Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.
It’s happening faster than we thought, and the implications deserve greater attention. https://t.co/OVVPJO7VQx
Introducing Magenta RealTime 2 (MRT2): the live music model you can play as an instrument.
MRT2 offers MIDI and prompt controls, and runs natively on a MacBook with <200ms latency.
Open weights. Open source inference engine. Suite of apps and plugins.
Hear what it can do and try it out for yourself below 🧵
Internet users reacting to AI images. A brief history.
- OMG, puppy slugs!! WTAF
- Uncanny Valley, LOLS
- It's sh*t
- It'll never be as good as human artists.
- What about the human artists!?
- I will destroy your family!!!
- Check out what I just made!
I'm a traditional artist and film industry professional who now also builds AI tools for the film industry, so I think I have some unique perspective to offer here.
Every single person I consult with points out that AI is useful to them in the domains that support them, but will certainly never be able to operate in their specific role.
What this means is that Directors will use AI for research, for concepts, boards etc, but AI will never direct.
Storyboard artists will see the benefits of using AI to create their films, because 'storyboarding is the real craft', and directing is just putting it on camera.
Researchers will use AI to collate the data into a digestable form, but AI will never research.
Writers will use AI to do research, but AI will never write.
Each the master of their domain, supported by any technology that can aid them in THEIR vision. It's the goal, and the agency that is retained, not the interplay between roles.
Everything else is a supporting act that can "obviously" be automated or streamlined.
Behind the scenes, of course, to support this you must build general intelligence that can handle - to some reasonable degree - the entire pipeline end to end. It's only from this place that each individual can insert themselves back in to conduct the orchestra in their own unique way.
It's my personal opinion that AI is a collaborative medium first and foremost, and that you want to bring your collaborators into your latent space to maximise the human surface area. Like so many people holding up a canopy.
But it's still fascinating to me that everyone's perception of the "irreplaceable human spirit" generally begins and ends at their own passions.
So yes, of course AI "may be useful" for VFX because VFX isn't this director's domain.
And he will gladly accept the title of "defender of humanity" while acknowledging that SOME roles (certainly not his) may be usefully augmented by AI.
Rest assured that everyone feels the same. The only honest anti-AI position here is that you will use AI where it elevates you, and reject it where it usurps you.
An arguably more objective position is that AI allows people the option to choose to collaborate with any number of people from any set of diverse skills, knowing that you can still ship an outcome from that specific configuration of talent and budget.
And that of course you will honour your creative collaborators and use the technology only where it helps bridge the gap between collaborators, and between the human experience of the creator and of the audience.
As the great physicist Max Planck put it, scientists must have a vivid intuitive imagination, for new ideas are not generated by deduction, but by an artistically creative imagination.
@a16z Or numbers. World models use numbers. Language and mathematics describe the world beautifully to us, we developed those languages. Even so, I love this approach.
Three Predictions:
1. Some form of AI, probably neurosymbolic in nature, will come that is far more economical and data- and energy-efficient than LLMs, and it will make an absolute fortune.
2. LLMs, on the other hand, will never be all that profitable (aside from the chip companies selling shovels in the gold rush).
3. Today’s gigantic bets are premature, and most won’t pay off.
❌ AI music products that promise democratization, but don’t make you smarter at music nor coding, and everything sounds the same
✅ Learning to code your own insane Ableton setup unique to your brain
AI music is dead.
Long live music hacking.
Build your own sh!t
I’m pleased to share that we’ve launched the TRACE Institute.
A new research effort bringing together physicists, mathematicians, biologists, neuroscientists, and philosophers on a key question in science: what are observers, and how are they related to spacetime and reality?
How animals sense Earth’s magnetic field is one of biology’s enduring mysteries.
Researchers in Science have now identified superparamagnetic macrophages in the livers of rock pigeons to be crucial for magnetic sensing.
The finding uncovers an unexpected role for immune cells in sensory perception and may fundamentally change our understanding of animal navigation.
Learn more: https://t.co/afMu07Io5l
Using our models, ESP and @unileon found that most crow communication happens through soft, close-range murmurs.
At @raincoast, ESP's tools are accelerating killer whale research by processing field note recordings in hours instead of months.
📄 : https://t.co/uL8IbCZmYk