Shulou(Shulou.com)11/24 Report--

On the morning of March 30, Beijing time, it is reported that OpenAI not only succeeded in surpassing Google with artificial intelligence chatbots, but also accidentally contributed to another almost impossible thing: this forced Alphabet's Google and DeepMind's two major artificial intelligence research teams to put aside years of feud and decided to join hands to catch up with OpenAI.

Software engineers at Google Brain are working with DeepMind employees to develop software that competes with OpenAI, according to people familiar with the matter. DeepMind is an artificial intelligence laboratory owned by Alphabet. Google launched the Bard chat robot to challenge OpenAI's ChatGPT, but accidentally flipped over at the demonstration site. Shortly after that, Alphabet decided to launch the joint project called Gemini.

The resignation of a prominent Google AI researcher, who warned Alphabet CEO Sandal Sundar Pichai and other executives about the development of the project, also had an internal impact on Bard's launch.

Google bought DeepMind in 2014, but the relationship between DeepMind and Google brain is more like a competitor than a partner. Both companies are scrambling to improve Google's products and services and seek scientific breakthroughs. But now, both of Alphabet's artificial intelligence labs believe that OpenAI has surpassed them. In addition, startups are poaching their key engineers and researchers.

OpenAI launched ChatGPT last November, a chat robot that mimics humans to provide answers and has become one of the fastest-growing apps of all time. Google's own chat robot, Bard, was just opened to some users last week, and it seems to be less effective than ChatGPT on some tasks.

Jeff Dean, Google's most senior head of artificial intelligence research and head of Google's brain, has played a technical role in the Gemini project, writing code to help Google develop machine learning models comparable to OpenAI's GPT-4, according to people familiar with the matter. This is enough to see the importance of Gemini within Alphabet.

The move also reflects Google's adjustment of product roadmaps in several sectors, such as search and cloud, to cope with the competition brought about by OpenAI's partnership with Microsoft. OpenAI has integrated technologies such as ChatGPT into Microsoft's products, and Microsoft has provided financial and computing support for the startup.

But Gemini is more or less forced to do it. The two artificial intelligence labs under Alphabet rarely cooperate with each other and rarely share code with each other. But people familiar with the matter said that because both sides now want to develop machine learning models that compete with OpenAI, and both need a lot of arithmetic support, Alphabet can only reluctantly force the two to cooperate.

The road to Bard, full of controversy and concern, the accidental rollover during the Bard demonstration revealed Google's problems. The development of the chat robot was controversial, leading Google's famous artificial intelligence engineer, Jacob Devlin, to join OpenAI soon after he resigned in January, according to people familiar with the matter. Devlin is the lead author of a groundbreaking paper on how to train machine learning models to improve their understanding of sets of sentences-an innovation that OpenAI has integrated into its own language model.

Devlin left after expressing concerns to Pichai, Dean and other executives on the Bard team. The Bard team received help from Google brain employees and used OpenAI's ChatGPT data to train its machine learning model. Specifically, Devlin believes that the Bard team seems to rely too much on ShareGPT-ShareGPT is a website on which people post their conversations with ChatGPT.

Some Google employees believe that using such chat logs would violate OpenAI's terms of service because the latter made it clear on its official website that it forbids the use of "output... to develop models that compete with OpenAI". Devlin also told executives that he was concerned that if he relied too much on ChatGPT chat logs shared on ShareGPT, Bard's answer would be too similar to that of ChatGPT.

Google stopped using ChatGPT data to train Bard after Mr Devlin expressed concern, according to people familiar with the matter. Part of the leadership of the Bard team is led by Sissie Hsiao, a vice president who previously headed Google's Assistant, a voice assistant similar to Apple's Siri.

A Google spokesman did not comment on Devlin's departure and concerns. A spokesman for OpenAI could not be reached for comment.

At the time of the search, attack and defense, OpenAI is already in the lead in the recent AIGC (generative artificial Intelligence) battle, but the exact business impact on Alphabet is unknown. Although ChatGPT can make many mistakes, millions of people have used it and the technology behind it to quickly generate blog posts, summarize meeting points, make spreadsheets, and accomplish a variety of tasks.

Revenue from the product has been small so far, but that could change. Microsoft has integrated ChatGPT-like features into its own search engine with the help of OpenAI, hoping to increase search market share. However, Microsoft has not yet taken much share from Google. But OpenAI seems to have ushered in a whole new era: applications can already understand the requirements of users through natural language.

Gemini's goal is to develop a large language model (a computer program that can understand and generate a similar human language) and will use up to 1 trillion parameters (an indicator that can be used to measure the computational complexity of machine learning models). It is reported that GPT-4 also has about 1 trillion parameters. Google's project requires tens of thousands of TPU. TPU is a microchip similar to Nvidia's GPU developed by Google, which is very suitable for training large machine learning models. However, it could take months for Gemini to produce real results.

At the same time, as Google scrambles to integrate artificial intelligence into its own products, other teams at the company are developing their own large language models. Pandu Nayak, the Google executive in charge of search sorting, has been developing a separate model to handle specific search requests. Google Cloud is also developing its own model to sell to cloud service customers, similar to those currently offered by OpenAI or Microsoft.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.