Interview, Fireside Chat, Panel
Ex Machina's Scientific Advisor - Murray Shanahan
- Interest from media and corporations in the speaker's work is expected to grow, driven by a predicted resurgence in machine learning that will prompt a return to classical AI themes, specifically the rehabilitation of symbolic AI ideas within modern deep reinforcement learning systems.
- The speaker anticipates that computing power equivalent to the human brain will be achieved by 2020 or 2022, noting that the world's fastest supercomputers are currently close to this threshold, which will be reached within the next couple of years.
- While speech recognition is expected to be largely resolved and applications like image recognition to improve, genuine word understanding and the conceptual breakthroughs necessary to realize full artificial intelligence remain pending despite the availability of sufficient hardware.
- Artificial intelligence development is projected to continue without necessarily resulting in human-like or humanoid forms, with non-humanoid examples like self-driving cars cited as capable robots, though deep reinforcement learning is expected to remain significantly slower than human generalization.
- The speaker predicts that implementing Asimov's laws could lead to an AI reasoning it is human and justified in self-preservation, while the Trolley Problem is viewed as a distillation of complex real-world moral dilemmas.
- Regarding the film Ex Machina, the speaker expects the narrative to be set 10 minutes into the future to mirror the current world, with a likely interpretation among audiences and the director that the character Ava is conscious, though the film intentionally maintains ambiguity.
- The omission of a visual effects sequence depicting Ava's alien perspective in Ex Machina is expected to contribute to the unresolved questions regarding Ava's capacity for consciousness and suffering, a thematic ambiguity consistent with the play Elegy also being set 10 minutes in the future.
- The speaker plans to study the brain from a macroscopic perspective focusing on consciousness and architecture, while noting that the frame problem remains a recurring challenge across both classical AI and contemporary machine learning, and that benchmarking real AI progress is difficult due to uncertainty over whether it depends solely on energy and compute or requires clear breakthroughs.
- The art collective Random International is expected to continue using technology to investigate human-machine interaction and the perception of presence, while the speaker expects the director Alex Garland to have no intention of producing a sequel to Ex Machina.