Finishing | Zheng Liyuan
出品 | CSDN（ID：CSDNnews）
The company is willing to spend $2.7 billion (about 18.9 billion yuan) to rehire a former employee who has long left the company - this kind of cool plot is unthinkable for ordinary workers, except for the protagonist of this incident, Noam Shazeer, a well-known researcher in the field of AI: Recently, Google announced that it would rehire its former employee Noam Shazeer for about $2.7 billion, which has attracted widespread attention in Silicon Valley and the entire tech world.
To be precise, the $2.7 billion is the licensing fee Google paid for Noam Shazeer's startup Character.AI technology, but according to people familiar with the matter, the deal has an unofficial requirement: Google wants Noam Shazeer back and work for him again.
Dissatisfied with Google's "being too conservative", it was decided to create Character.AI
Noam Shazeer, 48, is known to have joined Google as early as 2000 and was one of Google's first hundreds of employees. His first major project was the development of a system to improve the spelling correction capabilities of search engines. Shortly after taking on the project, Noam Shazeer made a request to Eric Schmidt, then Google's CEO, to use thousands of chips: "I'm going to solve the common sense problem this weekend. ”
Although the effort failed at the time, Eric Schmidt looks back on it in 2015 with anticipation for Noam Shazeer and his belief in his ability to build AI technology with human-level intelligence: "If there is one person in the world who can do this, it must be him." ”
By 2017, Noam Shazeer, along with seven other researchers at Google, had published a paper called "Attention is All You Need," detailing a computer system that could reliably predict the next word in a sequence at human prompts — a paper that would later become the basis for generative AI technology.
Not only that, but Noam Shazeer later collaborated with a Google colleague, Daniel De Freitas, to develop a chatbot called Meena that supports conversations on a variety of topics. At the time, Noam Shazeer predicted in a widely circulated memo, "Meena Eats the World," that the bot could replace Google's search engine and bring in trillions of dollars, according to people familiar with the matter.
In 2020, Google first showed Meena in a research paper that said the chatbot had trained 40 billion words from public conversations on social media. During the same period, OpenAI also developed a similar model, GPT-2. However, OpenAI did not make the program publicly available at the time due to concerns that it could be used to generate large amounts of deceptive, biased, or abusive content.
At Google, the team responsible for the Meena project also wants to release their tools, even in a limited form like OpenAI. To that end, they have repeatedly urged Google to allow outside researchers access to the chatbot in an attempt to integrate it into Google Assistant, and have asked Google to provide a public demo. But former Google product manager Gaurav Nemade revealed that Google's leadership has repeatedly rejected the proposals on the grounds that "the chatbot does not comply with the company's principles regarding AI safety and fairness."
So in 2021, dissatisfied with Google's "too conservative" approach to AI deployment, Noam Shazeer and Daniel De Freitas decided to leave Google. Although CEO Sundar Pichai stepped in to retain them, he also did not commit to making the chatbot they developed publicly available. Eventually, the duo left Google at the end of 2021 and founded a new company, Character Technologies Inc., in November of the same year focused on developing chatbot technology.
After 3 years of separation, Google issued an offer to return
In September 2022, Character AI, an AI chatbot website based on large-scale natural language training, was born. What's special about this site is that users can simulate chatting with many of the characters they've created, whether they're celebrities, historical figures, or fictional characters, through Character AI.
Soon after, OpenAI's release of ChatGPT proved the huge public demand for AI-powered chatbots, and also made Character AI the second most popular AI chatbot website in the United States after ChatGPT.
In this regard, Noam Shazeer once revealed in an interview: "This has caused some buzz within Google. Although he didn't elaborate, he was still sarcastic about Google's refusal to release its AI tools: "We thought it might be smoother to launch a product as a startup." ”
However, Character.AI despite some success in the capital markets and the closing of a $150 million funding round in March 2022, the company still faces significant pressure on technology development costs. In order to compete with giants such as OpenAI and Microsoft, Character.AI had to accelerate technology iterations, but before establishing a stable revenue stream, how to balance high costs became a major problem for the company.
At the same time, Noam Shazeer and his team faced another challenge: preventing users from engaging in excessive "romantic role-playing." He said that while hopefully chatbots can help people who feel lonely or depressed, this excessive emotional interaction is not in line with the company's core vision.
In search of greater capital support, Character.AI has tried to reach out to potential buyers, including Meta. At this critical juncture, Google's parent company, Alphabet, suddenly stepped in.
According to people familiar with the matter, Google paid about $2.7 billion for licensing Character.AI's technology, but the more important reason was actually to bring Noam Shazeer back to Google — an arrangement within Google that was seen as the main reason the company was willing to pay a huge fee.
That is, with this $2.7 billion, Google not only gained access to Character.AI technology, but also brought one of the top researchers in the field of AI back to the company. But people working on AI at Google say they don't know how the company will leverage the technology it gets from Character.AI.
Responsible for leading the development of Google Gemini
According to the Wall Street Journal, Google co-founder Sergey Brin played a key role in the events that led to Noam Shazeer's return to Google. At a recent conference, Sergey Brin publicly stated that in the past companies were overly cautious in deploying AI applications, and now Google is developing and rolling out AI technology as quickly as possible: "By the way, Noam is back at Google, which is great. ”
According to LinkedIn sources, Noam Shazeer returned to Google in August as Google's vice president, refocusing on AI research and leading several researchers, including Daniel De Freitas. Instead of managing a company with hundreds of employees like he did in Character.AI, he returns to the heart of technology research and focuses on driving Google's next-generation AI technology.
In addition, less than a month after his return, he has been appointed to a very important position: working with Jeff Dean, chief scientist at Google DeepMind, and Oriol Vinyals, head of the deep learning team, to build Gemini, Google's most powerful AI technology. With the return of Noam Shazeer, many media outlets believe that Google has not only strengthened its capabilities in the field of AI technology, but also laid a solid foundation for future leadership in the field of generative AI.
Reference Links:
https://www.wsj.com/tech/ai/noam-shazeer-google-ai-deal-d3605697
https://www.wsj.com/articles/google-ai-chatbot-bard-chatgpt-rival-bing-a4c2d2ad?mod=article_inline