Tech

Google Shows Off New AI Products to Take on Microsoft

Google drops a bombshell on the AI ​​circle

  At the 2023 I/O Developers Conference, Google released a series of new AI products. Among them, the new generation of large language model PaLM2 has attracted much attention. This large language model, which is regarded as “confronting GPT-4″, is there? What’s so special about it?
  Google said that PaLM2 will support Google AI chatbot Bard. PaLM2 is currently available to developers through Google’s PaLM API, Firebase, and Colab. PaLM2 includes 4 sizes, from small to large, Gecko, Otter, Bison and Unicorn, among which Gecko can work on mobile devices.
  The functionality brought by different size models means that PaLM2 can be fine-tuned to support the entire category of products in multiple ways, helping more users, and allowing Google to penetrate into different market segments.

  In terms of multilingual ability, PaLM2 has carried out more training, and the text spans more than 100 languages, which has greatly improved its ability to understand, generate and translate subtly differentiated texts, and solved problems such as idioms, poems and riddles. The ability to understand text, the processing threshold of this kind of text is relatively high. In addition, PaLM2 is capable of passing very high level language proficiency exams.
  In terms of reasoning ability, since PaLM2 is trained on web pages containing scientific papers and mathematical expressions, it has the ability to improve in logic, common sense reasoning, and mathematics.
  In terms of coding capabilities, PaLM2 supports more than 20 programming languages. Because it is pre-trained on a large number of publicly available source code datasets, it is good at popular programming languages ​​​​such as Python and JavaScript, and can also program in niche languages ​​​​such as Prolog, Fortran and Verilog.
  According to Google, users of PaLM2 include consumers around the world, as well as developers and businesses of all sizes.
  Pichai also announced that Google is already working on the development of the next-generation multimodal large model Gemini, which is very efficient in terms of tool and API integration, and has considerable innovative breakthroughs in memory and planning capabilities. It is still training among. “Gemini has demonstrated multimodal capabilities never seen before compared to other models. Once fine-tuned and rigorously tested for safety, it will also have a variety of sizes and functions and can be deployed in different products, applications and devices, enabling each Personal benefit,” Pichai said.
  From PaLM2 to Gemini, Google seems to have identified the pain points of GPT-4 this time, using multi-modal capabilities to hit the bull’s-eye, and using dedicated domain models to widen the gap. It can be said that it has fought a beautiful counterattack.
More than AI’s Google I/O Conference

  PaLM2 is the star at the 2023 I/O Developers Conference, but it is by no means all of Google’s ambitions, and the debut of new hardware has successfully attracted the attention of most netizens.
  Not surprisingly, Google released a new generation of Android phone Pixel7a, priced at $499, equipped with Google’s self-developed TensorG2 chip, running memory 8GB, more affordable than Pixel7 and Pixel7Pro.
  Interestingly, Google also released the first foldable mobile phone PixelFold, priced at $1,799, also equipped with TensorG2 chip, when fully unfolded, the mobile phone display is 7.6 inches in size, from this Wednesday, the new version of the tablet Pixel will accept pre-orders, with the same configuration The TensorG2 chip starts at $499 and is available in 128GB and 256GB versions.
  In terms of software services, Google announced the launch of A3VirtualMachines, paired with Nvidia H100, for training and serving the most advanced AI models.
  As early as March of this year, Google announced that it had added a series of AI functions to its Workspace office suite. At Wednesday’s I/O conference, Google dubbed DuetAI, which the company calls an “AI-driven collaborator.”
  DuetAI covers a collection of generative AI tools for Google’s productivity apps. DuetAI can also help Google’s cloud computing customers complete a range of tasks. DuetAI will allow cloud computing users and application developers to use AI-powered prompts to write code and build custom functionality into their Google Workspace applications.
  Google also announced a new feature called “Sidekick” for reading, summarizing and answering questions on documents in different Google apps.
Microsoft temporarily leads the battle for AI head

  From the beginning of the year to the present, the competition between Microsoft and Google has continued to intensify.
  In the artificial intelligence (AI) arms race between the two giants, the focus of the market is whether Microsoft can take advantage of ChatGPT’s Dongfeng to realize the conversion of profits, and whether Google’s advertising business, which is worried by the industry, can resist the opponent’s attack? From this financial report, we can get a glimpse of it.
  In contrast, Microsoft exceeded expectations across the board, and this time outperformed Google. Although Google performed well in key businesses, under the attack of Microsoft and OpenAI, the crisis facing Google search was also reflected in the financial report. According to the financial report released by Microsoft, the company’s revenue in the third fiscal quarter of fiscal year 2023 was US$52.857 billion, exceeding analysts’ expectations of US$51.02 billion, an increase of 7% compared with US$49.36 billion in the same period last year; The profit was US$18.299 billion, an increase of 9% compared with the net profit of US$16.728 billion in the same period of the previous year. According to CNBC reports, from the perspective of sub-items, Microsoft’s revenue in this quarter also exceeded market expectations.
  From the perspective of revenue composition, most of Microsoft’s sales still mainly come from software sales and cloud computing services. Boosted by the AI ​​wave led by ChatGPT, Microsoft’s intelligent cloud services achieved impressive growth in this quarter.
  According to the financial report for the first quarter of this year released by Alphabet, the parent company of Google, the company’s revenue in the first quarter was US$69.787 billion, a year-on-year increase of 3%. From the perspective of revenue composition, Alphabet’s main sales come from Google services and Google cloud services. The financial report shows that in the first quarter, Google’s service revenue was US$61.96 billion, a year-on-year increase of approximately 8%; Google’s operating profit was US$21.737 billion, a year-on-year decrease of approximately 1.1%.
  With the release of new Google products, a new arms race has begun.

error: Content is protected !!