Apple Publishes Paper on Running Large Models in iPhones, 2024 May Become Key Year for AI Phones - How is the Generative AI Industry Developing?
-
According to Economic Observer, Apple recently published a paper on arxiv introducing a method to run LLMs (Large Language Models) on devices with limited DRAM capacity. The paper states that the computational intensity and memory requirements of LLMs pose significant challenges for DRAM capacity. The research builds a flash memory-based inference cost model, optimizing in two key areas: reducing data transfer from flash memory and enabling smoother block data reads. This breakthrough expands the applicability and accessibility of LLMs, potentially accelerating Apple's plans to integrate generative AI into iOS 18.
Generative AI has already transformed how we interact, with both domestic and international smartphone manufacturers like Samsung, Xiaomi, and Honor actively working to incorporate generative AI into mobile devices. Integrating large models into smartphones is becoming an industry trend. Samsung's new S-series phones, set to launch on January 18 next year during their spring event, will focus on AI features, with Galaxy AI also being introduced.
According to a recent report by market research firm Counterpoint Research titled 'Generative AI Smartphone Shipment Insights,' 2024 is projected to be a pivotal year for generative AI smartphones, with estimated shipments reaching 100 million units. By 2027, shipments are expected to grow to 522 million units. ChatGPT relies on large models, big data, and significant computing power, marking the starting point for general artificial intelligence and a turning point for strong AI. The integration of AI into smartphones is expected to become a new selling point, driving sustained growth in demand for high-end chips like computing power and expanding related market opportunities.
On the corporate front, according to Securities Times, Lechip Micro stated on an interactive platform that Samsung plans to launch the S24 AI phone in January 2024, and the company is a long-term stable supplier for Samsung. Lianchuang Electronics mentioned that its products can be applied to AI PCs and AI phones.
This year, China's generative artificial intelligence market is expected to exceed 10 trillion yuan. Generative AI is accelerating its penetration into four major industries: manufacturing, retail, telecommunications, and healthcare.
Data shows that this year, the adoption rate of generative AI by enterprises in China has reached 15%, with the market size approximately 14.4 trillion yuan. The adoption rate of generative AI technology in the four major industries—manufacturing, retail, telecommunications, and healthcare—has experienced rapid growth. Experts predict that by 2035, generative AI could contribute nearly 90 trillion yuan in economic value globally, with China accounting for over 30 trillion yuan, representing more than 40% of the total.
With the increasing demand for digital content, generative AI (AIGC) technology has been widely implemented this year, further integrating the digital economy with the real economy. AIGC has opened a new frontier in human technological vision.
Today, everyone is talking about AIGC, and enterprises are placing greater emphasis on the importance of data. Large models have driven an exponential increase in computing power demand. How to effectively utilize large models and implement AIGC technology in real-world industries is a topic many tech companies are actively exploring and practicing.
Bill Gates: Generative AI Has Reached Its Limits, the Next Breakthrough Will Be Explainable AI
Bill Gates recently stated in an interview with Germany's Handelsblatt that although many at OpenAI, including Sam Altman, believe GPT-5 will significantly outperform GPT-4, he thinks there are reasons to believe generative AI has already reached its limits.
Historical data shows that compared to the arms race in large models, Bill Gates is more optimistic about AI Agents. He has said that in the near future, any internet user could own an AI Agent. Agents will not only change how individuals interact with computers but also disrupt the software industry, marking the most significant computing revolution since the shift from command inputs to clicking icons.
Generative AI (AIGC, Artificial Intelligence Generated Content) is a key milestone in the transition from AI 1.0 to AI 2.0.
The convergence of technologies like GAN, CLIP, Transformer, Diffusion, pre-trained models, multimodal techniques, and generative algorithms has fueled the explosion of AIGC. Continuous algorithmic innovation, the qualitative leap in AIGC capabilities driven by pre-trained models, and the diversification of AIGC content through multimodal approaches have endowed AIGC with more general and robust foundational capabilities.
From the progression of computational intelligence to perceptual intelligence and then to cognitive intelligence, AIGC has opened the door to cognitive intelligence for human society. By training on large-scale datasets, AI has acquired knowledge across multiple domains. With appropriate adjustments, these models can now perform tasks in real-world scenarios.
AIGC is a milestone for human society and artificial intelligence. In the short term, it changes basic productivity tools; in the medium term, it alters social production relations; and in the long term, it drives a qualitative breakthrough in overall social productivity. In this transformation of productivity tools, production relations, and productivity, the value of data as a production factor is greatly amplified. AIGC elevates data to the status of a core resource of the era, accelerating the digital transformation of society to some extent.
How is the generative AI industry developing?
Since the beginning of this year, general artificial intelligence, represented by cognitive large models, has sparked a global frenzy. International companies like OpenAI, Microsoft, and Google have continuously increased their investments, while domestically, a "thousand-model battle" has erupted, with numerous high-tech enterprises competing in research and development.
According to the China AI Large Model Map Research Report released by the Ministry of Science and Technology, China and the US significantly lead in the number of large models released globally, accounting for over 80% of the total.
At the same time, large models are seen as a new engine for corporate transformation. Recently, McKinsey's report The Economic Potential of Generative AI: The Next Productivity Frontier revealed that applying the 63 analyzed generative AI use cases across industries could add $2.6 trillion to $4.4 trillion annually to the global economy. This estimate does not include all generative AI applications, and if unstudied applications are considered, the economic impact of generative AI could double.
Specifically, industry players generally believe that the demand-side generative AI market is about to experience explosive growth, potentially forming two trillion-dollar markets in both consumer (C-end) and enterprise (B-end) sectors. Currently, corporate demand for generative AI is surging dramatically. General-purpose models will inevitably transform the traditional AI market, and productivity improvements will significantly boost industrial development.
From the progressive development of computational intelligence, perceptual intelligence to cognitive intelligence, AIGC has opened the door to cognitive intelligence for human society. Through training on large-scale datasets, AI has acquired knowledge across multiple domains. With appropriate model adjustments, it can now accomplish tasks in real-world scenarios.
Notably, this year's popularity of ChatGPT has sparked an AI entrepreneurship frenzy. Since March, AIGC job postings have surged significantly, especially driven by the large model entrepreneurship wave. Major tech companies and unicorns have entered the field, competing for talent, causing AIGC job postings to nearly triple in April. In the second half of the year, new AIGC job postings stabilized.
From January to August 2023, average salaries for AIGC positions showed an upward trend. By March 2023, the average monthly salary for new AIGC positions exceeded ¥50,000, reaching ¥52,472 - an 11.6% increase from January's ¥47,015. Starting June, the average monthly salary for new AIGC positions stabilized around ¥60,000.
What new challenges and opportunities will this industry face in the coming years?
The report conducts an in-depth analysis of the supply-demand dynamics, current development status, and evolutionary changes within sub-sectors of China's AIGC industry. It particularly focuses on:
- Comparative analysis of domestic and international AIGC industry development landscapes
- Strategies to address industry challenges
- Development recommendations for sectoral growth
- Competitiveness assessment frameworks
- Investment analysis and future trend projections