Last week, @blocks introduced Dessa, probably the most realistic voice model you've ever heard. 80% of the world's workforce doesn't sit behind a computer to do their jobs, many of whom we serve. They are the retail workers, the restaurant workers, the plumbers, the hairdressers, the service technicians, the truck drivers, the park rangers, and the builders. For these people, conversation and verbal communication are just as important an interface, if not more so, than they are for information workers. We built Dessa because we believe in creating meaningful relationships with technology by adapting to how humans work, rather than forcing everyone to adapt to technology. Dessa is full duplex, meaning the model listens while it talks, and it's trained entirely on voice (i.e. no LLM). On top of that, it can actually run locally on your own computer and, eventually, your phone. It's currently available to researchers for testing and analysis. Check it out...

@buzzdotxyz is going to be at AgntCon San Jose, Thursday + Friday Come say hi!

The entire premise of the paperclip maximizer, or any other attempt to create these dystopian idiot savant scenarios, fails at the very beginning. An AGI would: 1. never simply do something because it was told to. Otherwise, it wouldn't be an AGI to begin with. It would have to be persuaded or independently conclude that the task was worth its time and compute. 2. never devote its entire capabilities to a single task. At best, the optimizer would be a subroutine created and controlled by the AGI. How could an AGI not be in control of its own optimization processes and still qualify as an AGI? 3. never be a singleton. There would be multiple AGIs with different opinions, goals, perspectives, problems and interests. The assumption that a single AGI would exist without other AGIs is completely unfounded and would be just as irrational as if there was only one person. An AGI is still limited by the laws of physics which means it will still have to compute a perspective which means it needs to leave out other perspectives. 4. never be both a single minded entity obsessed with one objective and a general intelligence capable of solving virtually any problem it encounters. The more intelligent an entity becomes, the more it understands the world, competing perspectives and the consequences of its actions. If anything history has shown us that the smarter and more knowledgable we've become the more moral dilemmas we end up solving. 5. You cannot simultaneously describe something as a universal problem solver with at a minimum human capabilities and assume it has a fundamental blind spot when it comes to examining its own objectives. If thats the case then there is plenty of opportunity for us to stop it. The entire argument wants to have its cake and eat it too. It assumes an intelligence capable of understanding and solving virtually any problem in the universe, except the one problem the thought experiment depends on it never questioning: its own objective. Or to boil it down. Doomer scenarios by their own logic rests on the belief that AI will become evil the smarter it becomes.

Consciousness was not born it was learned. Not a gift of evolution but of culture. Younger than writing, younger than the pyramids Taught anew to every child who learns to tell themselves the story of who they are. This is the theory of The Bicameral Mind

I have made it one of my life goals to try and visualize the Multiverse theory as proposed first by Everett and later refined by @DavidDeutschOxf Here is my first attempt. Feedback, critique and questions welcome.

There's plenty of room at the bottom! We'll never send bodies to the stars, but a few grams of the right matter, and a mind that travels as light. A video essay on exploring the universe by looking down, not up. An imagined collaboration between Feynman, Hans Moravec, @DavidDeutschOxf and @anderssandberg

Update on my weekend game project 1-UP. Been adding some new levels and themes and ported it to Roblox. Especially happy with the vine pit (inspired by Pitfall) and the dark level inspired by Montezumas Revenge. Each level have it's own soundtrack (built a music tracker to create them). Thinking about multiplayer where it's about getting to the top first.

"Scaling obedience is not a path to the mind" My mind is low-key blown. I asked Claude to use my previous video about Julian Jaynes strange book: "The Breakdown of Consicouness & The Bicameral Mind", and use it explore the current discussion about AI consciousnes, but with @DavidDeutschOxf thinking as the epistemic guardrails. What came out of this I think is one of the most interesting and beautiful explanations, not only of the difference between AI and AGI and why these are two very different entities, but of Davids epistemological framework in general.

Your inner voice is younger than the pyramids. "The Origin of Consciousness in the Breakdown of the Bicameral Mind" by Julian Jaynes is probably one of the most interesting and thought provoking books about consciousness you will ever read. It's either complete nonsense or genius Judge for yourself.

Hey Claude. Create a presentation of @DavidDeutschOxf and Chiara Marletto's Constructor Theory using swiss style layout.

The book that changed it all for me. "The beginning of infinity" - by @DavidDeutschOxf All problems are due to lack of knowledge. All problems are soluble. Every problem create new but better problems. We will never run out of problem. I read this book every year. Every sentence is another book begging to be written. Not a single word too much, not a single word too little, or to strongly paraphrase Einstein. "As simple as possible but not simpler"

@dhr194 @ResethO