Google I/O expresses Google's anxiety with 200 "AI" sentences

Tech 2023-05-11 07:22:49 Source: Network

Before starting the article, let me briefly introduce some cold knowledge: this year is the 15th year for Google to hold the I/O Developer Conference.Throughout the past seven years of I/O conferences, AI content has always been a top priority for Google

Before starting the article, let me briefly introduce some cold knowledge: this year is the 15th year for Google to hold the I/O Developer Conference.

Throughout the past seven years of I/O conferences, AI content has always been a top priority for Google. For example, before the explosion of ChatGPT, Google had already shown the LaMDA series language generative model at the I/O conference, but had not waited for the opportunity of the fire.

Nowadays, the outbreak of ChatGPT has brought AI to its best moment, but it also makes Google feel extremely anxious.

After watching the entire conference, Google mentioned the term "AI" nearly 200 times, almost including AI in all products including Gmail, search, Pixel, and others, eager to prove to everyone that they are still at the forefront of AI.

If this conference were held six months ago, it would be incredibly exciting: AI assisted office work, AI search, AI programming, each of which is an important technological application. Unfortunately, before that, Microsoft had already gone all the way.

The reaction of the on-site audience can even be said to be a bit cold. In fact, there were only two enthusiastic cheers at the scene: Bard's dark mode and Pixel tablet's free base.

At today's I/O conference, Google's new products and technologies are like a catch up with Microsoft and OpenAI - catching up with the AI future that was first implemented.

PaLM2, Google's Secret Weapon Against GPT-4

The big language model has become the top product in the technology industry, and companies hope to come up with self-developed models that can challenge GPT to prove their scientific research capabilities, and Google is no exception.

Google has released a new generation of big language model, PaLM2, to compete with GPT-4.

According to Google, they used a large number of mathematical and scientific datasets to train PaLM2. Compared to last year's PaLM model, PaLM2 has greatly improved its multilingual processing, reasoning, and encoding capabilities.

According to Google's official testing, some results of PaLM2 (such as math) are even better than GPT-4.

Google did not disclose the specific parameter size of PaLM2, only mentioning that PaLM2 is built based on the latest JAX and TPUv4 computing facilities. Google has optimized the algorithm for PaLM2, making it smaller in size than PaLM, but with better overall performance and higher computational efficiency.

PaLM2 supports over 100 different languages and over 20 programming languages, including JavaScript, Python, Prolog, Verilog, Fortran, etc. It can be said that it is a language expert.

It is not enough to only perform regular text comprehension and generation. The multi-modal PaLM2 supports understanding and generating audio and video content. Google has demonstrated a model called Med-PaLM, which is specifically tuned for the medical field and can analyze patient injuries based on X-ray images. Google has reached the level of a human doctor in this regard.

The PaLM2 model can also be 'split' into four versions of different sizes and deployed on various devices. For example, a relatively small-scale Gecko model can be run on a smartphone, allowing mobile devices to also have large language models.

It is worth mentioning that PaLM2 is currently open for preview, and Google's own chat robot Bard has used the PaLM2 model to enhance its answering ability. Google Content has over 70 product teams using PaLM2 to build products, and the vast majority of updates at this conference were designed based on the PaLM2 model. It can be said that PaLM2 is Google's core engine for comprehensive AI.

In addition to PaLM2, Google CEO Sandel Pichai also revealed a bit about the next generation big model Gemini, which he claims aims to achieve multimodality, efficiency, and build for future innovation..

Gemini is a new model jointly developed by two top AI teams, Google Brain and DeepMind, after merging. From the naming, it can be seen that Gemini will combine the strengths of both teams and is expected to be among the most powerful large language models.

Bard: Upgrade multimodal, capable of drawing and reading images

Better looking, more comprehensive, and more people can use it.

After less than 2 months of updates, Google has migrated Bard to the more powerful large language model PalM2, which is a competitor of ChatGPT and BingChat with significant updates and stronger capabilities:

Cancel the waiting list and open it to more than 180 countries and regions;
Add picture Q&A to make the conversation more intuitive;
Integrate Adobe Firefly to quickly generate images;
Source references, dark mode, code export button, updated for more details.

It is not easy to use Bard, which has many restrictions in the UK and US, only knows English, and has a waiting list. Starting today, Bard officially announced the cancellation of the waiting list and opened the experience to more than 180 countries and regions.

Bard now also supports Japanese and Korean, and will soon expand its support to 40 languages, including Chinese.

In terms of functionality, Bard has also undergone significant upgrades, which can be summarized in four words as' illustrated '. When asking Bard questions, it is no longer a robot that only replies to text dryly, but can also display richer and more intuitive answers in the form of pictures.

At the same time, Google has also brought the power of Google Lens into Bard, allowing the latter not only to reply to images, but also to understand them and draft content based on image information. For example, you can send a photo of a dog to him and ask him to write some titles based on the breed of the dog.

Perhaps the feature of the next door Wenshengtu has received too much praise, and Google has also added this feature to Bard. However, it is a collaboration with Adobe Firefly, integrating Firefly's Bard to generate high-quality images in natural language. And with Adobe's endorsement, there is no need to worry about the copyright or quality of the images.

Adobe is just the beginning. Google will also work with Kayak, OpenTable, zip recruitment company, Khan Academy and other partners to add more functions to Bard.

Compared to ChatGPT, Bard's positioning is more biased towards BingChat, which combines search engines and productivity tools. It will add citation functionality next week and showcase the source for the generated code to ensure accuracy. And the ability to export and run code has been added, making it convenient for developers to work efficiently. By the way, a new dark mode was added, which caused cheers on site.

Although there was some confusion during the initial appearance, Google still showed us a constantly improving Bard. Unfortunately, the word we hear the most in the introduction is still 'we'll'. After all the features released today are actually installed and launched, Bard will be a comprehensive and mature tool.

DuetAI: Hand over office tasks to AI

Competitors for Copilot have arrived.

Google has introduced a new tool, DuetAI, for its workspace, allowing us to see the role of AI in cloud based Google Office Suite:

Document: Summarize the entire text and automatically write it;
Slide show: Cultural and creative images to improve efficiency;
Table: Analyze data and develop strategies.

In Google Docs, you can tell DuetAI what content you want to create and have it automatically generate articles. You can also have it generate different styles according to your needs.

The "ChatPDF" feature for "reading" articles and key questions has also been discontinued.

The ability to generate, expand, and adjust these words is also applied in Google Mail.

In terms of slides, DuetAI plays the role of an image generator, which can help you save time searching for images on the internet and enrich slides with original content.

In the table, DuetAI can automatically organize data based on your questions, making the data more standardized.

These capabilities can also be linked with other Google services, such as when you create a table in Bard, you can generate a table document with one click, making it convenient for future use.

DuetAI has also added a set of machine learning driven Google cloud development tools that can provide real-time code correction, generation, and support multiple development languages such as Go, Java, Javascript, Python, and SQL.

We don't need to wait for Bard to pass, we can now wait for DuetAI again. If you want to experience these features, you still need to join the waiting list and wait for Google to start testing in the corresponding region.

Overall, DuetAI lacks some of the stunning features of the first Copilot, and its functional demonstrations seem to be weaker. But Bard caught up in less than two months, and we hope DuetAI can bring more possibilities.

Google Search: Redefined, Comprehensive and Versatile

Future searches may look like this:

Driven by PaLM2, Google Search has also ushered in innovation with the launch of the new search engine SGE (Search Generative Experience). You don't need to divide a big question into multiple small questions anymore. After searching, you can gather the answers together. With the support of generative AI, you can make search do some heavy work.

At the conference, Google highlighted the improvement of generative AI in shopping. When you search for a product using Google, you will see its product description, comments, price, and display image.

Google, which has a product list of over 35 billion, is constantly updating at a rate of 1.8 billion yuan per hour. Through Google search shopping, you can obtain fresher and more reliable results.

Google believes that advertising is an important component of the internet, and they will continue to uphold their commitment to advertising transparency and ensure the distinction between advertising and search results.

Today, Google will open the registration of the search laboratory and access it in the coming weeks, supporting both the Google app and Chrome browser. The application link is as follows: labs. google.com/search

Google's AI anxiety can only be solved with AI

Some people jokingly claim that this Google I/O session should be called Google AI/O, with Microsoft and OpenAI leading the way. Google has lost a bit of moments of boasting about Amazon, but this is undoubtedly Google's most resolute self revolution in recent years.

Google, which claims to be AI first, has repeatedly missed opportunities due to its overly cautious and conservative strategy. Now Google has finally realized that it must act quickly, as many Google executives have said, to be "bold and responsible" and not hesitate because of fear of making mistakes.

Of course, AI is not a panacea, and the upgraded Bard may not change the trend of Google search being disrupted. But there is no doubt that AI will reshape search and completely change this way of interaction.

Bing, which has a much lower share than Google's, has reached over 100 million daily active users in just a few months through chat robots, engaging in over 500 million chats, of which 70% of users who try the new Bing chat feature will use it for search related tasks.

When the world's largest search engine decides to abandon the baggage of the past and embrace this trend completely, it may not only change Google's future, but also potentially disrupt the operating mechanism of the Internet and change the interaction of billions of users.

A brand new and simpler world will come.

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])

Mobile advertising space rental

Tag: Google expresses anxiety with AI sentences