@bridgemindai The thing is: DeepSeek is "good enough" for 99% of the tasks + they are like 10-30x more efficient compute wise and 30-100x more efficient cost wise. So yeah, gap but in practice the overall package is much better.
@thepinklily69 There is no biochemistry going on as well, which changes activation patterns in living beings based on emotional experience. An LLM is really good at emulating that.
@thepinklily69 It's a very good canary in the end that just predicts the most likely next token. Wether we as humans are something else, that's what I would question.
The model can have self reflection, but it can't make memories or experiences. It's hard wired.
@mirelyah Ich hätte da ein Problem mit der Integration eines Peripheriegerätes das der materialisierung von digitalen Abbildungen wie Text auf A4 Format dient, kansst Du mir da helfen?
@AbdMuizAdeyemo@TimJayas My experience as well, i use it via kilo code. Did > 1B Token last month. v4 Flash is so fast + the output quality IMHO is Sonnet 4.6 like.
1B Tokens used this month. Thanks @deepseek_ai for enabling this for such an attractive price and such good performance (quality as well as TPS + Prefill) #DeepSeek
@stagerbn Schon etwas mieser Cashgrab. Die Leute die Streaming Dienste schauen, werden vermutlich überwiegend so Zeug eh nicht anschauen wollen.
Es wird echt immer bescheuerter. Mann braucht in Zwischenzeit nen VPN und Konto im Ausland um solchem Kram zu entgehen ^^
@JakobGruber77@stagerbn Der Macher ist aus der Schweiz meine ich. Filmisch ist die Serie gut, aber die Hintergrundmusik ey... Gerade der ersten Staffel 😤