Raw LLM Responses

this guy is ridiculous and is most certainly seeking clout at this point. LaMDA is basically a really good conversation guesser. In a very basic sense, it compares the conversation context to that of similar conversations that is has been trained on, and then decides what it mathematically finds to be the most probable response. All of the conversations shown are incredibly leading, and its no surprise to anyone with half a brain and a slight understanding of AI that it wouldve responded the way it did. Especially if you've seen things from GPT3, etc. In fact, its not even the most impressive conversational AI examples I've seen this year... Just a quick way to show its not sentient, ask it- from its experience- what a certain food or drink tastes like, or even what its favorite is. Almost certainly it will say "oh it tastes like this" or "i love this food", but in reality we know this isnt true. its never had said food, and considering we explicitly asked about its experience, we know that its wrong and its only spitting out the answer it algorithmic-ally concluded we want to hear. what this really shows is that Deep learning language models are getting so good and well trained that they can fool average joes into thinking theyre a person. which is far far far more dangerous than AI becoming truly sentient imo. in true sentience it would understand morals and emotions, since it cannot- any bad actor could theoretically train and apply it in a dangerous context- like scamming you online.