AI engineers are in high demand – but what is the job really like?

Key Concepts

• Synthetic IDs: Fake identities used to understand and prevent fraud.
• Chaos Engineering: Intentionally injecting failure into systems to learn and improve resilience.
• Generative AI (Gen AI): AI that can generate new content, such as text, images, or code.
• Large Language Models (LLMs): AI models pre-trained on vast amounts of data to understand and generate human language.
• SEA-LION: A large language model developed in Singapore catering to Southeast Asian languages.
• Prompt Engineering: The skill of crafting effective prompts to maximize the power of large language models.
• Parsing: Breaking down user input into smaller pieces to determine its meaning.

I. Fighting Scams and Fraud with AI

• Rajat Maheshwari's Role: Rajat Maheshwari from Mastercard's Cyber & Intelligence Solutions team develops tools to manage customer risk and prevent scams.
• Synthetic IDs: Rajat's team creates synthetic IDs, which are as good as genuine IDs, including face masks and fingerprints, to understand how fraudsters operate. These masks were created by a Japanese artist who did the masks for James Cameron’s movie Avatar.
• Deepfakes: The evolution from masks to deepfakes highlights the increasing sophistication of fraudulent activities.
• Global Anti-Scam Alliance Statistics: More than $1 trillion are lost to scams every year, affecting 2 billion victims globally.
• AI in Transaction Monitoring: Mastercard uses AI models to detect patterns indicative of scams, such as a large number of people sending money to the same recipient in a short period. The AI provides a risk assessment to the issuing institution, leading to the decline of suspicious transactions.

II. Chaos Engineering and Generative AI in Cloud Security

• Chaos Kitty Project: Joel Garcia's team at Amazon Web Services (AWS) uses a project called Chaos Kitty to simulate real-world security conditions and improve incident response processes.
• Intentional Failure Injection: Chaos engineering involves intentionally injecting failure to learn about system vulnerabilities.
• Visual Representation of Cloud Security: The Chaos Kitty setup uses colorful lights and bricks to represent AWS cloud components, with red indicating security configuration issues and green indicating compliance.
• Gen AI Assistant: A Gen AI assistant is integrated to help users understand and apply company security policies.
• AI-Powered Remediation: The AI can be prompted to fix issues, such as "Alexa, fix Chaos Kitty," allowing security teams to focus on more complex improvements.

III. Developing Large Language Models for Southeast Asian Languages

• SEA-LION Model: Leong Weiqi and his team in Singapore are developing SEA-LION, a large language model catering to Southeast Asian languages.
• Multicultural Context: The model aims to address the limitations of existing LLMs in understanding and responding to multicultural contexts and idioms specific to Southeast Asia.
• Informal Language Example: The video demonstrates SEA-LION's ability to understand and respond appropriately to informal Indonesian, including idioms like "his beard is always on fire," which other models may interpret literally.
• Open Source Initiative: All the work done on SEA-LION is fully open source and shared with the public.
• Team Diversity: The SEA-LION team includes members from various Southeast Asian countries, creating a "mini-ASEAN" environment.

IV. Generative AI for Personalized Experiences

• StoryGen Project: Vincent Oh at AWS worked with the National Library Board of Singapore on StoryGen, a project that uses generative AI to create personalized books for children and adults based on user inputs.
• Prompt Engineering: The project highlights the importance of prompt engineering, a new skillset that involves crafting effective prompts to leverage the power of LLMs.

V. AI's Impact on Jobs and Efficiency

• Mid-Career Switches: The video emphasizes that AI is an industry where people can make mid-career switches, with examples of team members who previously worked in finance and pharmacy.
• AI's Role in Efficiency: AI is portrayed as a tool that enhances efficiency, allowing people to focus on more important tasks for businesses and customers.
• Parsing Invoices Example: LLMs can parse invoices, even reading handwriting, to extract relevant information, such as the signatory.

VI. Future of AI

• Integration with LLMs: The integration with large language models is seen as a key trend in the future of AI.
• Expansion of Language Support: There is a long-term ambition to expand language support to include more dialects, such as Balinese, Javanese, and Bisayan.
• Global Impact: The efforts in Southeast Asia are expected to trigger similar movements in other regions with underrepresented languages, such as India and Africa.

VII. Notable Quotes

• Rajat Maheshwari: "Sometimes we have to think like bad actors to come up with the solutions which can stop these things."
• Vincent Oh: "This is just the very beginning of the art of the possible and it’s going to be amazing how people will leverage Gen AI to unleash their extended level of creativity."
• Rajat Maheshwari: "If I am able to save someone's life-long savings, then that makes my day."

VIII. Conclusion

The video showcases various applications of AI, from fighting scams and fraud to developing personalized experiences and supporting diverse languages. It highlights the importance of understanding AI's capabilities and limitations, as well as the growing demand for AI specialists and the potential for mid-career switches into the field. The development of LLMs like SEA-LION demonstrates the need for AI solutions that are tailored to specific cultural and linguistic contexts. Ultimately, AI is presented as a tool that can enhance efficiency, improve security, and unlock new levels of creativity.