Avatar
Kis Sean
0d6f3fb7f3c83755ea731380516167da6691cea0d7ddf4865505d291687ca343

那给够它 4TB,也不算特别高的难度啊。比ASIC和GPU都简单

不算最后跑到卫星上的情节。samaritan好像没这么干吧,the machine反而是自己拆分到了电线杆的盒子上

Replying to Avatar Kis Sean

Finally got to try bing AI chatbot, my overall experience is not impressed. Most of my search can be done in less than 5~10 seconds, 20 seconds top. If I am using voice assistant the whole experiences is even more seamless than typing. Bing still need around 10 seconds to search/crawl infos online, and another 10+ seconds to inference. (or it's a purposed latency to offload the system, but the user experience is slow)

With 20 seconds, if it's some general/popular topic, I can definitely find a proper introduction article and skim through the summary. Some easy optimizations/debuggings can be done, for example, when I ask to translate a paragraph, it would search the whole thing online first. Or when I said "translate what I said earlier", it would search "translate what I said earlier" and translate that sentence, or doesn't reply at all. Seems not even as smart(dumb?) as ChatGPT.

Maybe it's good for some questions which the information that's too scattered online, or the wording/keyword is difficult for current search engine. For example, "python code to concatenate two strings and remove all spaces in between."

Overall, I wouldn't expect anything useful to me until the chatbot system (not just the AI model) can figure out when it can answer my question without searching online first. And it doesn't need to scraw the whole world wide web when I just need to know the weather. Google and amazon probably has tons of API for some easy questions like this, AI chatbot should be able to leverage those. Or implement the whole system the other way around, only let AI chatbot handles the questions that current voice assistant can't handle.

终于有机会试试Bing AI聊天机器人,我的整体体验并不是很满意。我大部分的搜索都可以在5~10秒内完成,最多20秒。如果我用语音助手的话,整个体验就更无缝了。Bing 还需要大约10秒来搜索/爬取网上的信息,再加上10多秒来边跑神经网络推理边显示结果。(或者这是为了减轻系统负载而故意延迟,但用户体验就是很慢)

用20秒的时间,如果是一些普通/流行的话题,我肯定能找到一篇合适的介绍文章,并足够时间浏览一下摘要。不少简单的优化/debug可以做,比如当我要求翻译一段话时,它会先在网上搜索整个内容。或者当我说“翻译我之前说的”,它会搜索“翻译我之前说的”并翻译那句话,或者根本不回答。看起来甚至没有ChatGPT那么聪明(笨?)。

也许它对一些网上信息太分散,或者用词/关键词对当前搜索引擎来说很难的问题有用。例如, “python code to concatenate two strings and remove all spaces in between”

总的来说,除非聊天机器人系统(不仅仅是AI模型)能够明白哪些情况可以不搜索就能直接回答我的问题,否则我不指望它对我有什么用。让它回答天气,赛事等简短的问题的时候,它也没必要把整个互联网扒下来才能找到答案。Google和Amazon的AI bot能这么快回答一些简单的问题,应该是已经有很对对应的API了,AI chatbot想变得更有用,必须要应用上那些。或从另一个方向搞,只让现有的语音助手不能直接回答的时候才调用LLM的AI聊天机器人。

#AI #OpenAI #ChatGpt #Bing #Microsoft #Chatbot #Assistant

https://twitter.com/stephen_wolfram/status/1638951773040672768

finally, hopes this will be implemented in the bing chatbot soon

前几天我看到说被黑客占领了很多私钥之类,我还以为项目方会先处理完

L1容量辣鸡,“现在任何L1都接纳不了多少用户”是共识了。BTC主链的频繁转账费用已经超过一般人能承担的范围很多了。

L2需要L1的安全性啊,和BTC 和 LN的原理一样的,LN当有信任不过的转账时候,就broadcast结账

以上只是说blockchain系统共有的特性,至于ETH能不能活不在这个话题内

把剩下一点给你,不多,不好意思

把剩下的一点给你,不多,不好意思

已经满员了,不好意思

哦,可能是因为你的地址是getblby.com;如果你有WOS的话可以post在下面或私信给我

也不是,之后的话说不定了。不过我觉得存着这些sat在这个平台对我来说也没什么用了

不好意思,已经满员了

不知道为啥你的地址失败了,它说 Error Fail to decode invoice, 试了几次都这样。要么你留一个 3k7的收据在这里下面我试试吧

我要把我的sat都发出去,谁想要的话请在下面评论。只分给三位(因为确实很少),我会多刷新几次尽量保证没错过。

Comment below if you want sats, only for the first 3, because I only have around 11k

#[2]

我觉得真的是晶哥在冲KPI的话,这个话题里,你互动下暗示的内容就够了。