Google's Big Model Transformation: Gemini Ultra vs. OpenAI
楚一帆
发表于 2024-2-9 16:03:47
229
0
0
21st Century Business Herald reporter Kong Haili reports from Beijing
The battle for big models is raging, and Google, which has been snatched the first mover advantage by OpenAI, is catching up.
On the night of February 8th Beijing time, Google took a series of major actions: Bard, which originally provided AI interaction functions, was officially renamed Gemini, and Gemini model functions will be fully introduced into Google products, including Workspaces (such as Documents, Sheets, Slides) and Google Cloud.
In order to facilitate user access to Gemini, Google has specially introduced the Gemini App for Android models, and the iOS system can directly use Gemini from the Google App.
The Gemini Ultra, known as Google's strongest AI model, has also been officially launched. Users can experience the Gemini Ultra through the Google One AI Premium program, which costs $19.99 per month.
This design is quite similar to OpenAI's ChatGPT Plus subscription service, which offers the GPT-4 version for a monthly service fee of $20, with a clear sense of confrontation.
After more than a year of accumulation, Google intends to regain its lost ground in the big model competition with the new version of Gemini, and even regain its leading rhythm.
Gemini has undergone a facelift
In November 2022, ChatGPT emerged, revolutionizing the way human-computer interaction works. The battle of big models became unstoppable, and OpenAI and Microsoft took the lead.
As a pioneer in artificial intelligence, Google launched the chatbot Bard in March 2023 and the Gemini model in December, in order to benchmark against ChatGPT. At that time, according to benchmark tests released by Google, Gemini scored above GPT-4 in 30 out of 32 performance test criteria.
However, users have mixed reviews of Gemini, and its demo videos have been widely questioned as a carefully edited collection with suspicions of hype marketing.
Subsequently, Oriol Vinyals, Vice President of Research at Google DeepMind, admitted that all human prompts and AI outputs in the video were authentic, but were shortened for simplicity.
Now, by renaming Bard Gemini, Google has extended the big model horizontally.
According to Google CEO Sandal Pichai, the Gemini model is about to become widely available in Google products, including Workshop and Google Cloud.
It is reported that over one million people are using features such as "Help Me Write" from Duet AI in Google Workspace to improve work efficiency and creativity. Next, the Duet AI functionality will be uniformly integrated into Gemini.
For Google Cloud users, Duet AI will also be changed to Gemini in the coming weeks. Gemini will play a role in helping companies improve productivity, improve developer coding speed, and protect organizations from cyber attacks.
Developers are the testing ground for every major technological advancement, and they will also play an important role in the Gemini ecosystem. Sandal Pichai stated that hundreds of thousands of developers and businesses are already using the Gemini model for construction, and Google is also training the next iteration of the Gemini model.
In order to make it more convenient for users to use the Gemini model on their phones, Google has launched a new app specifically for Android.
Similar to Siri's role in iOS, Google's app developed for Android is positioned as a mobile assistant, just like using "Hey, Siri" to wake up a phone, "Hey, Google" will also become a new way to call the Android system.
Android users can call Gemini by pressing the power button, swiping from a corner of the screen, or shouting "Hey Google". This will bring a brand new overlay experience, where users can easily use the large model function on the screen by waking up Gemini, such as generating titles for newly taken photos or asking questions about articles they are reading.
According to Google Assistant& Sissie Hsiao, Vice President and General Manager of Bard, stated that many of the features originally provided by Google Assistant will be provided through the Gemini App, including setting timers, making phone calls, and controlling smart home devices.
And iOS phones that already have Siri assistants will be able to use Gemini directly from the Google App in the coming weeks.
Starting from the evening of February 8th Beijing time, Gemini will launch an English version on Android and iOS phones in the United States. Starting next week, more countries will gradually be able to access Gemini in English, Japanese, and Korean.
Gemini Ultra New Experience
The legend of Gemini Ultra has already laid the groundwork and is hailed as Google's most powerful and advanced model version, carrying Google's expectations and user curiosity.
Ultra 1.0 is the first model to outperform human experts in MMLU (Massive Multi Task Language Understanding), which uses a combination of 57 subjects including mathematics, physics, history, law, medicine, and ethics to test the knowledge level and problem-solving ability of large models.
Now, the Ultra 1.0 product has finally landed. In order to keep pace with the Gemini universe, the Ultra 1.0 version is named Gemini Advanced.
Sissie Hsiao said that in third-party evaluations, users feel that Gemini Advanced is currently the most popular chatbot.
According to Google's official introduction, Gemini Advanced not only supports longer and more detailed conversations, but also can better understand contextual prompts, and can handle more advanced programming solutions, more complex logical reasoning, deeper data analysis and other functions.
Gemini Advanced has now launched an English version in over 150 countries and regions, and over time, Google will expand it to more languages.
But just like OpenAI's paid service for GPT-4, Google has also set payment standards for Gemini Advanced.
Google's new Google One AI Premium program will include Gemini Advanced at $19.99 per month, with a free trial for the first two months.
That is to say, the Google One Premium program, which originally charged $9.9/month, will be offered as an additional bonus in addition to existing services, with Gemini Advanced offered at a bundled price of $19.9/month.
In addition, AI Premium subscribers will soon be able to use Gemini in Gmail, Documents, Slides, Sheets, and more (formerly known as Duet AI).
Google also emphasized that they will continue to boldly and responsibly promote AI technology to the world in parallel. Moreover, in order to reduce issues such as unsafe content or bias, Google places great emphasis on integrating security into AI products. Before launching Gemini Advanced, they conducted extensive trust and security checks and improved the model's capabilities based on feedback.
In the deployment of AI big models, Google did miss out on the first mover opportunity over the past year. Now, with the release of the Gemini Ultra model, there may be significant upgrades and improvements compared to Gemini 1.0. For more real answers, please refer to the follow-up feedback from Google One AI Premium members.
If there is not much difference in capabilities between the two parties, Gemini Advanced at $19.9 per month will have many ecological advantages over GPT-4 at $20 per month.
Google, which is busy training big models, is still far from its goal of providing the world's most advanced, secure, and responsible AI. However, Google relies on the largest search database and has a massive user advantage. These factors may add some buffs to its competition for big models.
CandyLake.com 系信息发布平台,仅提供信息存储空间服务。
声明:该文观点仅代表作者本人,本文不代表CandyLake.com立场,且不构成建议,请谨慎对待。
声明:该文观点仅代表作者本人,本文不代表CandyLake.com立场,且不构成建议,请谨慎对待。
猜你喜欢
- The delivery time for two iPhone 16 models has been shortened! What signal?
- Apple lowers prices of various iPhone models in India
- Baidu Shen Dou: Upgrade computing platform capability for 100000 card computing power cluster, Wenxin large model daily usage exceeds 700 million times
- Meta releases heavyweight new products: $299 Quest 3S headset, AR glasses prototype, multimodal AI model
- Baidu World 2024 will be held on November 12th, and the daily average number of adjustments for the Wenxin large model has exceeded 700 million times
- 挑战Model Y 蔚来的品牌下沉“阳谋”
- Ford CEO tired of making 'boring' car models, personalized and electrified products become 'new favorites'
- Dialogue | Baidu Li Tao: The overlap between automotive intelligence and the wave of big models is a historical inevitability
- Boeing announces 10% layoffs, first delivery of 777X model postponed to 2026
- Faraday Future plans to launch the first model of its second brand by the end of next year
-
苹果知名分析师郭明錤周四(10月31日)在社交媒体上发文表示,苹果明年可能会减少对芯片制造商博通Wi-Fi芯片的依赖,并推出自己的处理器。 郭明錤在社交媒体平台X上写道,“在2025年下半年的新产品(例如iPh ...
- uturn
- 昨天 14:42
- 支持
- 反对
- 回复
- 收藏
-
10月30日,小鹏汽车生态企业小鹏汇天宣布,旗下分体式飞行汽车“陆地航母”即将亮相2024中国航展,11月12日将在中国航展第二展区(斗门莲洲)进行全球首次公开飞行,同时“陆地航母”也将在珠海国际航展中心8号 ...
- yxtianyouyou
- 前天 11:43
- 支持
- 反对
- 回复
- 收藏
-
交易所监管文件显示,当地时间11月1日,亚马逊创始人杰夫·贝索斯拟出售约1635万股亚马逊股票,预计套现约30.5亿美元。今年7月,贝索斯已申请额外出售约2500万股亚马逊股票,按当时股价计算可套现约50亿美元。 ...
- blueskybb
- 13 小时前
- 支持
- 反对
- 回复
- 收藏
-
近日,凯撒海湾目的地(山东)运营管理有限责任公司(简称“凯撒海湾”)与携程旅悦集团签署战略合作协议,双方将围绕“海上目的地运营”、“旅游产品与服务创新”、“研学旅行”、“日韩及海外旅游市场开拓”等 ...
- llyyy2008
- 11 小时前
- 支持
- 反对
- 回复
- 收藏