• Posts
  • Transformers: the Google scientists who pioneered an AI revolution

Transformers: the Google scientists who pioneered an AI revolution

Also: Apple Tests ‘Apple GPT’


Today, we delve into the rich history and ongoing influence of the transformer AI model, originally developed at Google. We also bring to light Apple's foray into AI chatbot technology, an exciting development for the tech giant. Our coverage extends to Google and OpenAI's collaborative venture into digital watermarking to enhance the transparency of AI-generated content. Lastly, we scrutinize the need for a robust AI governance framework in Southeast Asia. Furthermore, we will share updates on AI tools and venture capital happenings that will keep you abreast of the latest trends. Grab your coffee, and let's dive in.


  • 🤖 Transformers: the Google scientists who pioneered an AI revolution

  • 🍎 Apple Tests ‘Apple GPT’

  • 💧 Google And OpenAI Plan Technology To Track AI-Generated Content

  • 🐯 Southeast Asia needs a robust AI governance framework

The transformer is a groundbreaking AI model developed at Google in 2017 by a diverse team of researchers. Originally created to improve machine translation, it works by analyzing entire sentences at once rather than word-by-word. This allows it to understand language and generate text, images, music and more. Though invented at Google, bureaucracy and slow product development led the brilliant scientists behind it to leave and found their own startups, unleashing a wave of innovation in generative AI.

Apple is stepping into the AI chatbot arena with a new tool called "Apple GPT", aiming to compete with prominent players like OpenAI's ChatGPT, Microsoft Bing, and Google Bard. Bloomberg reports indicate the company is in the early stages of developing Apple GPT on a proprietary foundational model, Ajax, that runs on Google Cloud. While this internal-only tool is currently available to some Apple employees with special approval, its public rollout remains uncertain due to ongoing security concerns. This move follows the rapid popularity of OpenAI's ChatGPT and the subsequent entrance of other tech giants into the generative AI field. Despite minimal mention of AI in their recent Worldwide Developer's Conference, Apple CEO Tim Cook has voiced his belief in the potential of generative AI, hinting at future advancements despite current challenges.

Google and OpenAI have announced plans to develop digital watermarking technology to track AI-generated content in response to increased concerns about distinguishing between AI-generated and human-generated content. Digital watermarking, a concept which dates back to the 1990s, involves embedding a small, unobtrusive piece of data in digital content files. This data can be extracted and used to verify the source and authenticity of the content. Despite the maturity of the technology, its implementation faces significant hurdles due to the lack of standardized watermarking algorithms and patent issues. Hence, it is expected that each AI vendor will have to develop its own scheme and consider its patent liabilities. A cooperative move towards standardized payload formats and a common database is urged to streamline the process. Yet, the use of watermarking will be voluntary, and there will be a need for after-creation identification of AI-generated content, foreseeably leading to an arms race between AI detection and AI content creation tools.

The potential for AI to boost Southeast Asia's economy is profound, with forecasts of up to $950 billion gain and a GDP uplift of 10-18% by 2030. However, obstacles such as the need for inclusivity in AI development, cybersecurity, and labor market disruption must be addressed. Governments must shift from a centralised approach, ensuring AI isn't used discriminatorily and robust data protection standards are implemented. The threat of job displacement due to AI's widespread adoption underlines the need for investment in education, training, and research. Initial steps towards AI governance have been taken by various countries in the region, including Singapore's Model AI Governance Framework and national strategies in Indonesia, Thailand, Malaysia, and Vietnam. However, these existing frameworks lack specific guidelines to mitigate the potential risks associated with AI technologies, necessitating a regional adoption of a risk-based approach akin to that set by the EU. A sense of regional cooperation, similar to the EU model, and a clear, specific AI governance framework will not only fortify Southeast Asia's competitiveness but also safeguard citizens' digital rights.

🛠️ AI tools updates

💵 Venture Capital updates

The fortuitous meeting of Databricks CEO, Ali Ghodsi, and MosaicML CEO, Naveen Rao, at the Cerebral Valley conference in March set the stage for one of the year's most significant A.I. deals. A subsequent dinner revealed shared goals to democratize A.I., setting the foundation for a merger. Despite Rao's disinterest in joining large corporations after his past experiences, Ghodsi's direct involvement and personal communication won him over. Their companies, Databricks, a data-analytics giant, and MosaicML, a disruptor enabling cost-effective A.I. model training, agreed on a $1.3 billion acquisition deal in June. By the time the deal was finalized, all 62 MosaicML employees had joined Databricks, contributing to its billion-dollar revenue.

🫡 Meme of the day

⭐️ Generative AI image of the day