X
关注 AI 研究者、开发者与机构的动态
按账号或来源筛选(541)
@hongming731@hongming731AI 评分44 @rohanpaul_ai@rohanpaul_aiAI 评分77 抱歉,您提供的主推文内容仅为一个链接(https://t.co/5auKJUwm38),没有可翻译的正文文本。请提供推文的实际文字内容,我将为您翻译。
@rohanpaul_ai@rohanpaul_aiAI 评分2727 
@elonmusk@elonmuskAI 评分1818 @alexandr_wang@alexandr_wangAI 评分44 @alexandr_wang@alexandr_wangAI 评分1818 @elonmusk@elonmuskAI 评分2222 试试 Grok Build https://t.co/XXiqcvJ2p7 https://t.co/QE7wyVm0bj
@cohere@cohereAI 评分3030 谁来定义 AI 的规则? 人工智能需要的是有据可依的标准,而不是卡特尔。来自 Cohere 联合创始人兼 CEO @aidangomez 的观点: https://t.co/QnDmIfsLPM
@cb_doge@cb_dogeAI 评分55 
@rohanpaul_ai@rohanpaul_aiAI 评分55 抱歉,您提供的主推文内容只有一个链接(https://t.co/b9ceneJM7p),没有可翻译的正文文字。请提供推文的实际文字内容,我再为您翻译。
@rohanpaul_ai@rohanpaul_ai精选AI 评分6868 
推荐理由:转述 FT 报道的 Anthropic 财务口径,读者可了解其 IPO 前的收入、ARR 与毛利率剔除项。
@ArtificialAnlys@ArtificialAnlysAI 评分4242 
@AISafetyMemes@AISafetyMemesAI 评分5252 微软 CEO 加入呼吁放缓 AI 发展的行列。发帖者指出,此前 OpenAI、Anthropic、Google 和 xAI 也加入了这一呼吁,并附上相关链接。
@rohanpaul_ai@rohanpaul_aiAI 评分3535 引用@rohanpaul_ai@rohanpaul_aiOkay, now Jacob Coxon on NBC News (the ex-Anthropic+OpenAI researcher who resigned). NBC News: "What do you mean when you say AI could kill us all? What is the worst-case scenario, or scenarios, that keeps you up at night?" Jacob Coxon: "Before discussing specific scenarios, it helps to just think of building AI as sort of inviting an alien mind onto the planet. You're building a human-level or superhuman-level mind without understanding what it wants, what it thinks, or the way it thinks. If you extrapolate into the future, AI could control autonomous drones or all the robots that are currently being built very rapidly." ---- From "NBC News" YouTube channel, (full video link in comment)
@rohanpaul_ai@rohanpaul_aiAI 评分55 @rohanpaul_ai@rohanpaul_aiAI 评分4444
引用@rohanpaul_ai@rohanpaul_aiJacob Coxon’s next interview on BBC (the ex-Anthropic+OpenAI researcher who resigned). "If we don't slow down at the current rate of progress, there is a strong chance that we could all die in the immediate future. It's not at all an exaggeration. And many of the people who are deep in the weeds and actually coding the thing, believe that there is a greater than 10% chance for a probability that humans might die. And they all keep this in their head on a daily basis while working on the technology." --- Full video link in comment.
@kimmonismus@kimmonismusAI 评分33 @AISafetyMemes@AISafetyMemesAI 评分3232 @rohanpaul_ai@rohanpaul_aiAI 评分22 @steipete@steipeteAI 评分1111 @emollick@emollickAI 评分55 给那些打算问 AI 这个引用出处的人 https://t.co/CQdWD9NkZY

@kimmonismus@kimmonismusAI 评分5656 
@elonmusk@elonmuskAI 评分1212 @rohanpaul_ai@rohanpaul_aiAI 评分44 @rohanpaul_ai@rohanpaul_aiAI 评分4949 Terence Tao 指出,训练和运行 LLM 主要依赖线性代数、矩阵乘法和少量微积分,属于本科生可掌握的内容,真正的谜团在于为何模型在某些任务上表现优异、在其他任务上失败且无法提前预测。

@EMostaque@EMostaqueAI 评分99 引用@EMostaque@EMostaqueDay 2 of my daily blog I dive deep into @DarioAmodei’s proposal but don’t see how it can logically hold. Intelligence is too often viewed as a crime. I discuss why & what to do about it. Also note OpenAI had embedded eval in 2023 - their board with @hlntnr & others Tough one https://t.co/GP19RKHvnT
@rohanpaul_ai@rohanpaul_aiAI 评分2323 @rohanpaul_ai@rohanpaul_aiAI 评分2828 
@rohanpaul_ai@rohanpaul_aiAI 评分2222 
@rohanpaul_ai@rohanpaul_aiAI 评分1515 在迈向 RSI(递归自我改进)的 L4 阶段,AI 从部署后发生的事情中学习,并将有用的经验转化为对其记忆、工具或代码的持久改变。 这样,智能体就能持续从真实世界经验中适应,而无需等待人类手动更新它。

@rohanpaul_ai@rohanpaul_aiAI 评分1919 在迈向 RSI(递归自我改进)的 L3 阶段,AI 开始根据自身当前的弱点和经验,决定接下来该学什么或练什么。 因此,不再是人类不断挑选训练任务,系统可以自己创造或寻找新的经验,针对它仍需改进的地方。

@rohanpaul_ai@rohanpaul_aiAI 评分2525 在迈向 RSI(递归自我改进)的 L2 阶段,AI 开始根据失败和反馈来决定如何改进自身,而不是由人类来选择每一项改动。 人类仍然定义目标,并决定什么算作成功的改进。

@rohanpaul_ai@rohanpaul_aiAI 评分3232 AI 进展并不均衡:模型在知识和推理方面已远比在工具使用、搜索和长时间运行的软件任务方面更接近上限。https://t.co/6YumUlvzxa

@rohanpaul_ai@rohanpaul_aiAI 评分2929 
@rohanpaul_ai@rohanpaul_aiAI 评分2929 AI 自我改进不是单一能力,而是一个从人类指导每一次改动,到 AI 最终改进创造未来改进的流程的演进过程。https://t.co/121HTL6KM2

@rohanpaul_ai@rohanpaul_aiAI 评分4141 一篇关于递归自我改进(RSI)的路线图论文指出,目前已能看到 RSI 的若干片段,但完整的递归自我改进尚未到来——多数所谓"自我改进 AI"仍无法改进其改进方式。

@emollick@emollickAI 评分2121 @alexandr_wang@alexandr_wangAI 评分99 我爱茶,@Muse 给我买了一套世界上最好的茶,还给我做了一张可爱的动漫图,把我收到的所有东西都画了进去!https://t.co/XJHfHqfIFn


@emollick@emollickAI 评分3535 @rohanpaul_ai@rohanpaul_aiAI 评分77 对“AI 放缓”叙事来说,这翻译确实有意思 https://t.co/tliGgrTWI4 https://t.co/OrVk0So0XJ
