Symbolic AI Makes a Comeback With Augmented Intelligence’s Apollo Model

In a world increasingly dominated by neural networks and transformer-based AI models like OpenAI’s o1, a different approach to artificial intelligence is gaining fresh traction: symbolic AI. Long considered outdated due to its reliance on rule-based logic instead of machine-learned patterns, symbolic AI is experiencing a revival thanks to new research showing it can be both scalable and effective in domains where traditional neural models stumble.

This renaissance has sparked a wave of startup activity. Companies like Orby, TekTonic, Symbolica, and Unlikely AI—the latter founded by William Tunstall-Pedoe, co-creator of Amazon Alexa—are exploring how symbolic architectures can automate complex workflows, especially in the enterprise space.

One of the latest entrants, Augmented Intelligence, has just emerged from stealth with an eye-popping $44 million in funding. The investment group includes notable figures like former IBM President Jim Whitehurst, underscoring confidence in this emerging paradigm. At the core of the company’s offering is Apollo, a conversational AI that combines the strengths of symbolic reasoning with neural network adaptability — a hybrid model known as neuro-symbolic AI.

Apollo isn’t just designed to chat. According to founder Elhelo, the goal is for the AI to act. That means performing tasks — such as booking flights or integrating with business tools to pull live data — rather than merely responding with information. This agent-like behavior sets Apollo apart from models like ChatGPT, which typically offer guidance but don’t take direct action unless connected to external APIs or tools.

Where Apollo claims to shine is in its explainability and privacy. Unlike many large language models that require fine-tuning on proprietary data, Apollo operates using pre-set rule-based instructions and accesses only the resources it’s authorized to use. This could be a game-changer for security-conscious companies. Businesses wary of using tools from providers like OpenAI, which have faced scrutiny over data usage, might find comfort in Apollo’s more constrained and transparent approach.

Additionally, Apollo offers detailed logs explaining how decisions were made — an essential feature for auditing AI performance and making improvements. While Elhelo’s claim that the model can “eliminate hallucinations” may be a stretch, the company’s commitment to reducing unpredictable outputs is resonating in the market. So far, it’s been enough to secure a strategic partnership with Google Cloud.

Despite being relatively new and small (around 40 employees), Augmented Intelligence was recently valued at $350 million following a $10 million raise. That’s a bold valuation in a crowded AI space, but it highlights investor belief in the promise of neuro-symbolic AI.

As the AI world looks beyond neural networks for more reliable, secure, and agentic systems, Apollo might just be the symbol of what’s next.

Cited: https://techcrunch.com/2024/09/30/augmented-intelligence-claims-its-symbolic-ai-can-make-chatbots-more-useful/