The Current State and Impact of AI Technologies in 2024
TABLE OF CONTENTS
- Summary
- Advancements in AI Hardware
- Legal and Ethical Issues in AI
- AI Applications in Various Sectors
- Societal and Environmental Impacts of AI
- User Experience (UX) and AI
- AI in the Context of Global Development
- Conclusion
1. Summary
The report 'The Current State and Impact of AI Technologies in 2024' comprehensively explores the advancements in artificial intelligence (AI) hardware and software, legal and ethical issues, applications across different sectors, and societal implications. Among the key findings, NVIDIA's preparation and anticipated high demand for its H20 GPUs in China, advancements in chip technology with ASIC Sohu outperforming NVIDIA GPUs, and SK hynix's AI memory solutions stand out. The report also addresses legal concerns, including the New York Times' copyright lawsuit against OpenAI and ethical debates surrounding AI-powered medical chatbots like ChatGPT. In the business and education sectors, AI as a Service (AIaaS) and AI tools are revolutionizing efficiency, with large language models (LLMs) like GPT-4 significantly enhancing various applications. The societal impact of AI, from increased energy consumption by data centers to mental health issues exacerbated by social media, also garners attention. EU regulations, such as the Digital Services Act, aim to mitigate harmful content and promote user protection. Finally, the report discusses AI's role in enhancing user experiences and South Korea's notable advancements in AI technology.
2. Advancements in AI Hardware
2-1. Demand for NVIDIA GPUs in China
When NVIDIA started preparing the H20 GPU for China, the company anticipated a great demand from sanction-obeying GPUs. According to a Financial Times report, citing SemiAnalysis, NVIDIA will sell over one million H20 GPUs in China, far surpassing the 550,000 Ascend 910B chips from Huawei. With an average price of $12,000 per NVIDIA H20 GPU, this translates to an estimated $12 billion in revenue. The higher memory configuration and better performance in real-world applications make NVIDIA H20 chips a preferable choice over Huawei's Ascend 910B.
2-2. ASIC Sohu's Performance Over NVIDIA GPUs
Etched, a company specializing in application-specific integrated circuits (ASICs), introduced a new chip called Sohu, designed explicitly for transformer models. Sohu outperforms NVIDIA's H100 and the latest B200 'Blackwell' GPUs significantly, delivering up to 500,000 tokens per second, compared to the 25,000 tokens/s and 43,000 tokens/s provided by eight-cluster configurations of NVIDIA H100 and B200 GPUs, respectively. Sohu's architecture enables 90% FLOP utilization, compared to traditional GPUs' 30-40%. This efficiency addresses power waste and supports AI applications requiring real-time output.
2-3. SK hynix's AI Memory Solutions
SK hynix showcased its AI memory solutions at HPE Discover 2024 and COMPUTEX Taipei 2024. Key products include HBM3E, which offers data processing speeds of 1.18 terabytes per second and superior heat dissipation. Also featured was CXL Memory Module-DDR5 (CMM-DDR5), promising up to 50% more system bandwidth and 100% greater capacity compared to systems with only DDR5 DRAM. SK hynix's AI memory solutions, including the groundbreaking HBM3E and advanced CXL technologies, are tailored to meet the high-performance demands of AI systems.
2-4. Samsung's 1nm Process Mass Production Plan
Samsung plans to advance its 1nm process mass production timeline from 2027 to 2026. This follows the successful mass production of its 3nm wafer in June 2022 and anticipated mass production of the second-generation 3nm process in 2024, and the 2nm process in 2025. These developments come amid speculation denied by Samsung concerning the performance issues of their HBM3 and HBM3E chips, which reportedly run too hot for NVIDIA's validation requirements. Samsung has addressed these claims, asserting that their process of optimizing products through close collaboration with customers is on track.
2-5. NVIDIA's Challenges with Samsung's HBM3 Chips
According to multiple sources cited by Reuters, NVIDIA has faced issues validating Samsung's HBM3 chips. The chips reportedly run hot, exacerbating NVIDIA's existing cooling challenges for higher-end products. Power consumption is another reported issue. Samsung has attempted to get HBM3 and HBM3E parts validated by NVIDIA since 2023, but full validation has yet to be achieved. Presently, NVIDIA is sourcing HBM3 chips from Micron and SK Hynix, the latter having supplied HBM3 since mid-2022 and HBM3E since March 2024. Samsung maintains that these reported failures are untrue, attributing them to ongoing customization for customers' needs.
3. Legal and Ethical Issues in AI
3-1. New York Times vs. OpenAI Copyright Case
The New York Times (NYT) has filed a lawsuit against Microsoft and OpenAI, accusing them of copyright infringement. The NYT alleges that OpenAI used data scraped from the internet, including NYT articles, to train its generative AI models without proper authorization. OpenAI responded with a motion to dismiss, introducing a 'hired gun hacker' defense. This defense claims that the evidence presented by NYT, consisting of one hundred examples of ChatGPT generating responses identical to NYT articles, was fabricated by an expert hacker allegedly hired by NYT. Additionally, the case highlights procedural irregularities and the substantive copyright issues at play. The lawsuit, if successful, could significantly impact OpenAI and other generative AI companies by challenging their data harvesting practices.
3-2. Ethical Concerns in AI-Powered Medical Chatbots
ChatGPT, developed by OpenAI, has emerged as a popular AI chatbot with potential applications in the medical field. However, several ethical concerns have been raised regarding its use. Unlike specialized medical chatbots, ChatGPT has not been trained on specific medical datasets, leading to potential inaccuracies in the health information it provides. This raises significant patient safety, privacy, and cybersecurity concerns. Researchers argue that while ChatGPT could revolutionize medical interactions, it must undergo stringent testing, evaluation, and regulation to ensure ethical and accurate performance. The current knowledge cutoff for ChatGPT is September 2021, making its data potentially outdated. As AI-driven medical tools like ChatGPT continue to evolve, establishing ethical guidelines, quality assurance systems, and international standards will be crucial to addressing these challenges.
3-3. Debates on AI Chatbots in Education
The integration of AI chatbots like ChatGPT into educational settings has ignited intense debates. Critics argue that these technologies could facilitate academic dishonesty and propagate misinformation. Several public school districts, including those in Seattle and New York City, have banned AI chatbots on campuses. However, supporters contend that AI chatbots can enhance learning by providing interactive and informative experiences. They advocate for teaching students to critically evaluate AI-generated content to develop essential digital literacy skills. A national survey by the Walton Family Foundation revealed that 75% of students believe ChatGPT can help them learn faster, and 73% of teachers agree. These findings highlight the potential of AI chatbots to transform education, provided that their use is carefully monitored and integrated within ethical guidelines.
4. AI Applications in Various Sectors
4-1. AI as a Service (AIaaS)
AI as a Service (AIaaS) is revolutionizing how businesses integrate AI solutions. Virtual assistants like Alexa utilize advanced machine-learning models to provide personalized assistance, helping users manage tasks like scheduling, reminders, and smart home device control. AI agents like Devin automate and execute operational workflows, such as coding based on user specifications, ensuring efficiency without manual intervention. Providers offer machine learning frameworks to streamline model development and deployment. For example, Google Cloud AI facilitates the summarization of large documents, ML image processing pipelines, and chat app creation via APIs. These comprehensive AI services enable users to build and deploy advanced AI applications quickly and efficiently.
4-2. AI Tools in Business and Education
Artificial Intelligence (AI) is being harnessed extensively in business and education to drive efficiency and innovation. According to Gartner, around 37% of companies use AI to operate their businesses, a number that has surged 270% from 2015 to 2019. AI tools span various industries, with some notable implementations including: - **Proptech + Marketing**: Digital marketing software for real estate professionals leverages AI for predictive analytics, significantly reducing the labor-intensive processes of lead conversion and sale completion. - **HR Tech**: Tools like the Wonderlic Test use AI to efficiently and objectively evaluate job candidates' competencies, ensuring equitable hiring processes. - **Edtech**: Platforms like Mainstay incorporate AI to engage students with behavioral intelligence, providing reminders and support opportunities for improved educational outcomes. - **Software Development**: Solutions from companies like GitLab and Notion include AI enhancements for code suggestions and research, streamlining development processes. Oracle AI offers tools for customer feedback analysis and predictive modeling, while TensorFlow supports building machine learning projects. These tools exemplify the transformative impact of AI on traditional business operations, boosting productivity and decision-making across various sectors.
4-3. Large Language Models (LLMs)
Large Language Models (LLMs) are advanced AI models designed to understand and generate human language, significantly impacting the field of Natural Language Processing (NLP). These models, such as Generative Pre-trained Transformer (GPT) series, BERT, and others, comprehend semantics, capture nuances in language, and provide contextually relevant responses through deep learning techniques. - **GPT-3, GPT-3.5, GPT-4**: These models from OpenAI are known for language translation, summarization, chatbot development, and even creative writing assistance. GPT-4, released in 2023, stands out with over 170 trillion parameters, capable of processing both text and images, and powering applications like Microsoft Bing's AI chatbot. - **BERT**: This model excels in understanding language context and nuances, aiding in sentiment analysis, named entity recognition, and question answering systems. - **PaLM 2 and Bard**: Developed by Google, these models are tailored for content creation, multilingual communication, and integrating with various content management systems. LLMs are integral to applications such as machine translation, sentiment analysis, conversational AI chatbots, and generating diverse content types. Their sophisticated capabilities enhance search accuracy, improve customer interactions, and support decision-making processes in businesses, demonstrating the extensive influence of AI-driven language models.
5. Societal and Environmental Impacts of AI
5-1. Energy Consumption by Data Centers
Data centers running artificial intelligence (AI) programs have significantly increased electricity demand. Google reported a 48% increase in carbon emissions since 2019, attributing the rise to AI. The International Energy Agency (IEA) indicated that data centers utilize about 40% of their electricity for computing and another 40% for cooling. The exponential growth in AI services, driven by large language models, has contributed to the surge in energy usage. As per the IEA, in 2022, data centers, cryptocurrencies, and AI collectively consumed 460 TWh of electricity, which equates to nearly 2% of global electricity demand. Researcher Alex De Vries estimated that if NVIDIA's AI-specialized servers operate at full capacity, their annual electricity consumption could range between 85.4–134.0 TWh, comparable to the energy consumption of Argentina or Sweden. Despite efforts by Amazon, Google, and Microsoft to reduce their carbon footprints by buying renewable energy, the construction of new data centers and the increased energy usage of existing ones have led to higher greenhouse gas emissions. Google and Microsoft’s greenhouse gas emissions increased by 48% and 30% respectively in recent years, attributed primarily to the expansion of AI.
5-2. Impact on Mental Health through Social Media
The impact of social media on mental health, particularly among teenagers, has raised significant concerns. During a parliamentary inquiry in Canberra, Antigone Davis, Meta’s vice-president and global head of safety, stated that teenage mental health issues are complex and multifactorial and dismissed the notion that social media alone is harmful. Nevertheless, Prime Minister Anthony Albanese criticized Meta, calling the company arrogant for not acknowledging the potential harm social media can cause, such as online bullying, social exclusion, and grooming. Albanese emphasized that social media companies must recognize their social responsibility to protect their users, especially young people. He noted that parents are concerned about the negative impact of social media on their children's mental health and that companies like Meta should do more to address these issues responsibly.
5-3. EU Regulations on AI and Harmful Content
The European Union (EU) has intensified its regulatory actions against major tech firms concerning the management of harmful content. Elon Musk's social media platform, X (formerly Twitter), is facing a formal warning from the EU for not adequately addressing dangerous content, potentially leading to fines of up to 6% of the company's revenue. This is part of the EU’s broader enforcement of the Digital Services Act (DSA) and Digital Markets Act (DMA), which mandate social media platforms to curb misinformation, hate speech, terrorist propaganda, and harmful advertisements. The investigation into X began following the October 2023 Hamas attacks on Israel. The EU has also scrutinized companies like Meta, AliExpress, and TikTok for their content moderation practices. The DMA, effective since March 2024, aims to prevent abusive practices by large tech companies, ensuring fair competition in the market.
6. User Experience (UX) and AI
6-1. AI in Enhancing UX
Artificial Intelligence (AI) plays a crucial role in revolutionizing the field of User Experience (UX). By leveraging AI tools, UX professionals can accomplish tasks more efficiently, generating multiple design variations and conducting rapid iteration processes. This capability leads to quicker outputs compared to traditional methods, which rely solely on human input. Notably, AI aids in content curation and creation, enabling creatives to produce high-quality articles, infographics, images, and other media swiftly. However, the AI-generated content often requires human oversight for fine-tuning and ensuring emotional and aesthetic appropriateness. Moreover, AI’s potential to directly integrate into UX allows applications and websites to offer 'predictive' experiences. In this context, AI fosters the creation of more personalized and engaging user interactions by analyzing behavioral patterns and user preferences, making the digital interface more intuitive and user-friendly.
6-2. NLP for Human-Machine Communication
Natural Language Processing (NLP) is a significant branch of AI responsible for enabling computers to understand and process human language. NLP aims to bridge the communication gap between humans and machines, making interactions feel more natural and accessible. This technology powers virtual assistants like Siri and Alexa, enhances chatbot functionalities, and enables applications to conduct sentiment analysis, voice recognition, and language translation. By understanding users' emotional states, chatbots equipped with NLP can provide more tailored and empathetic responses, potentially improving user satisfaction. Recent advancements, such as the implementation of AI-edited product review summaries on platforms like Amazon, illustrate NLP's growing impact on digital experiences. Furthermore, NLP's role in translation services and automated documentation in healthcare highlights its potential to aid professionals by reducing time spent on administrative tasks and allowing greater focus on core responsibilities.
6-3. Personalized Experiences and Chatbots
AI's ability to analyze user data makes it indispensable for creating personalized digital experiences. Through detailed analysis of user behavior and preferences, AI can tailor interactions to individual needs, enhancing user engagement. This is evident in personalized recommendations on streaming platforms like Netflix and social media apps like TikTok. Personalized UX feeds emotional satisfaction, leading to richer user experiences and sustained engagement. Additionally, AI-powered chatbots and virtual assistants have become ubiquitous in the digital space, providing immediate assistance and streamlining user interactions. These tools free up human customer service resources by handling routine inquiries. Though the effectiveness of chatbots can vary, advancements in AI and Artificial General Intelligence (AGI) promise more intuitive and emotionally intelligent interactions in the future. For instance, AI chatbots like Pi offer more natural, conversational engagement, resembling human communication patterns and making digital interactions smoother and more enjoyable.
7. AI in the Context of Global Development
7-1. OpenAI's Global Tour and AGI
Sam Altman, CEO of OpenAI, emphasized the continuous and sustained growth of AI technologies during his global tour. He noted that while the rate of AI advancement is accelerating, achieving a fully developed artificial general intelligence (AGI) remains a major challenge due to the high standards people have for AI. Altman pointed out that even if AI reached significant milestones such as curing diseases or addressing climate change, human curiosity and the drive for further advancement would persist. His tour, which included visits to Tokyo, London, Tel Aviv, New Delhi, and Seoul, aimed to foster relationships with political and business leaders to discuss AI advancements and address societal concerns. The tour came in the wake of the popularity and scrutiny of OpenAI's generative AI solution, ChatGPT.
7-2. National Surveys on AI in Education
A comprehensive overview of the definition and functioning of AI, as discussed in the University of Illinois Chicago's online Master of Engineering with a focus on AI and Machine Learning program, reveals the growing importance of AI in the educational sector. AI's ability to learn from experience, adapt to new inputs, and execute tasks traditionally requiring human intelligence is transforming various aspects of education from content generation to personalized learning experiences. The program highlights AI's applications in machine learning, neural networks, natural language processing, and game playing, which are pivotal for developing educational tools that can cater to diverse learning needs. The burgeoning field of AI is also fostering career opportunities, evidenced by the anticipated growth in AI-related roles such as machine learning engineers and data scientists.
7-3. AI Development in South Korea
South Korea is making significant strides in the development of AI technologies, positioning itself as a key player in the global AI ecosystem. Companies like Naver, Kakao, and LG are at the forefront, developing their own generative AI solutions utilizing large language models. The emphasis is on creating sovereign AI systems that operate independently and securely to protect sensitive data. This approach is particularly important given concerns about data leakage to entities like OpenAI and other AI users. Jung-woo Ha of Naver stressed the necessity of building an advanced AI ecosystem to safeguard data sovereignty for both local businesses and the nation. South Korea's robust AI capabilities are paving the way for innovative solutions that can compete globally while addressing national security concerns.
8. Conclusion
The findings underscore the substantial influence of AI technologies like NVIDIA hardware and ASIC Sohu on modern industry and society, driving advancements in business operations, medical applications, and educational tools. However, this growth accompanies significant challenges such as the energy consumption surge from data centers and the ethical dilemmas posed by AI applications like ChatGPT. While AIaaS and LLMs offer productivity enhancements, the ongoing debates and legal issues, exemplified by the New York Times vs. OpenAI case, necessitate stringent regulatory oversight. The European Union, through policies like the EU Digital Services Act, plays a pivotal role in enforcing accountability among tech giants. To balance AI's benefits and risks, continuous adaptation of legal frameworks and ethical guidelines is imperative. Future prospects include stricter regulations and the evolution of AI to provide more personalized and efficient user experiences. Applying these findings practically involves leveraging advanced AI tools to address global challenges, such as environmental sustainability and mental health considerations, while ensuring the ethical deployment of AI technologies.