>>/dobrochan/2404@603
The patient tried to wipe the brown spots on his pants for a long time, but spat and decided to disguise them with new ones. As if in the new year, no one will remember his smelly quiz about the openness and honesty of ML, because new layers will grow on the patient’s intra-trench fecal stalactitis.
>> I already know about half a dozen open-source ChatGPT replication projects.
Informative, Father. So where did the ChatGPT source and source datasets — or did you, without them, in your divine illumination, determine the extent of what you say is replication?
>> From left to right: Dall-e 2, Imagen, Muse. The understanding of the prompt is obviously growing, and the resource intensity of the models is falling.
Undoubtedly, sick, understanding grows. The new neural network for a request with three books gives not four, but two books. The progress that's been made is amazing.
Ahaha, the cover of the book has different colors!
What little things, really. For example, if you introduce you, Father, instead of haloperidol mounting foam per rectum, you will not, as you say, get fucked for such trifles?
Has their probability decreased from almost 100% to 20-40% in less than a year?
The most hilarious, the sick, is when a schizophrenic patient who thinks he is a scientist assesses the probability of successful output of a neural network from four pictures - and in addition gets values like "20-40%".
>> Let me take Gowers' assessment on faith.
Excellent, respectable. All you have to do is complete your dirty pants twerk with a spectacular pa with a quote in which Gowers claims that there are several "Fields Award level" people turning to him for advice.