Another generative AI assistant will be "on board"! Google releases multi-modal Bard assistant, interactive AI era is near?

Original source: Financial Associated Press

Image source: Generated by Unbounded AI

**October 6 News, At a new product launch conference a few days ago, Google officially released the new generation of Android flagship phone Pixel 8 / Pro series, equipped with Tensor G3 chip, ** which can run more complex ML (machine learning) model, adding a number of AI enhancements to the new phone, such as reading web pages to users in different languages and "more natural" voices, and virtual assistants speaking more naturally.

Google points out that Pixel 8 Pro is the first phone to run Google's basic large model directly on the device, which requires 150 times the calculation of the largest ML model on Pixel 7.

At the same time, Google announced the launch of "Assistant with Bard" for Android and iOS devices, which combines the personal assistant function of mobile phones with generative AI. Users can interact with Bard Assistant through text, voice or images. ——In other words, it is multi-modal**.

For example, when a user asks "What important emails did I miss this week?" the Bard assistant will list the key points and specific content, and attach links to the specific corresponding emails. It can also help the user extract active addresses and display them in Google Maps. show.

If a user wants to post a photo of a puppy to social media, they simply summon the Bard Assistant floating dialog box and ask it to write the post. The Bard assistant will recognize the image and write the corresponding content.

Google said it will roll out Bard Assistant to early testers soon to get feedback and roll it out to the public in the coming months.

In addition, Mustafa Suleyman, co-founder of DeepMind, said in a recent interview that the current generative AI is only a transitional technical stage, and will next enter the interactive AI era**, where AI will be based on the different task needs of users. Arrange for other software and or contact a real person to get the job done.

He believes that the first wave of AI mainly focuses on classification - deep learning shows that humans can train AI to classify input data such as images, videos, audios, and languages. Humanity is currently in the second wave of "generative AI", which means "input data and generate new data." The third wave in the future will belong to "interactive AI". "Dialogue is the interactive interface of the future." Users not only click buttons and type text, but directly talk to AI. By then, interactive AI will be able to take independent actions. action.

Tianfeng Securities pointed out that the importance of scenarios during the implementation phase of C-side AI applications has become more prominent, with chatbots, AI companions and content production tool scenarios being the first to be implemented. The development speed and commercialization progress of AI applications in these scenarios may exceed expectations.

Analysts predict that AI iterations and later event catalysis will continue to accelerate. Starting from the second half of the year, the iteration speed of overseas giants' applications and models will increase significantly. The capabilities of general chatbots are expected to be further strengthened, which may lead to improvements in user experience and further growth in the number of users.

In addition, Huajin Securities added that the shift of large models from general to vertical scenarios is more of an exploration of commercialization and is the driving force for large models to move from training to inference. With the development and improvement of vertical large models, the application of large models is the key to opening up greater room for growth. Edge computing is a clear and huge incremental market, which has now reached the stage of industry implementation. Cloud computing companies, telecom operators, equipment manufacturers, CDN companies, etc. are all actively promoting the implementation of the industry.