In addition to Weibo, there is also WeChat
Please pay attention
WeChat public account
Shulou
2025-01-29 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >
Share
Shulou(Shulou.com)11/24 Report--
Bing can read the picture! With a picture at the beginning, all the needs are met. This wave, GPT-4 is going to lose hemp.
What, Microsoft's Bing can recognize pictures? Or are you ahead of ChatGPT?
Compared with the original GPT-4 preview, which only painted cakes and did not land, Bing was in the limelight this time.
Some Reddit netizens found that an option to upload pictures suddenly appeared on the interface of Bing.
It is said that Bing can do anything after uploading a picture.
It doesn't matter whether it's programming, writing code, doing problem mapping, or even seeing a doctor.
(but it seems that it is still being tested on a small scale, such as the editor hasn't experienced it yet.)
After a large number of netizens found this function, the first thing to test must be a variety of stem pictures.
Type the outline into Bing and see if it can figure out where the punchline is.
For example, in the picture below, a man is ironing in the back of a taxi, and the car is still driving.
Bing says there are too many unusual parts of the picture. First of all, ironing in a taxi is outrageous, and secondly, does it imply that men wear the same color as taxis? Or is it some kind of coincidence?
Third, the ironing board is fixed to the taxi, which risks instability. Finally, the clothes the man is ironing are blue and collide with yellow. is it implying something?
It can be seen that Bing still captured all the details of the picture as much as possible, and even the analysis was quite reasonable.
And in the figure below, "Why add layers to the neural network?" Bing also gives his own analysis in the cartoon.
This cartoon exaggerates the difference between statistical learning and neural networks, contrasting the serious and technical language used by the former with the latter's simple and direct advice, creating a humorous juxtaposition effect.
In general, the cartoon takes advantage of stereotypes and common perceptions related to the two methods of machine learning, highlighting the difference between statistical learning and neural networks.
Of course, the same picture is also shown in the preview version of GPT-4.
GPT-4 hit the nail on the head and pointed out that the cartoon satirizes the differences between statistical learning and neural networks in ways to improve the performance of models.
But in another set of figures, Bing doesn't seem to have a get to the point.
Bing also made a sample analysis of the analysis, it said that maybe it is the contrast between the joint and the packaging? It could be that the bottom of the plug looks like a face.
I even analyzed what the three pictures were without words, but didn't mention where the stem was.
However, if Bing is asked to do some objective analysis, such as identifying wire brands, the results are still very accurate.
By contrast, when GPT-4 analyzes the following set of drawings one by one, he can react immediately--
The "Lighting charging line" in the picture looks like a large and out-of-date VGA port, plugged into this small and modern smartphone.
When writing code, when GPT-4 was painting the cake, the most amazing thing was its excellent understanding of the code.
In the demonstration, Greg Brockman drew a sketch directly on the paper, took a picture, sent it to GPT-4 and said, write the web code for me according to this layout, and I wrote it.
This little brother also asked Bing to give it a try and see if he could set up a joke website with the same style.
The final effect looks really good, although the interface does not look very beautiful, but there should be.
The program code is as follows:
Then, the little brother asked Bing to continue to improve the code and make the corresponding web page.
Bing says of course, just fine-tune the code. However, it is pointed out that JavaScript is not a compiled language, but an interpreted language.
So you don't need a compiler, just a browser or a JavaScript engine.
After the popular science of friendship, Bing also successfully completed the task.
As a doctor, the following use case is more powerful.
The little brother directly took a picture of his red arm and asked Bing what was wrong with him.
Bing was directly diagnosed as bruising and subcutaneous vascular rupture.
There are even a number of ways to relieve symptoms:
Apply ice pack in the first 24 hours to relieve swelling and inflammation.
Heat 24 hours later to promote blood circulation and remove congestion.
Lift the bruised part to a higher position than the heart, lowering blood pressure and blood accumulation.
Take OTC painkillers.
Eat more foods rich in vitamin C, vitamin K, zinc and bromelain to speed up wound healing and collagen synthesis.
Apply ointment to relieve bruising and swelling.
It's pretty professional.
And the little brother sent a cross-section of the organization and asked Bing to break it off.
Bing not only identified the content of the picture, but also explained in detail what is the organization, what is the cross section.
It even explains when a cross section of the organization is used.
Then the little brother asked, what do you think is the section of the organization?
Bing analysis shows that this is a cross section of muscle tissue.
Even, Bing points out very seriously that he can't judge whether a person is healthy or not based on a single picture.
There are also netizens who personify college students in suffering, asking Bing to be a good teacher and tell him what is in the picture.
Bing said, listen carefully to the teacher (delete), this is the kidney yuan, there are four main functions, respectively, is blah blah blah...
Boy, in the future, this kind of learning task of combing and summarizing will be directly dumped to AI. AI will never lose patience.
Someone even threw Bing a question about cell meiosis.
Bing said the uploaded picture is a schematic diagram of meiosis, dividing from one diploid cell to four haploid cells.
Then it explains the process and significance of meiosis.
Sometimes it is not very bright. Of course, there is still a lot of room for improvement in the picture recognition function of Bing.
For example, the elder brother asked how many xs are in box An in the picture (according to the legend, A.trifida is xx).
At first glance, there are 11 human eyes, but how can Bing be wrong?
Bing said there were a total of nine, five in box An and four ○ in box B. that adds up to nine.
Box A said, box X said! How ○ and B frame.
Or, for example, in this Nintendo character diagram, Bing recognizes only seven of the characters.
Then, the little brother asked a question of chess: "in this opening game, how should White go next?" "
But Bing answered wrong as soon as he came up: "now it's black's turn to go. As for how white goes, it depends on black." "
Netizens hastened to correct: "No, you first see the position of the pieces, and then I repeat, now it's White's turn to go."
However, Bing still did not answer correctly this time-there are several places where there are actually no chess pieces.
It can be seen that the newly added map recognition function of Bing is quite powerful.
The beginning of a picture, behind all rely on the strong generation ability to play.
It is not a big problem to identify and generate factual feelings. I just don't know if the party with cultural metaphors in the future can do better.
Reference:
Https://www.timesnownews.com/technology-science/snap-and-ask-microsofts-bing-chatbot-now-accepts-picture-prompts-article-100900618
This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)
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.
Views: 0
*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.
Continue with the installation of the previous hadoop.First, install zookooper1. Decompress zookoope
"Every 5-10 years, there's a rare product, a really special, very unusual product that's the most un
© 2024 shulou.com SLNews company. All rights reserved.