|
|
Well, in that case, let’s have some side dishes and ask it if it knows about the Gemini that Google just released two days ago? It gave the answer accurately, and also attached some links to Geminirelated tweets that day. It probably wanted to tell us that its information was summarized from here. Next, let’s make it a little more difficult and let it summarize what major events are happening in the technology field today. The result this time is really hard to judge. It is true that the news happened recently, but a lot of it is not from today. It is a little timesensitive but not much .
First, Gemini was announced several days ago, and the rest, such as Apple system updates, the new EU AI agreement, etc., were also very vague. Anyway, after looking at the things it summarized, Shichao didn't get any incremental infor Armenia WhatsApp Number mation , and some of the news didn't come from the source and I couldn't believe it, so I had to check it out myself . . . From the above rounds of experience, Grok feels pretty good to Shichao. He has entertainment ability , can quickly get some jokes, and is quite humorous when answering. Regarding the realtime database, it can indeed be seen that Grok is constantly updated, but its summary ability is at a mediocre level at best.
It is far from enough to use it as a productivity tool, and it is not as convenient as a search engine. Of course, this is not just a problem with Grok. Other large models with realtime retrieval capabilities have more or less a la carte summary capabilities. After experiencing the relatively unique functions, as usual, we have to test some basic issues with large models on Grok. This time, in order to ensure that the true strength of Grok can be measured, Shichao specially found the paper from the last time Google tested Gemini's MMLU (largescale multitask language understanding) capabilities, and picked out some questions from it to ask it.
|
|
發表於 17:03:07