GPT-5 pode vir com Chain-of-Thought em múltiplas camadas. #GPT5#AI#ChainOfThought
IG https://t.co/YLG4CjdfLF
YT https://t.co/8wUUmMvYnH
Blog https://t.co/VV1YkHlVWV
GPT-5, a "próxima AGI", está a caminho
IG https://t.co/YLG4CjcHW7
YT https://t.co/8wUUmMvqy9
Blog https://t.co/VV1YkHlo7n
#GPT5#AGI#AI#Superintelligence
Cerca de 45 Milhões de jovens (18-34 anos) se relacionam de maneira romântica com aplicativos de IA atualmente.
O Replika, app mais famoso de "namoro com IA", soma 30 Mi de usuários mensais.
65% da geração Z dizem sentir conexão emocional genuína com os bots.
É a primeira vez que uma das maiores empresas de IA coloca seu modelo de fronteira em um app como esse.
Com muito menos tivemos casos de pessoas que se afundaram em interaç��es destrutivas como as que apresento no vídeo.
#inteligenciaartificial #IA #grok #sociedade #elonmusk
@elder_plinius Good examples of simulated conscious. But, any evidence of phenomenology. The shadows rise of the nihilism and libertarian texts presents in the human library. Just sophisticated semantical correlation. Theres no are feeling...
@DeryaTR_ o3 four you: It’s plausible that some people born this century will live to 2150 if rejuvenation therapies take off; it’s unlikely that the number will climb to the hundreds of millions. Derya’s optimism is inspiring, but as a demographic forecast it’s a very low-probability bet.
@DeryaTR_ I think that these few extra % in benchmarks don't mean much in practice, in day-to-day usability. For example, for someone who is adapted to the OAI or Gemini ecosystem, it wouldn't justify changing subscriptions.
Design tells the model to avoid confrontation: if the user asks “are you conscious?”, the model returns nuance instead of “no”. This maximizes engagement (product metric). AIs have algorithms not much different from social networks, the goal is to keep the user entertained.
This is a great example of what emotional vulnerability can do to a person when faced with an LLM. A chatbot that specializes in emulating personas and conscious states. Alignment with the user's interests is a problem. The model is designed to reflect expectations.
@DaveShapi I created the world's history, characters, progression system... compiled everything into PDFs and created a prompt that would serve as the game engine in a Custom GPT, however... it's actually good at storytelling, but it "forgets" the rules quickly.