Empathetic AI

Early last month, Y Combinator (YC) published a video titled ‘𝙒𝙝𝙮 𝙑𝙚𝙧𝙩𝙞𝙘𝙖𝙡 𝙇𝙇𝙈 𝘼𝙜𝙚𝙣𝙩𝙨 𝘼𝙧𝙚 𝙩𝙝𝙚 𝙉𝙚𝙬 $1 𝘽𝙞𝙡𝙡𝙞𝙤𝙣 𝙎𝙖𝙖𝙎 𝙊𝙥𝙥𝙤𝙧𝙩𝙪𝙣𝙞𝙩𝙞𝙚𝙨’, underscoring why these agents go far beyond being ‘general AI wrappers.’ Sequoia Capital also stated in its recent 'Generative AI Act 01 report' that 'for any given domain, it’s hard to encode 𝐝𝐨𝐦𝐚𝐢𝐧 𝐚𝐧𝐝 𝐚𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧-𝐬𝐩𝐞𝐜𝐢𝐟𝐢𝐜 𝐜𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞𝐬, that’s where last-mile app providers might have the upper hand in solving the messy real-world problems'. I was grateful for the opportunity to speak at Google’s Sydney office on GenAI Lab's 𝟭𝘀𝘁 𝗯𝗶𝗿𝘁𝗵𝗱𝗮𝘆 last week, where I shared what we do at Empathetic AI, how we do it, and why we build a 𝘃𝗲𝗿𝘁𝗶𝗰𝗮𝗹 𝗱𝗼𝗺𝗮𝗶𝗻 𝗔𝗜 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻 𝗶𝗻 𝘁𝗮𝘅. Here are my reflections on building a domain specific GenAI product: 🍋 For any specific domain with deep knowledge, the real challenge is turning domain expertise into AI. The value for an AI startup is how quickly and accurately you can translate that expertise into AI code, essentially, building a cognitive architecture that embodies the domain’s nuances. 🍎 In certain industries, the stakes are much higher comparing to others, because information, data, and knowledge all come with their own layers. This goes way beyond using AI to summarize an article or draft an email. If we use a tool to tackle the 'know-how', it needs to be smart enough to provide solid reasoning, which means a ton of data re-ranking and hierarchy work on the backend. 🍐 We're moving from 'interact with process' to 'interact with result', where there is a big question mark in between, therefore, 'explainable AI' has to play a key role, people need to see the 'why' behind it. 🍇 Designing an AI tool for a specific field means you have to really get the ins and outs. Never underestimate the complexity here. Also, you need a solid evaluation mechanism in place. If there isn’t one, then you’ll have to build it yourself. 🍓 Your users are the ultimate judges of your product, they’re the only ones who can confirm product-market fit. Never build a solution and assume it will solve the problem, it should always be the other way around. Thanks to Alex Zinoviev for the invite and to the incredible team Kyra Blake, Rebecca Dengate, Kristin de la Fuente for organizing such a fantastic event. In just one year, the GenAI Lab has grown into a thriving community with 1,600+ members. How incredible! 🚀 It was also amazing to see Australia’s first open-source LLM, Kangaroo LLM presented by Sarina Hinduja. 🦘 A big thanks to Google and Vectara for hosting and sponsoring the event. More to come in the GenAI ecosystem! 🌟

EVENT RECAP 🎉 'Building GenAI Products' 🎉 Last night we celebrated the first birthday of GenAI Lab with a meetup at Google sharing the stories of three different founders who are reshaping the future of Generative AI. Key Takeaways: 🎤 Angela Shi - Founder, Empathetic AI 💫 Founder Perspective: As a non-technical founder, Angeli applied her finance expertise to address Australia’s complex tax system with Empathetic AI's Luna, the country’s first AI tax co-pilot. 💫 Advanced Analysis: Luna is trained on the ATO’s legal database dating back to 1936, offering scenario analysis, real-time updates on private and public rulings and legislation. 💫 Ethical Transparency: Partnering with CSIRO's Data61, Empathetic AI prioritizes transparency by safeguarding sensitive information and using a “human-in-the-loop” approach for accurate, expert-supported insights. 💫 User-Centric AI Design: Luna was built with NLP and clustering tools, allowing users to customise outputs for their unique tax issues, aligning AI assistance with individual needs. 🎤 Jacky Koh - Co-Founder - Relevance AI 💢 Domain-Expert AI Agents: Jacky highlighted the potential for domain professionals to train AI agents with their expertise, allowing these agents to be cloned, licensed, or even work within multi-modal teams, broadening accessibility to specialised knowledge. 💢 Multi-Agent System for Scalability: He discussed the scalability of a multi-agent system, where AI agents duplicate, communicate with shared memory, and work collaboratively, achieving 10x productivity and an economy of scale not possible with human-only teams. 💢 AI Workforce Models: Jacky described three models for AI systems: Assembly Line (rigid tasks), Omni-Direct (creative tasks), and Leader (a hybrid), highlighting how sharing memory across agents creates a horizontal, high-efficiency workforce. 🎤 James Gabb - Co-Founder Sauce AI 💬 Bias Awareness in LLMs: James noted that large language models (LLMs) can carry bias depending on how they’re trained, cautioning developers to consider alternatives like clusters and custom tools when appropriate, as LLMs aren’t always the answer. 💬 Building Trust in AI Products: He addressed common issues—high expectations, reduced control, and overestimation of AI’s capabilities—that can erode trust in AI, explaining how Sauce AI aims to counter these through a bottoms-up approach. 💬 Real-Time Insight and Personalisation: Sauce AI provides a 'living digital footprint' for the organisation by capturing data across inputs, clustering trends, and creating real-time, prioritised insights that help ICs address product & customer needs accurately and effectively. We got to hear from Sarina Hinduja (CPO) at Kangaroo LLM who is developing Australia's first LLM which is as Aussie as vegemite on toast. #GenAILabs #Community #AI

A while back, I saw a comment on LinkedIn asking "Can Australia’s tech sector really make an impact globally?" Well, if Australia isn’t widely recognized as a tech innovation hub, that’s even more reason for the entrepreneurs, industries, and academics pushing boundaries here to have a platform to shine. Great night at InnovationAus Gala Dinner yesterday, thanks to National AI Centre and Department of Industry, Science and Resources for the opportunity. Stela SOLAR Rita Arrigo Eloise Leaver Learning from other founders, industry leaders, and investors was definitely one of the best parts of the night. Belinda White of Commonwealth Bank, Kristen Migliorini of KomplyAi, Charles Guan of Rich Data Co, Anthony Chung and David Anstee of Conja, Sha-mayne Chan of Friyay.ai, Monica Ly Matthew Philip Charlotte Yinlun Pan of EdgeRed, Howjer Gu of Planoverse, Carolina Ferreira Neha Panwar of Octopusbot, Can (Alan) C. of Boson Ventures, Chelsea C. Eva Zhuang of Boman Group Our global impact depends on the vision and drive of these incredible people. What we can achieve depends on how we support and empower them. We should be celebrating those out front, leading the way in innovation and showing what’s possible.

We certainly can’t say we are the best on the market, but each time we receive emails like this, it warms our hearts. Seeing AI help SMEs makes us proud that we’re adding some real value. 👏✨

Reflections from the Generative AI Leaders lunch by Bondi Innovation. I had the privilege of joining insightful discussions on the opportunities and challenges AI brings to the workplace. Here are some key takeaways: AI in the Workplace: • Unlocking AI’s potential requires investment in people, data, and systems – but also navigating complex regulatory and privacy risks. • Generative AI demands more than traditional tech rollouts. Managing workforce disruption and driving intentional, human-centric change is essential. • Upskilling and reskilling employees will be critical to working alongside AI, addressing concerns about job displacement. • The future lies in reimagining possibilities with AI, not just automating processes – bridging technical and business roles will be key. Governance, Risk, and Regulation: • Australia is playing catch-up on data privacy regulations, with GDPR-style laws expected by 2025 – businesses need to prepare for compliance challenges. • Critical skills in governance, cyber, and risk management are currently underdeveloped, posing risks in this fast-moving landscape. • Striking the right balance between innovation and regulation will be crucial to avoid stifling progress while ensuring trust and safety. • Digital literacy and AI awareness at the board level are vital to guiding organizations through these transformations. Was great to meet fellow AI leaders and exciting times ahead as AI reshapes the way we work! It’s clear organizations that proactively adapt and embrace these changes will be the ones that thrive in the future of work. A big thanks to Ross Dawson and speakers Angela Shi and William Ashford. #AI #GenerativeAI #FutureOfWork #Leadership #Governance #Innovation

Join us at the 5th GenAI Leaders Lunch on Monday 28th at AWS office, where we’ll share the fun (and chaos) of building a GenAI startup! Thanks for inviting us Ross Dawson 🌟

Excited to share that Empathetic AI will be presenting at 2nd Symposium on Diversity and Inclusion in AI by CSIRO's Data61 on November 14-15, 2024!🌟

Magic of how Luna being a tax domain AI model, outforms ChatGPT as a general model for tax related questions, with deeper insights, reference checks, data desensitisation and reasoning. 🤗 https://lnkd.in/dicqcp-p

Weekend at SXSW Sydney Tech and Innovation Expo 🤗 The best part? It's always about live-testing what you've built. In the era of AI, being a doer than a thinker, showcasing what you have without too much jargon, is fascinating - because we do not want to use one complexity to replace another, we simplify it. Loved chatting with everyone who came by our booth! Huge shoutout to Kylie Parker and Adam Harris - both domain experts in tax and finance - who LOVE what we’re doing. We believe in the power of GenAI in finance and tax, and we can’t wait for more domain experts and business leaders to join us on this exciting journey. I’m also a big fan of Leonardo.Ai, check out the awesome background picture we generated for Empathetic AI using their platform! So cool! It was lovely catching up with old friends Wilhelmina and Jonathon. From finance and tax to creativity, AI’s power is unstoppable. Big thanks to the National AI Centre for sponsoring the booth, and shoutout to Dragonfly Thinking, we’ve been hearing great things about you at the booth!

I’m always excited for the opportunity to share how AI can work its magic in tax and finance. Thanks SXSW Sydney for offering the incredible platform to do so! 🌟 Here are some reflections I'd like to humbly share from my talk today: 🍎 Building Effective AI Solutions: Successful AI applications rely on a few important principles: ✅ Sufficient and right data ✅ Well-designed architecture ✅ Human-in-the-loop approach ✅ Forward-thinking mindset 🍐 General AI vs. Domain-Specific AI: While general AI models are great for broader tasks, fields like tax and finance need more specialized AI that can offer deeper insights and accuracy tailored to specific needs. 🍇 The Importance of Human Expertise: AI’s performance, particularly in specialized fields, requires ongoing evaluation. Human expertise plays a key role in ensuring AI delivers results that are accurate and applicable to real-world scenarios. 🍋 AI’s Growing Role in Finance: AI’s scalability extends beyond data retrieval, summarizing, analyzing, and highlighting content. Its future strength lies in financial optimization, making it much more than just a productivity tool. What do you think AI can do for the future of finance? 🤗 #AIinFinance #TaxTech #FinTechInnovation #AIOptimization #FutureofFinance #AIEmpowerment #FinanceTransformation #SXSWSydney #GenAI