• BuzzRobot
  • Posts
  • Exploring the ethical and societal impact of AI assistants with a Google DeepMind researcher. Virtual talk

Exploring the ethical and societal impact of AI assistants with a Google DeepMind researcher. Virtual talk

Plus: Alternative architectures to Transformers

Hello, fellow human! Here we come – the movie 'Her' is no longer science fiction.

We are getting closer to a time when AI assistants could be our creative partners, research assistants, tutors, or life planners. We may even fall in love with them… All this is happening in the very near future.

That's why the development and deployment of AI assistants demand careful evaluation and foresight. The new reality will require us to address the following questions:

  • What might a world populated by advanced AI assistants look like?

  • How will humans relate to new, more capable forms of AI that have human-like traits and with which they can easily speak?

  • How might these dynamics play out at a societal level – in a world with millions of AI assistants interacting with one another on their users' behalf?

On July 11th, our speaker, Iason Gabriel, a staff research scientist from the Ethics Research Team at Google DeepMind, will explore a range of ethical and societal questions that arise in the context of assistants: value alignment and safety, anthropomorphism and human relationships with AI, and questions about collective action, equity, and overall societal impact.

Just a quick reminder that this Thursday, July 11th, we are having Aleksandar Botev from Google DeepMind, who’ll be diving deep into the technical aspects of Griffin, a new architecture and potential competitor to Transformers.

Speaking of Transformer alternatives, check out a video recording of the lecture we hosted a couple of weeks ago. Dan Fu from Stanford University spoke about emerging efficient architectures competitive with Transformers.

What I found especially interesting from the lecture is that many companies' tech stacks are so hooked up serving Transformer-based models, that by Dan’s estimate, it would take them at least 6 months and the best ML systems engineers to transition to serving non-Transformer models.

Reply

or to participate.