline.
ai-sidekick.app/en/product/f…
15
The adoption of AI moves fast. Stack Overflows traffic dropped by 18 percent.
t3n.de/news/programmierer-ch…
50% of A.I. researchers believe there's a 10% or greater chance that humans go extinct from our inability to control AI
youtube.com/watch?v=xoVJKj8l…, original source aiimpacts.org/2022-expert-su…
23
🧠 I've been diving deep into the world of AI and ChatGPT, and it's interesting to see how the GPT-4 API model differs from ChatGPT. An important one is Chain-of-Thought! 🤖💬
🔗 So, what is Chain-of-Thought? It's a method used by ChatGPT to think step-by-step out loud🌐🔀
5
100
🎉 Isn't it incredible how AI prompt engineering keeps evolving? 😃🚀 Currently, I'm looking into self-criticism prompting and self-refining selfrefine.info/, in the context of improving marketing content. Have you played with it already? Share your thoughts below! 💭👇
19
The big difference, ChatGPT is reasoning out loud and therefore finds the correct results.
💡 It's based on this research [1]. You can make the API do the same by just adding "Think step-by-step." and it gets it right.🎯
arxiv.org/abs/2201.11903
10
✅ ChatGPT:
After the two switches, the order of people sitting is as follows:
Initially: Alice, Brett, Chris, Doris
Brett and Doris switch places: Alice, Doris, Chris, Brett
Alice and Doris switch places: Doris, Brett, Chris, Alice
The final order: Doris, Brett, Chris, Alice.
18
🤔 Picture this: Four people are sitting, from left to right Alice, Brett, Chris, Doris. Brett and Doris switch places. Then Alice and Doris switch places. How are they sitting? 🎲
❌ The GPT-4 API just returns "Chris Brett Alice Doris", which is false.
34


