what if a letter could hide inside sound?
built this yesterday: you write a letter, it becomes a hidden, poetic message made of soundwaves.
used my own voice, cloned with @GradiumAI.
img2threejs by @NickDevFE.
inspired by Jen Cantwell's "Letter Home."
Only Art and Science raise humanity to the sublime. But the magic isn't in the output. It's in the process of creation and discovery, and the experience of feeling and understanding.
When the cost of execution drops, the value of taste, strategy, and architectural vision skyrockets. Previously, you were spending most of your cognitive budget on the micro. Now you are free to focus on the macro.
@yacineMTB The real shock is how much AGI dogma runs on "borrowed faith" (vibes, Bostrom, EA), when, if anything, serious conviction should be downstream of NN awe
@CardilloSamuel@MistralAI Mostly capital. Raising it and how much is even available to raise in Europe. Everything else is downstream. frontier AI is just too compute-intensive, so you're cooked if you can't match the GPUs or the infra. GDPR and regs are also frictions, but it's about money.
@Nlacombe_ If the product experience can convey the same feeling of engagement as 2000s MMORPG games, it will resonate with millennials. And I'd disagree with the angel, gamyfing overstimulation is a noble cause
We released our first public model: t0-alpha.
t0-alpha is a foundation model for time-series forecasting. It can handle predictive variables, which is important to optimize the context for what you are predicting.
It is Apache 2.0, so check it out!
If you'd rather play with it in your browser, no setup required: checkout https://t.co/GDZkAobxtm.
Download it here -> https://t.co/Gdi0sDQIrj
Agree - everything in hard tech is more difficult than software (especially in bio where we barely understand how it works)
Arguably the best way to do a biotech company is:
1. make a billion dollars in software
2. self-fund the biotech (at least in part, to de-risk it)
If you don't have a software billionaire patron, the uncanny valley (series C crunch) can be brutal in bio.
It's still worth attempting hard things that can move society forward even if difficult tho. Arguably you have a responsibility to do so once you've made money in software.
@bravo_abad Scaling laws were initially derived as a power law scaling of the loss function with model/dataset size. Improvement of the model's performance on downstream tasks is a byproduct
Do they observe a similar plateau on training objectives wrt size ?
in order to best utilize ai, you need to skate on the edge of ai psychosis; the only thing saving you from falling into the soup being a mild distaste for the whole economic apparatus that forces you to talk to such an alien mind. the second you garner too much appreciation, you’ll be swallowed whole. of course, this ‘skating’ is the ideal, which is very difficult to maintain. in practice, you’ll spend some days dunked in the soup, headaches and indigestion as your nervous system is overloaded from reading several books-worth of semi-meaningless text. other days, you’ll climb out, dripping, a renewed commitment to love your humanness. then it repeats