Google’s Late Night King has a super update! PaLM 2 large model is shockingly released! Bard writes code and doesn’t know what to do!

This time, Google is anxious!

They have been really frustrated these past six months! The development speed of AI in the past six months has been like a rocket. In November, ChatGPT was launched. New Bing released in February this year and GPT-4 released in March have made it difficult for many major manufacturers to catch up.

Not to be outdone, Google immediately launched the chatbot Bard. As a result, the stock price plummeted due to a small error in the demonstration video (in fact, GPT also made errors). Now, at Google’s annual I/O conference , Google needs to convince the public and shareholders of its strength in AI to "avoid its previous shame."

It is worthy of recognition that Google is indeed a leader in artificial intelligence research.

You must know that the "T" of GPT is the Transformer model, which is a classic NLP model launched by Google in 2017. With such a good architecture in hand, Google lags far behind OpenAI in developing artificial intelligence products, and has also missed the trend of artificial intelligence.

So can Google turn the tide this time? Let’s take a look at what cool updates Google released at the I/O conference!

1. New features of Bard are coming: personal assistant, fully automatic photoshop

Too lazy to write an email? Don’t know what tone to use to be more polite when writing an email to your instructor?

Google Bard's new "Help me write" not only uses generative artificial intelligence to assist users in composing emails, but also provides a variety of tone options for users to choose from. For example, if you need to request a refund from an airline, you just tell it what you want and it will generate a draft for you based on the flight details you provide. It's more concrete and clear when you say it this way.

Road lovers are also in luck, Bard has launched a new immersive route view, which allows you to experience the entire journey in advance before reaching your destination.

Imagine you want to ride a bicycle in New York. With immersive route views, you can choose a route along the seaside and enjoy the beautiful scenery. You can zoom in and out of the map to get a bird's-eye view and even see bike lanes in detail. This is definitely a new way to travel!

Immersive Route View is also planned to be available in 15 cities by the end of the year, including London, New York, Tokyo and San Francisco. This means you can experience your trip in these cities in advance and feel the charm of the city and the beauty of the attractions.

The one-click image recognition function is also here! It doesn’t just help you identify who the person in the picture is! Not only can you search for specific people, scenery and other elements in photos, you can also find photos of your friends!

Just enter their name in the search bar and the album will find the relevant photos for you. Moreover, you can also use editing tools to improve the quality of your photos. For example, if you want to remove some interfering objects, adjust the light or change the color of the sky, you no longer need Photoshop. These can be achieved through the AI-driven editor.

# Google not only applies AI technology to photo albums, but also launches some amazing features in other products. For example, Google launched Magic Eraser on Pixel phones, which uses AI technology in computational photography to eliminate unnecessary interfering objects in photos.

Don’t know what major to apply for in college? Afraid of falling into the wrong profession for the rest of your life? Bard can help too! If you like playing games and are interested in animation, you would like to know if there are any universities in Pennsylvania that offer excellent animation courses. Next, you can use Google search to find some universities with excellent faculty and great achievements in the field of animation.

Bard can display the universities it recommends in a table and display the locations of these schools through Google Maps, so that your parents can understand the location of the schools more intuitively. Bard can also help you mark which schools are public schools and which are private schools.

This service is not over yet. Bard will use Google Sheets to create a table to display your choice of schools in a clear and orderly manner, and then share your choices with your family through Google Docs and Google Drive so that they can help You further search and select.

In the coming months, Bard will also be partnering with Adobe Firefly’s tools, allowing you to instantly create something out of your imagination and transform it into a completely new image using Bard .

For example, if you have a 7-year-old who loves unicorns and has an upcoming birthday, you want to send a fun birthday invitation with a picture in the shape of a unicorn and a cake. , then by partnering with Adobe Firefly, you can turn this idea into a reality.

This tool can turn your imagination into visual images. The designers are probably sweating again!

White-collar workers can also free their hands in their daily work. Bard's "Help write me" not only writes articles, makes sales presentations, writes project plans, and develops customer copywriting, but also Good at it. One of the most popular use cases is writing job descriptions, which is very important for any size business.

Using this feature in Docs is very simple. For example, let's say you run a fashion boutique and need to hire a senior fashion designer. You only need to enter a few prompt words, such as "Senior Fashion Designer Job Description", and then Docs will send these prompt words to the backend to generate the required job description. Tools like this can save a lot of time and effort and streamline the hiring process.

In addition, Google Sheets also provides convenient tools for organizing information. For example, if you run a dog walking company, you need to record customer information, logistics arrangements, dog walking times, etc.

In a new form, you just enter a few descriptions, such as "Clients and Dog Walking Business Rates" and "Pet Roster" and click Create Form. Sheets will be sent to the trained model based on your commands to generate a table containing dog names, customer information notes, etc. Tools like this greatly simplify form work, allowing you to focus more on the things you love.

In addition, Google’s slide tool also provides convenient functions. People often use slides to tell stories, both at work and in their personal lives.

For example, if your family members are creating a slideshow for their parents’ 50th wedding anniversary, each family member can contribute something.

You can select a sentence as the image generation prompt and click Create Slideshow. The system will send the input command to the text image generation model and generate 6 to 8 images for selection. This way you can choose whatever you want and ultimately design a great slideshow.

The introduction of these functions provides great convenience for people’s work and creation, allowing them to complete tasks and realize ideas more efficiently.

It’s also easy to use Bard. Google has removed the waiting list process. It is now available in more than 180 countries and regions, and more regions will join soon!

Not only that, Google has also launched Japanese and Korean versions of Bard, and plans to support the top 40 languages ​​​​around the world soon.

At the GoogleI/O2023 Developer Conference, the most eye-catching aspect of AI was the launch of an upgraded version of Bard (Source: Technology Big Guys’ Insights )

2. PaLM 2 large language model is released, supporting multi-modality, and there is nothing you don’t understand about the code!

Google also launched the latest PaLM 2 model, which is based on Google’s basic research and the latest infrastructure. Its new base model Gemini was also released. It is Google's first multi-modal model built from scratch, with high capabilities at different scales and efficient integration with other tools and APIs.

Med-PaLM 2 can help answer questions and summarize insights from a variety of dense medical texts, and in the future will add the ability to process medical images.

Sec-PaLM is a model trained on security intelligence that powers new Google Cloud products to help organizations and security analysts detect, understand, and defend against threats. .

In addition, PaLM 2’s models perform well when fine-tuned on domain-specific data and can be trained according to the requirements of specific tasks.

It is worth mentioning that Google merged the two teams into Google DeepMind and received support from Google's computing resources to build a more capable system and ensure safe and responsible deployment.

According to reports, with the support of PaLM 2, Bard’s mathematical logic and reasoning capabilities have been greatly improved, which can help developers perform better programming. Now, it can help us explain and collaborate throughout the code generation and code debugging process, and has learned more than 20 programming languages, including C, Go, JavaScript, Python, etc.

For example, a user wants to learn to play chess and wants to write a four-step killing algorithm in Python. Using the Python language, Bard wrote a script that implemented the algorithm and formatted it very clearly and understandably.

But if the user doesn't understand this algorithm, the user can click on a certain part of the code and Bard will be able to associate it to the original address and be able to understand the code and provide an explanation, such as explaining the meaning of the chessboard code and what it means throughout the code. role in the line. This explanation function is very useful and can help simplify the complex and make the code clearer and easier to understand.

Want to improve your code? Bard can help you too. With some suggestions, such as using compressed lists and generator functions, users can incorporate these improvements into a single block of Python code. Bart's ability in this area is constantly improving and he can re-correct the code.

Poor English? Don’t know Korean? Bard can also help you communicate codes with Korean colleagues through cross-border collaboration without any barriers! If you need to debug code with a Korean colleague, you can ask him to fix errors and help your teammates by adding Korean comments to the code. First it recognizes that the code is actually recursive, so it will suggest a fix, then it will explain the reasons behind the fix, and add Korean comments as per your request. It’s really great, Kang Sangmi hit Google xi!

3. Google Pixel Fold and Pixel ecosystem are fully equipped with AI

Google also showed off the latest Pixel series.

Google showed off how they are expanding the Pixel line with two new products: the Pixel Fold, a foldable phone, and the Pixel Tablet, a tablet. These new devices are designed to provide more choices and experiences, enrich the Pixel product line, and strengthen Google's position in mobile devices.

Pixel Fold is a foldable phone that can be used as a regular phone when needed, or unfolded into a larger tablet for a richer visual experience. Through this design, users can switch between different scenarios and enjoy more flexible functions.

Pixel Tablet is a tablet computer for handheld and home use. It uses a high-resolution display and built-in speakers to deliver clear images and audio. Equipped with a Google Tensor chip, it provides users with an excellent Android tablet experience and supports personal AI skills.

Google highlights the Pixel Tablet’s advantages in video calling, photo editing, and voice recognition. It makes video calls clearer by optimizing camera functions and AI customization, and provides some convenient editing tools. At the same time, the Tensor G2 chip is used to connect to the PC to achieve a faster voice input experience.

In order to improve the application experience on large-screen devices, Google has done a lot of application re-adaptation work and provided more than 50 Google applications. Pixel Tablet also supports switching between multiple users, so each user can easily access their own apps and content. In addition, it also has content transfer capabilities between mobile devices, allowing users to share media between different devices.

After watching the entire press conference, do you think Bard can make a comeback, compete with ChatGPT, and lead the AI ​​wave again?

The above is the detailed content of Google’s Late Night King has a super update! PaLM 2 large model is shockingly released! Bard writes code and doesn’t know what to do!. For more information, please follow other related articles on the PHP Chinese website!