• AI KATANA
  • Posts
  • OpenAI Touts New AI Safety Research. Critics Say It’s a Good Step, but Not Enough

OpenAI Touts New AI Safety Research. Critics Say It’s a Good Step, but Not Enough

Also: More than 40% of Japanese companies have no plan to make use of AI

Author

AI KATANA
July 18, 2024

Check out our new AI jobs section below, we will be featuring AI jobs in Asia. If you would like to feature a role, please contact us.

Morning!

In today’s newsletter, we explore OpenAI’s new AI safety technique aimed at enhancing transparency, while critics call for increased external oversight. DBS Bank is rolling out a generative AI-powered virtual assistant to streamline customer service operations, showing promising results in early trials. Despite global advancements, a significant portion of Japanese companies remain hesitant to adopt AI due to various concerns, as revealed by a recent survey. The Japan Newspaper Publishers and Editors Association demands consent and accuracy from generative AI providers to protect news content integrity. Additionally, the UK’s new government plans to regulate powerful AI models to balance innovation with safety. In other news, GoTo introduces an AI-based voice assistant in Indonesia, Japan leverages AI for foreign worker training, and significant venture capital investments highlight the growing AI sector in medical research and enterprise solutions.

Sliced:

  • ⚠️ OpenAI Touts New AI Safety Research. Critics Say It’s a Good Step, but Not Enough

  • 🏦 DBS equips its customer service officers with gen AI-powered virtual assistant

  • 🇯🇵More than 40% of Japanese companies have no plan to make use of AI

  • 📰 Japan news media association demands consent and accuracy from generative AI

  • 🇬🇧 Britain's new government aims to regulate most powerful AI models

OpenAI has unveiled a new AI safety technique aimed at making AI systems more transparent. This method involves two AI models engaging in a dialogue, encouraging the more powerful model to explain its reasoning, which the second model verifies. This approach, tested on simple math problems, is part of OpenAI’s broader strategy to ensure the safety of advanced AI systems. Despite these efforts, critics argue that more external oversight and governance are necessary. Recent organizational changes at OpenAI, including the disbandment of a team focused on long-term AI risks and departures of key figures, have fueled concerns about the company's commitment to balancing rapid AI development with safety considerations.

DBS Bank is set to enhance its customer service operations with a generative AI-powered virtual assistant named CSO Assistant, developed entirely by the bank’s AI engineers. This assistant integrates a large language model with voice telephony and speech recognition, tailored to local languages. The CSO Assistant can transcribe customer queries in real time, perform live searches on the bank’s knowledge base, and assist with post-call documentation. Initial pilots since October 2023 have shown nearly 100% accuracy in transcription and solutions, and it is expected to reduce call handling times by up to 20% once fully deployed. Following its rollout to 500 customer service officers in Singapore by the end of 2024, DBS plans to expand the implementation to other markets, including Taiwan and Hong Kong, within the next year. This initiative is part of over 20 generative AI use cases DBS is pursuing, aimed at driving efficiency and quality improvements across its operations.

🇯🇵More than 40% of Japanese companies have no plan to make use of AI

A recent survey conducted by Nikkei Research for Reuters revealed that over 40% of Japanese companies have no plans to adopt AI, highlighting a significant divide in the technological advancement within the country’s corporate sector. The survey, which included responses from roughly 250 out of 506 companies approached, showed that while 24% of firms have already integrated AI into their operations and 35% are planning to do so, the majority remain hesitant. Primary objectives for those adopting AI include addressing worker shortages, reducing labor costs, and accelerating research and development. However, concerns such as potential job losses, lack of technological expertise, high capital expenditure, and doubts about reliability are major barriers. Additionally, the survey noted that 15% of companies experienced cyberattacks in the past year, prompting a mix of outsourced and in-house cybersecurity measures. This reluctance to embrace AI, coupled with cybersecurity challenges, reflects the broader issues faced by Japanese companies in modernizing their technological capabilities.

The Japan Newspaper Publishers and Editors Association has issued a statement demanding that providers of generative AI services, like Google and Microsoft, obtain proper permits from media organizations to use their news content and ensure the accuracy of generated information. The association expressed concerns that AI-assisted online search services, known as retrieval-augmented generation (RAG), often infringe on copyrights and may damage the trustworthiness of news organizations. It highlighted issues such as identical reproduction of original stories, inaccuracies, and lack of corrections by AI. The association warned that unregulated use of content could harm democracy and national culture, urging the government to review intellectual property laws to prevent "freeriding" on news content. Additionally, it raised concerns about "zero-click searches," where users get sufficient information from RAG services without visiting the original news sites, disadvantaging traditional media.

The newly elected Labour government in Britain has announced plans to explore the regulation of the most powerful AI models, aiming to establish legislation that would place requirements on developers of these advanced technologies. This initiative follows the previous administration's efforts to position Britain as a leader in AI safety, including hosting a global summit on AI and launching the world’s first AI Safety Institute. The new government, led by Prime Minister Keir Starmer, is taking a cautious approach to AI regulation, seeking to maintain a competitive edge while addressing the rapid integration of AI into daily life and public services. AI experts are urging the government to expedite the legislative process to address the risks and opportunities presented by these technologies.

🧑🏽‍💻 AI Jobs

🛠️ AI tools updates

GoTo, Indonesia’s largest digital ecosystem, has launched a new AI strategy featuring "Dira by GoTo AI," an AI-based voice assistant designed to enhance user experience, safety, and AI capabilities within its ecosystem. Dira, the first AI voice assistant in Indonesia's financial technology industry, operates in Bahasa Indonesia and is integrated into the GoPay app. This assistant aims to streamline user transactions and improve accessibility, enabling users to perform various functions such as bill payments and money transfers with simple voice commands. GoTo plans to expand Dira’s availability to the Gojek app and further develop its AI features to continue improving user experience and operational efficiency.

Japan is utilizing AI to create educational videos for foreign workers, addressing the growing demand for labor due to an aging population and labor shortages. The software, called Teachme AI, developed by Tokyo-based company Studist, can produce instructional videos in 20 different languages, significantly reducing the time needed for video editing. The tool helps companies efficiently train their foreign workforce, with the ability to generate 30-minute videos in just 15 minutes, complete with subtitles and chapters. This initiative aligns with recent legislative changes aimed at encouraging foreign workers to stay longer in Japan, reflecting a broader strategy to integrate and support the foreign labor force.

💵 Venture Capital updates

CytoReason, an Israeli AI medical startup, has raised $80 million in a funding round supported by major investors including Nvidia and Pfizer. Established in 2016, CytoReason employs AI to create computational disease models that simulate the human body, aiding in drug discovery and development. This technology integrates diverse data sources to provide insights into disease biology, potentially reducing trial phases and development costs while increasing drug approval rates. With this new capital, the Tel Aviv-based company plans to expand its disease models, enhance its molecular and clinical data, and establish a US hub in Cambridge, Massachusetts. CytoReason collaborates with six of the world's top ten pharmaceutical companies, including Pfizer, Sanofi, Roche, and GSK, leveraging its platform to advance research in immunology, inflammation, and other therapeutic areas.

Amity, a tech startup founded by Korawad Chearavanont, the grandson of Thai billionaire Dhanin Chearavanont, has successfully raised $60 million in its latest funding round. The company, which specializes in generative AI chatbots that assist enterprises with issue resolution and employee queries, plans to use the funds for acquisitions to expand its product offerings and enter international markets. Amity, which has now raised a total of $90 million since its inception in 2012, aims to list its Thai unit publicly in 2025. The company's AI solutions are popular for generating reports, analyses, and recommendations by connecting to customer databases, and it generates a significant portion of its revenue from Thailand, servicing local banks, government agencies, and CP Group companies.

🫡 Meme of the day

⭐️ Generative AI image of the day