Batuhan4

Yayınlarda hep dedim: kriptocular her şeyi gördü. Başka sektörde rekabet var sananlar bir kere kriptoda bir şey üretmeyi denesin, tüm dünyayla yarışmak neymiş anlar. Kriptoda herkes defalarca dolandırıldı. Ben bir ara günde ortalama 100 rug yiyordum. Milyon farklı tail senaryodan patlamış insanlar bunlar. Yapay zeka o yüzden kriptoya bu kadar sert çöktü; yetenekli adamları tek tek topladılar. bkz. Hermes ekibi @NousResearch :D

para piyasası fonu adı altında fon çıkarıp topladıkları paraları kendi şirketlerine yüksek faizli kredi olarak verip fonun gelirini yüksek gösteren bir fonda, herkesin parasını çekmeye çalıştığı bir senaryoda neler yaşanır sizce… böyle bir fonun ismi ppf olmamalı

Türkiye’nin ilk agentic trading hackathon’u başladı! Youtube’dan canlı takip edebilirsiniz👇🏻 youtube.com/live/NvKa0SkVDLQ…

Mesela tp2 fonu, halka açık, 1000 liralık bile alabiliyorsun istediğin bankadan, sende al!! Ne diye kotuluyorsun?!

Anthropic periodically does things I strongly disagree with, but specifically in the case of text watermarks, I didn’t really understand why all the criticism was directed at them. First, this is not Anthropic’s initiative, but an attempt to comply with the EU AI Act. For some reason, there are almost no questions about the law itself. Second, as far as I understand, this practically does not affect output quality. The technology itself is not new, Google has been using SynthID-Text for quite a while, and Claude now uses a version of the same approach. I have no questions about why watermarks are needed for generated images, audio and video. There, provenance really is crucial, because the probability of fakes is very high. But when it comes to text, the logic seems less obvious to me. A watermark can show that an LLM was involved in writing the text. But what exactly does that tell us about the text itself? I can come up with the entire idea myself and ask an LLM only to translate it. I can give the model the theses, structure, and arguments, and it will simply format them. I can actually ask it to come up with everything from scratch. But for the watermark, all of these are just different degrees of the same phenomenon: AI was involved. And all of this will be viewed through the lens of even greater distrust and bias toward AI-generated texts, because the watermark will be present in every case. But this says nothing about who the source of the idea is, whether the text is true or who is responsible for it. So I don’t really understand what exact problem this initiative is trying to solve. Because the provenance of text and the provenance of an image or video are far from being the same thing.

If your friends aren’t getting this codex message, you should get new friends