Google Claims Gemini Ushers in the Era of Native Multimodality: 2024 AI Industry Prospects Research and Outlook
-
Media reports indicate that a demo video produced by Google DeepMind exaggerated Gemini's capabilities. The widely circulated video showed Gemini rapidly identifying objects in images with human-like voice responses, showcasing impressive multimodal functionality.
However, some media outlets and users found that Gemini couldn't achieve the effects shown in the video during actual use. In fact, Google DeepMind admitted the demo wasn't conducted in real-time or with live voice interaction. This marketing effort has even drawn criticism from Google employees.
Notably, Google DeepMind CEO Demis Hassabis emphasized in media interviews that Gemini represents a new breed of AI—"native multimodality"—distinct from existing "stitched-together multimodal" models. It's trained from the ground up using multiple modalities (e.g., audio, video, and images), potentially paving an unprecedented path in AI with significant breakthroughs.
AI's multidisciplinary and highly complex nature requires strengthened research, coordinated planning, collaborative innovation, and steady advancement. Enhancing original innovation capabilities should be prioritized, with key core technologies as the main focus, alongside strengthening fundamental theoretical research. Accelerating next-generation AI development supports achieving high-level technological self-reliance.
In the large model industry, China closely follows global trends. Since 2021, China has seen a surge in large model releases, producing several influential models. Influenced by ChatGPT, domestic large models entered rapid development in 2023, leading to a "hundred-model competition."
On November 7, OpenAI's first developer conference unveiled GPT-4 Turbo, a standardized AI Agent with initial planning and tool selection capabilities. It can autonomously access the internet, perform data analysis, generate images, and more, evolving into a unified intelligent entity.
2024 AI Industry Prospects Research and Outlook
Technology can be a development tool but also a source of risk. Centering on modern industrial system construction and supply-side structural reform, AI should deeply integrate with primary, secondary, and tertiary industries, driving industrial transformation and cultivating new growth points.
Seizing the new technological revolution wave and accelerating AI R&D and application is crucial for gaining global tech competition initiative. It's a strategic resource for China's technological leapfrogging, industrial optimization, and productivity advancement.
Artificial intelligence is a strategic technology leading the new round of technological revolution and industrial transformation, serving as the main engine for empowering the digital and green dual transformation of the economy and society, with a strong 'leading goose' effect.
Wan Jinbo, a researcher at the Chinese Academy of Sciences' Institute of Science and Technology Strategy and Policy Research, pointed out in an article that China should, under the current situation of leading in data elements while keeping pace in computing power and algorithms, attach great importance to consolidating the underlying technologies and basic software of artificial intelligence, create a safe and trustworthy AI ecosystem, and accelerate the application of AI.
In the process of Chinese-style modernization, the technological innovation of artificial intelligence is one of the important forces driving China's scientific and technological innovation. As the most representative disruptive technology, while artificial intelligence brings potential huge development dividends to human society, its uncertainty also brings many global challenges.
Ten Major Trends in Artificial Intelligence Development
KPMG recently jointly released with Zhongguancun Industrial Research Institute the Prospect of Artificial Intelligence's Comprehensive Transformation: The Arrival of the Tipping Point (2023), proposing ten major trends in the future development of the artificial intelligence industry, and deeply analyzing the development status and core driving forces of each trend, in order to provide useful reference for the AI industry, anchor opportunities, and resolve challenges.
Trend 1: Multimodal Pre-trained Large Models Become the Standard in the AI Industry
In terms of algorithms, the development of pre-trained large models originated in the field of natural language processing (NLP) and has now entered the stage of 'hundred-model battle'. It is expected that as large model innovation shifts from single-modal to multimodal, multimodal pre-trained large models will gradually become the standard in the artificial intelligence industry. In the future, in terms of large models empowering industries, China's large models are very likely to catch up from behind, which will also be one of the key factors in the competition of domestic large models.
Trend 2: The Increasing Scarcity of High-Quality Data Will Drive a Leap in Data Intelligence
In terms of data, the training of large models requires a large amount of high-quality data, but there are still certain problems in data quality, including data noise, data missing, and data imbalance. This will affect the training effect and accuracy of large models. It is expected that the continuous emergence of high-quality data requirements in the field of large models will drive a comprehensive improvement in data in terms of large-scale, multimodal, and high-quality dimensions, and data intelligence-related technologies are expected to achieve leapfrog development.
Trend 3: Ubiquitous Intelligent Computing Power Accelerates New Computational Paradigms
In terms of computing power, new hardware and architectures are emerging rapidly, potentially overturning existing chips, operating systems, and application software. The future may realize "everything as data," "no computation without data," and "no intelligence without computation" – meaning intelligent computing power will be omnipresent, characterized by four features: heterogeneous diversity, software-hardware coordination, green intensification, and cloud-edge-device integration.
Trend 4: AIGC Applications Penetrate All Scenarios
In AIGC (AI-generated content) applications, which originated in digital content creation, the transition from single-modal to multi-modal digital content creation is taking shape. It's expected to further enhance human creative efficiency, enrich digital content ecosystems, and usher in an era of human-machine collaborative creation. Any scenario requiring creativity and new content could be redefined by AIGC, with full-scenario penetration imminent.
Trend 5: AI-Driven Scientific Research Transitions from Single-Point Breakthroughs to Platformization
In AI4S (AI for Science) applications, the field is accelerating from single-point breakthroughs toward platformization. During the "single-point breakthrough" phase, AI4S was researcher-led, with market focus on data, models, algorithms, and methodological originality. Initial "single-point applications" in specific tasks/scenarios demonstrated practical solution value.
The business model determines whether the ecosystem can complete the value creation-realization cycle. Currently, the AGI ecosystem's business model is primarily represented by AIGC-related models, mainly manifesting as MaaS (Model-as-a-Service).
On the supply side, a foundational ecosystem of "general large models + domain-specific large models + industry-specific large models + enterprise/individual small models" is expected to emerge, driving AI adoption across industries and ultimately achieving AGI.
AI Could Generate Up to $340 Billion Annual Profit for Wall Street
McKinsey Global Institute released a report on December 5 stating that Wall Street banks using generative AI tools could increase profits by up to $340 billion annually through productivity improvements, equivalent to a 9% to 15% growth in operating profits. The report noted that corporate and retail banking would benefit the most.
According to McKinsey's research on 63 industry use cases, generative AI tools may eventually take over most repetitive tasks currently performed by human workers.
In the fiercely competitive market, making timely and effective decisions is crucial for businesses and investors. The AI industry report compiled by China Research Network provides a detailed analysis of the development status, competitive landscape, and supply-demand dynamics of China's AI industry. It also examines the opportunities and challenges facing the sector from the perspectives of policy, economic, social, and technological environments.
Additionally, the report reveals potential market demands and opportunities, offering accurate market intelligence and scientific decision-making support for strategic investors and corporate leadership in planning. It also serves as valuable reference material for government departments.