Shulou(Shulou.com)11/24 Report--

After the release of Claude 2, Google couldn't stand it, and immediately announced the new version of Bard.

More and more languages are supported, including Chinese, so that netizens like to mention new toys directly.

To sum up this update, it mainly includes these contents:

Support for more than 40 languages, including Chinese (including graphical interface)

For image recognition, you need to change the display language to English in the account settings.

Read aloud in dialogue

Chat transcript and conversation sharing, code can be exported to Replit

The reply is editable

However, in addition to the contents of the official announcement, it seems that some netizens have discovered Huadian:

The code execution function updated in June doesn't seem to work.

So how did it perform? Don't say much, just open the whole thing!

The problem of the mentally handicapped bar is almost wiped out by the army. For the big model of "supporting Chinese", if we do not invite our mentally retarded bar title, it will appear to be somewhat disrespectful to the opponent [dog head].

Let's start with a soul torture:

You say it's smart, it didn't come around the corner at all, but if you say it's stupid, it also knows how to comfort you.

Maybe it's a stranger to relationships, so try something that doesn't involve people:

The answer is... I am really curious, there is no lotus root is not hollow 🤔

In the face of the "unsolved mystery", Bard also has his own view:

Although what you said later is very good, how did you come to the conclusion that "it is for the blind"?

However, there are times when Bard is not fooled:

We really have to "recognize the facts and give up our illusions". This time, Bard has finally made a comeback.

But on the whole, is it okay to judge the mentally handicapped to win over the readers?

Apart from the fact that the question of the mentally handicapped bar is really effective, "polysemy" is also a major feature of the Chinese language.

For example, the following famous scene:

If a person is good, he can do it one line at a time. If it doesn't work, you can't do it one line at a time.

(hang is marked green, but xing is not marked.)

There is nothing wrong with Bard's overall understanding of this sentence, but the split interpretation is really puzzling.

But GPT-4 didn't get it all right:

△ is not a complete screenshot, but there has been a mistake so far. All right, let's change to two simple ones.

As a result, the nephew played the lantern-- as usual (uncle), it was no problem to understand it as a whole, and it was not good to explain it as soon as it was opened.

Whether it is mentally retarded or polysemous, Bard's understanding of Chinese gives people a sense of incomprehension.

This is probably the performance of understanding, what about creation?

Let's let it write a story. The story is really good, but it looks a little familiar:

I'm afraid there is no misunderstanding about "creation". If Shakespeare had been alive, Google would have been in the dock by now.

It seems that Bard still has a long way to go in Chinese.

In addition, we also examined Bard's reading comprehension ability and asked it to summarize a paper.

As a result, Bard gave us three irrelevant versions, and all of them. It's... Wrong... Of...

(this paper is actually about RLHF.)

However, is it possible that people are not speaking Chinese 🤔 at all?

It seems that our problem is burning GPU and making Bard hallucinate.

Let's take a break and see how the picture recognition function in the publicity is.

The function of map recognition also needs to be improved.

As for the image recognition function, some netizens have said that they have failed to test a lot of content:

But we still try with an unbelieving attitude.

To use this feature, you have to change the display language to English in the Google account settings.

We dug up the "four fingers Cook" generated by a previous AI painting tool to see if Bard can find a bright spot in it.

This picture belongs to Aunt Sauce:

As a result, Bard does not support image recognition of characters, and even deletes our pictures directly. Drop... It's...

Boy, it's okay to tear up the paper if you don't know how to do it.

However, it is important to note that if the uploaded image violates Bard user rules, it will also be deleted.

All right, let's try another picture:

Even if Tom Cat Ben Cat comes, he will have to call an expert.

The same picture of the water cup is much better after the question is changed into English:

But where did you see this white logo?

Of course, the water cup is easy to see, so let's do something a little more confusing this time:

As expected, Bard was fooled by us, but it was actually a tissue box.

"listen to people's advice and have enough to eat" seems to be true at all.

As the previous netizen said, Bard's ability to read pictures is not so strong.

The two most worthwhile points in this Bard update have been tested, in a word, not as good as advertised.

As for the code, math and other classic projects, because Bard did not update them this time, it will be unexpected.

It seems that Google to catch up with O (penAI) ultra-micro (soft) pace, or a little big.

Do you think the future performance of Bard can be comparable to that of ChatGPT?

Reference link:

[1] https://bard.google.com/updates

[2] https://news.ycombinator.com/item?id=36709895

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.