Google hardened OpenAI, and the Silicon Valley war continued to escalate

Chinese entrepreneur

2024-05-15 12:04Published on the official account of Beijing "China Entrepreneur" magazine

Image source: Visual China

Text: "Chinese Entrepreneur" reporter Yan Junwen

Editor|Li Wei

A tech race is taking place in Silicon Valley.

On May 15th, Google held the 2024 I/O Developer Conference, AI was the theme of the whole audience, Google CEO Sundar Pichai and his colleagues released a new version of Gemini 1.5 Pro, which further expanded to 2 million Tokens on the window capacity of 1 million Tokens, and their ultimate goal is infinite context.

Google also showcased new search products, such as the "AI Overviews" feature that will be rolled out in the U.S. starting this week, with the search engine summarizing search results directly. In addition, Google also launched a new TPU, the sixth-generation TPU chip Trillium, which has a 4.7-fold increase in computing power performance compared to the previous generation of chips.

Just one day before this conference, OpenAI held a spring conference to release the latest GPT-4o multi-modal large model, which is faster and cheaper than GPT-4 Trubo. In the next month, Silicon Valley will also usher in the blockbuster conference of two tech giants - around May 21, Microsoft will hold its annual developer conference Build; Around June 10, Apple will hold its Worldwide Developers Conference. A big competition about AI started on the other side of the ocean.

With the rise of OpenAI and Google's failure in AR glasses, the outside world once commented that Google has become a machine built for "making money" and has lost its passion for innovation.

However, Google still owns Google Search, YouTube, cloud services, Gmail, Android and self-driving Waymo and other businesses, YouTube alone has 2 billion monthly active users, Android also has 2 billion monthly active users, it can be said that Google has the world's largest number of user scenarios, data and device interfaces.

Google-C shares edged up 0.6% to close at $171.93 after Google released the updated product, with a market capitalization of $2.12 trillion, up 23% from the start of 2024.

In the past fiscal quarter, Google showed strong competitiveness. In the first quarter of 2024, Google achieved revenue of $80.539 billion, a year-on-year increase of 15%, and non-GAAP net profit of $23.662 billion, a year-on-year increase of 57%.

Google struggled to hold its ground

As soon as the conference began, Pichai explained that more than 1.5 million developers are now using Google's native multimodal large model, Gemini. Gemini went live in December 2023, and it is believed to surpass GPT-4.

Pichai announced that Gemini 1.5 Pro is open to all developers worldwide and supports a preview of the Gemini 1.5 Pro model with 2 million tokens, with their goal of "infinite context".

At this developer conference, Google focused on search the most, which is Google's starting business, and now Google wants to redefine search, it wants to make search into a multi-round conversation reasoning, and users only need to ask.

With the posture of "redefining search", Pichai has launched a number of Gemini-based products in succession, such as the "AI Overview" function that summarizes the results of Google's search engine, and the AI search Ask Photos, which allows users to store images.

In the case of yoga, Google will help users find more top-ranked studios and introduce their services, and you can see the distance between each store. A Google search breaks down the bigger problem into sections, such as rankings, reviews, hours of operation, and much more.

But the biggest initiative to "redefine search" may not be in the hands of Google, but on OpenAI's side. At the beginning of this year, it was rumored that OpenAI planned to launch an AI search business, but at OpenAI's press conference on May 14, the service was not shown, which may be related to the cost of computing power.

Large, mature companies are still helpless to deal with the rise of innovative companies. Benchmarking against OpenAI, Google is still trying to show its presence.

At the event, Google also showcased a project called Astra, a universal human assistant developed by Google, in which Astra can recognize objects in the real environment through the phone camera, so that it can see, speak, and hear. This is similar to the GPT-4o function launched by OpenAI the day before, which realizes a new form of human-computer interaction. But Google doesn't seem to be satisfied with this, with Astra emphasizing the triple interaction between humans, machines, and the real world.

The biggest challenge for large models is to solve the problem of response time. Based on the Gemini model, Google said that they have developed agents that process information faster by continuously encoding video frames, and at the same time, combine video and voice input into the timeline of events to capture and effectively remember. These agents can better understand the user's environment and respond quickly in conversations, making the speed and quality of interactions more natural.

Gemini will also support real-time voice interaction, and real-time video interaction will be available later this year. In the coming months, Google will also launch Gems, a custom AI assistant similar to GPTs. It will be able to interact with the Google Family Bucket.

"Gems are easy to create, they can act as your yoga girlfriends, your personal sous sous chef, smart calculus tutors, peer reviewers of code." The relevant person in charge of Google said.

Later this year, the multimodal Gemini Nano model that can run locally will be available on Pixel phones, making it smarter and more privacy-preserving, alerting users if they receive a scam call, for example.

Without OpenAI, the above-mentioned innovative technology of Google can be called remarkable, but it is "both Shengyu and He Shengliang".

The Silicon Valley Tech Race Escalates

At the press conference, Google only mentioned one big tech company, and that was Nvidia. Pichai released the sixth-generation TPU (Tensor Processor, Google's dedicated chip for machine learning) chip Trillium, which is Google's own GPU chip, which is 4.7 times more powerful than the previous generation chip.

At the same time, Pichai said that Google Cloud will use Nvidia's latest Blackwell architecture GPUs in early 2025. Over the past few years, Google has invested 2 million miles of ground and submarine fiber.

In March of this year, Nvidia released the B100 chip and B200 GPU accelerator based on the "Blackwell" architecture, Nvidia founder Jensen Huang said that this is Nvidia's most powerful GPU chip to date, and after producing the first servers, Huang gave a server to OpenAI.

GPUs have become gold's hard currency for more than a year, with tech companies having to queue for months to get their hands on them, making Huang the most popular CEO in Silicon Valley, with people lining up to meet him, and Nvidia is a partner of all tech companies.

Behind the GPU boom, Silicon Valley is entering a new round of technology competition.

At the end of 2022, OpenAI released ChatGPT, and the entire Silicon Valley entered a state of euphoria, and even an AI street was born here, where five or six people can form a company and rely on an idea to obtain financing, such as Pika, founded by Stanford doctoral student Guo Wenjing and his classmates, which completed $55 million in financing in just a few months and launched the video generator Pika 1.0.

Relations between tech giants have become extremely strained, mainly because AI has reconstructed boundaries and big companies have become ambitious again:

Microsoft is back on search to support Bing's development; Meta opens up its VR headset operating system, Meta Horizon OS, to third-party hardware manufacturers, further encroaching on Apple's ecosystem. In response to the rise of OpenAI, Apple had to abandon its decades-long car-building plan and concentrate resources on AI.

Evolution is taking the form of crowds in Silicon Valley, and the convergence between large companies is also complicating the AI race.

Apple plans to work with OpenAI to bring OpenAI's technology to the iPhone, but Microsoft is the largest funder of OpenAI; Google is in a battle with OpenAI to support the AI evolution of its open-source system, Android, in order to compete with Apple's iOS system.

The wave of technology races that have swept the world is a bit of Silicon Valley in the seventies and eighties of the 20th century. At that time, native technology was emerging, and in 1968, Intel was founded; In 1969, AMD was founded; In 1976, Steve Jobs founded Apple Computer; In 1977, Oracle was founded; In 1984, Cisco was founded.

It's just that 50 years ago, it was hardware such as chips that created the entrepreneurial tide and created wealth; Now, AI and productization are the latest driving forces.

In the era of AI, China's Internet has also ushered in new development opportunities, but as most Chinese AI entrepreneurs say, although one company after another will transform or even exit in this wave, do not underestimate the vitality of Chinese entrepreneurs.

View original image 580K

• 

•