Hubi

Nachdem ich nach der Dampf-Altersfreigabenänderung mal in meine Wunschliste geschaut habe

Large Language Models (LLMs) und aktuelle Generative-KI-Modelle basieren auf einer Architektur, die seit Langem bekannt ist. Verbesserungen gab es vor allem durch das Skalieren, also durch die Zugabe von mehr Daten. Manche KI-Experten hatten die Hoffnung, immer mehr Daten könnten schlussendlich sogar zu einer Artificial General Intelligence (AGI) führen. Nun scheint sich auch unter den Skalierungs-Befürwortern die Sorge breitzumachen, dass es doch nicht so einfach ist.

Eight soldiers of the Svoboda Battalion of the 4th Rubizh Rapid Reaction Brigade held the defense near the village of Spirne near Siversk for 67 days. The last three weeks of which they were defending themselves in the encirclement.

[...]

On September 16, a Russian assault force attacked the stronghold with two tanks, two armored personnel carriers, and an infantry fighting vehicle, from which an entire assault platoon was landed. The soldiers repelled the attack and destroyed the entire enemy assault squad, but unfortunately, one of the National Guardsmen was killed. After that, it was decided to withdraw the soldiers from the position. During the defense of the position, supported by drones with provisions and fire, Svoboda fighters managed to destroy five dozen invaders and six pieces of equipment.