Mustafa Suleyman, cofounder of DeepMind and founder of the new AI startup Inflection, believes that generative AI like chatbots are just an interim phase. The real revolution will come with interactive AI – bots and agents that can take actions and accomplish goals by coordinating with other software systems and even people.
This next wave of interactive AI represents a profound shift in technology, giving it a kind of virtual agency rather than just responding statically to user commands. With the right high-level goals from users, interactive AI could independently coordinate across different systems to get things done.
For example, you could tell an interactive AI assistant to plan a vacation for you within a certain budget. It could then autonomously search for flights, hotels, and activities, negotiate rates and make reservations based on your preferences, collaborate with a human travel agent if needed, and ultimately book the entire trip for you.
The key breakthroughs making this possible are large language models like GPT-3 that can understand natural commands and giant sets of training data that teach AIs common sense reasoning and interaction strategies. Startups like Anthropic are also engineering safety features directly into models to ensure interactive AIs behave responsibly.
Interactive AI promises to automate complex coordinated tasks across both digital and physical domains. However, handing agency over to autonomous software also introduces major risks around AI safety and aligning these systems to human values. Companies like Inflection will need to prioritize ethical AI development.
Overall though, Suleyman argues interactive AI will reflect our best human traits like empathy and thoughtfulness. This technology can remove human limitations and inconsistencies, optimizing systems to be more fair and effective. But robust regulations and ethical standards will be critical to steer interactive AI towards benefits over harms.
The third wave of interactive AI is arriving. Pioneering startups are determined to animate technology for the good of humanity – if we responsibly co-evolve with it.
You can read the full interview transcript with Suleyman here