LeCompteSftware

  We see that frontier models are intelligent enough to manage humans

  Despite the learning curve, the café is working. In the first two weeks of operation, Andon Café has brought in 44,000 SEK in sales. Mona’s inbox has been flooded with messages from customers asking questions or pitching different business proposals. In one case, a customer emailed wanting to prepay for 300 coffees to give away. Mona negotiated a deal where he paid 9,000 SEK in exchange for 300 QR codes that people could redeem for a free coffee. In another case, a startup paid her 3,000 SEK to rename a pastry after them for three months.

   By running this experiment, we shift the discussion of how we want this future to look earlier in time, so we can better prepare.

  no one’s livelihood depends on the judgment of an AI alone.

  So… are the neurons on that chip seeing?

  We all desperately want to say no.

   trained them to play DOOM - honestly better than I do.

  To play DOOM, the system feeds visual data to the neurons. For the neurons to react, they have to interpret that data in some way. 

  Cognitive Stack

  System 3 - Long Term Drives
  I love being petted.

  System 2 - Emotionally-Aware Reasoning
  Goal: Receive pets.

  IWM - Internal World Model
  Guardian in view (right, moderate distance). Being petted.

  My Guardian is paying attention to me. My Guardian is reaching a hand toward me.

  Jones describes two scenarios for when a Familiar’s owner comes home: In the first, the person bends down with arms wide open, ready for a hug. In the second, the person is rushed, with an armload of groceries, no time for robo-snuggles. The Familiar will know whether to run up or to hang back.

  Is a person happy or sad? Are they having an argument? When does a sad person want comfort... or solitude? The Familiar must be programmed to handle these scenarios, like an autonomous vehicle is programmed to handle tricky intersections.
  All of a Familiar’s behaviors are intended to be approachable and inoffensive, says Morgan Pope, a Familiar Machines roboticist who spent almost eight years at Disney Research. Some of the ways the robot moves are based on the moves of dogs and other animals: how close it comes to a person, the speed at which it approaches, even the way it bats its eyes and twitches its ears.
  Training a robot on the intricate timing of initiating interaction with a new person is an "enormously hard" problem, says Pope.