Interview, Fireside Chat
What is Wolfram Language? (Stephen Wolfram) | AI Podcast Clips
- The Wolfram Language is expected to evolve within a limited number of years into a symbolic discourse language capable of encoding actionable human conversations, including desires and basic common sense relevant to typical legal contracts.
- Computational contracts are predicted to become the future of autonomously executing agreements across sectors like health care and central banking, replacing traditional legalese with computable conditions that enable computers to decide execution and resolve arguments.
- Future AI development will rely on the interplay between symbolic knowledge and statistical methods, such as image identification and models like BERT, to close the gap between unstructured text and computable knowledge without starting from scratch.
- Ethical principles will be encoded into "AI ethics modules" allowing users to select specific ideological systems for content ranking, moving away from a single centralized module toward a market-based system with multiple modules operating under different value systems.
- The speaker plans to tackle the challenge of "infinite tails" in long-term projects by working on a fundamental theory of physics alongside the development of the Wolfram Language to maintain a coherent vision over decades.
- While high-end R&D users currently drive adoption of the Wolfram Language, a slower "trickle down" to other professionals is anticipated due to social network dynamics and the absorption of ideas relying on personalities, despite plans to expand reach through free engines and university licenses.
- The Wolfram Knowledge Base is expected to grow continuously as data flows in, aiming to eventually represent the finite scope of the world's knowledge, with the ultimate goal of enabling fully automatic answers to general knowledge questions.
- The speaker acknowledges that while progress is visible for a few years, the complete realization of these projects, including the digitization of all knowledge and the solving of complex scenarios like self-driving car decisions, will likely take decades.
- Risks associated with a single, centralized AI ethics module are identified as detrimental to the future of the species, prompting a strategy to "bite off the whole problem" of natural language understanding rather than solving it in pieces, while ensuring the integrity of the knowledge base through leadership rather than open distribution models.