An AI agent is only as good as its data
Generative AI tools today face a fundamental problem: outdated and biased data.
Hashtag AI empowers communities with the tools to challenge and enhance AI answers through contributing new information. Hashtag AI isn't just search integration or retrieval augmented generation tools for connecting external data sources, it's an evolving knowledge base connecting communities that grows stronger with every interaction, constantly aligning contributions to keep AI accurate and relevant.
Just as the simple-yet-powerful hashtag came to dominate social media as a tool to organise information and coordinate action, Hashtag AI provides the tools for new ways of organising and coordinating in the era of generative AI. Join the waitlistConnected Knowledge Communities
The first community we're supporting through Hashtag AI is CollectiveBrain.org, which connects early stage businesses with the data they need to perform market analysis. Unlike other providers that charge thousands of dollars for market analysis reports, CollectiveBrain.org aims to build a community that pools data and insights to provide small businesses a level playing field to compete with larger players.
If your organisation is building GenAI tools to empower communities, we'd love to work with you. Please reach out to Anj on LinkedIn or email anj@simmons.ai to see if we can adapt the platform to your needs.
Research
The platform draws on years of research experience from Dr. Anj Simmons, including a $306,220 competitive grant from the Australian Office of National Intelligence (2023 research presentation). In March 2024, Anj left academia to commit full-time to Hashtag AI, bringing her expertise and vision to create tools that redefine how we connect, reason, and build knowledge.
Follow Anj's research on Google Scholar or her technical blog at blog.anj.ai.Development Log
Prototype 4 Released
Prototype 4 explored integration of Hashtag AI with a conversational chat interface. It was designed to have a similar feel to using ChatGPT, with the key differentiator being that it automatically connects people making compatible requests/offers rather than just recalling information.
Test Prototype 4Lessons learned: We received a lot of positive feedback from people we shared it with regarding how easy the interface makes it to share information. However, without an established market or community on the platform, people didn't have a reason to return regularly.
What's next: We'll identify high value use cases and build communities around those.
Prototype 3 Released
Prototype 3 explored integration of Hashtag AI with social media. Anj started a subreddit for people to make requests and offers, and developed FindSimilarBot, an AI-powered bot that attempts to automatically match people requesting something with people offering it (and vice versa).
Test Prototype 3Lessons learned: Even without advertising the subreddit, posts showed up in search results and were quickly found by users selling logo design and tutoring services in response to sample requests we posted. However, many of these users were flagged by Reddit as spam accounts.
What's next: We'll focus on establishing a market that connects genuine users rather than just those looking to sell services.
Prototype 2 Released
Prototype 2 introduces the idea that users should be able to use Hashtag AI to both find and share knowledge.
Test Prototype 2Lessons learned: Knowledge graphs are very powerful when combined with GenAI to search/update data; however, they are a confusing way to present information to end users.
What's next: We'll provide more familiar ways to interact with information, while continuing experiments with knowledge graphs behind the scenes.
Prototype 1 Released
Initial prototype showing the idea of using GenAI to formalise the user's question as a mathematical problem and then reasoning over a knowledge base to arrive at a conclusion.
Test Prototype 1Lessons Learned: Our initial assumption—based on news coverage of AI hallucination issues—was that people would prefer deterministic answers that can be trusted over convenience. However, when we interviewed people who were experimenting with applying GenAI in their business, they claimed that recent chatbots like ChatGPT and Copilot were already accurate enough, and preferred to rely on cloud platform providers like Microsoft for their AI needs.
What's next: Instead of focussing on fixing the limitations of existing AI tools or competing with large providers, we'll focus on uncovering new opportunities.
Hashtag AI Founded
Anj Simmons leaves academia and dedicates her efforts full-time to Hashtag AI.