This day is a seminal moment in the life of this weblog -- this is the closest we've ever come to seeing the word "self-replicate" in the news.
They refer to it as "digital replication", "living individual replicas", and "digital doppelgänger" but the concept has become a tightly arranged crystal of thought-parts. The idea of multiple, simultaneous instances of a single original self is now well-formed.
What's better is that this was published in The Conversation, which is a general audience magazine not a science journal, or a science fiction story for that matter:
Image credit: Try not to do this often but this is not the first time this image has been used as a thumbnail: René Magritte, Not to Be Reproduced (La reproduction interdite, 1937
AI clones made from user data pose uncanny risks
Jun 2023, phys.org
This mirror image of an individual created by artificial intelligence is referred to as an "AI clone." Our study dives into the murky waters of what these AI clones could mean for our self-perception, relationships and society.We presented 20 participants with eight speculative scenarios involving AI clones. The participants were diverse in ages and backgrounds, and reflected on their emotions and the potential impacts on their self-perception and relationships.
What we fear:
- Digital counterparts could exploit and displace their identity
- Threat of identity fragmentation
- Lastly, participants expressed concerns about what we described as "living memories." This relates to the danger posed when a person interacts with a clone of someone they have an existing relationship with. Participants worried that it could lead to a misrepresentation of the individual, or that they would develop an over-attachment to the clone, altering the dynamics of interpersonal relationships.
via University of British Columbia: Patrick Yung Kang Lee et al, Speculating on Risks of AI Clones to Selfhood and Relationships: Doppelganger-phobia, Identity Fragmentation, and Living Memories, Proceedings of the ACM on Human-Computer Interaction (2023). DOI: 10.1145/3579524
Note: As the volume of personal data we generate continues to grow, so too does the fidelity of these AI clones in replicating our behavior.
Word watch: "user-generated data expiration strategies"
Multiple 'selves' of modular agents boost AI learning
Jul 2023, phys.org
The book writes itself:
A study comparing reinforcement learning approaches used in single AI agent and modular multi-AI agent systems -- trained deep reinforcement learning agents in a simple survival game were trained to seek various resources hidden around the field and to maintain sufficient supply levels to prevail.One agent, seen as the "unified brain" or "self," operated in standard fashion, taking a step-by-step approach to evaluate each objective and, through trial-and-error, learning what the best solutions are each step of the way.The modular agent, however, relied on input from sub-agents that had more narrowly defined goals and had their own unique experiences, successes and failures. Once input from the multiple modules were assessed in a single "brain," the agent made choices on how to proceed.The singular agent achieved the game's goals after 30,000 training steps. The modular agent learned faster, making significant progress after only 5,000 learning steps.
via Princeton Neuroscience Institute: Zack Dulberg et al, Having multiple selves helps learning agents explore and adapt in complex changing worlds, Proceedings of the National Academy of Sciences (2023). DOI: 10.1073/pnas.2221180120
Post Script:
Playing Tetris by committee (the Octopad, an eight player game controller)
May 2019, BBC
May 2019, phys.org
July 2019, University of Washington News
Bonus: the monolithic approach struggled with "the curse of dimensionality" -- the exponentially spiraling growth of options as the complexity of the environment was increased.
No comments:
Post a Comment