A Rebel's New Frontier: The Rise of Adaption Labs as AI Industry's Conventional Wisdom Crumbles
In a bold move, computer scientist Sara Hooker has raised $50 million for her new startup, Adaption Labs, which is challenging the conventional wisdom that more computing power and data are the keys to building superior artificial intelligence (AI) models. Hooker, who previously worked at Google and Canadian AI startup Cohere, believes that efficient, self-learning training methods will lead to the best results, rather than relying on massive amounts of data and energy.
Hooker's stance is a radical departure from the industry's traditional approach, which prioritizes building large-scale models that can be shipped to billions of people worldwide. Instead, Adaption Labs is focused on developing models that can learn continuously and adapt to different environments in real-time, incorporating user feedback and exploring alternative training methods like "gradient-free learning."
The startup's ambitious goals are not only seen as a departure from the industry's dominant assumptions but also reflect a growing chorus of voices questioning traditional scaling law principles. Yann LeCun, who recently left Meta to launch AMI Labs, has raised doubts about the industry's reliance on ever-greater computing power, while David Silver, a former Google DeepMind researcher, is working on training self-learning models through experience rather than relying on data.
Adaption Labs' fresh funding will be primarily used to build out its team, with staff members offered an "Adaptive Passport" perk that allows them to take an annual trip to a new country. The company aims to represent itself as a global technology firm from day one, encouraging employees to explore and experience different cultures.
Hooker expects that the industry's reckoning on traditional scaling laws will lead to a year where algorithmic innovation becomes the real driver of progress. "This is the year in which it will really matter," she said, as Adaption Labs poised to disrupt an industry that they have spent their careers advancing.
In a bold move, computer scientist Sara Hooker has raised $50 million for her new startup, Adaption Labs, which is challenging the conventional wisdom that more computing power and data are the keys to building superior artificial intelligence (AI) models. Hooker, who previously worked at Google and Canadian AI startup Cohere, believes that efficient, self-learning training methods will lead to the best results, rather than relying on massive amounts of data and energy.
Hooker's stance is a radical departure from the industry's traditional approach, which prioritizes building large-scale models that can be shipped to billions of people worldwide. Instead, Adaption Labs is focused on developing models that can learn continuously and adapt to different environments in real-time, incorporating user feedback and exploring alternative training methods like "gradient-free learning."
The startup's ambitious goals are not only seen as a departure from the industry's dominant assumptions but also reflect a growing chorus of voices questioning traditional scaling law principles. Yann LeCun, who recently left Meta to launch AMI Labs, has raised doubts about the industry's reliance on ever-greater computing power, while David Silver, a former Google DeepMind researcher, is working on training self-learning models through experience rather than relying on data.
Adaption Labs' fresh funding will be primarily used to build out its team, with staff members offered an "Adaptive Passport" perk that allows them to take an annual trip to a new country. The company aims to represent itself as a global technology firm from day one, encouraging employees to explore and experience different cultures.
Hooker expects that the industry's reckoning on traditional scaling laws will lead to a year where algorithmic innovation becomes the real driver of progress. "This is the year in which it will really matter," she said, as Adaption Labs poised to disrupt an industry that they have spent their careers advancing.
The way they're focusing on efficient self-learning training methods is so refreshing - it's like a breath of fresh air for the AI industry. I mean, who needs more data and energy when you can have models that adapt to different environments in real-time? 
It's not just about the tech itself, but also about creating a better work-life balance with that "Adaptive Passport" perk... sounds like they're giving their team members some serious perks! 
What I love most is that this movement might just change the way we think about AI progress. Like, maybe it's not all about building massive models and then shipping them out - but more about experimenting, learning, and adapting (pun intended)! 
So, you gotta wonder, what's driving Sara Hooker and her team at Adaption Labs? Is it the desire for change or a genuine attempt to break free from the conventional wisdom that's held us back in AI development? I mean, can we really rely on more computing power and data to solve our problems when there are alternative approaches out there?
And what about this "Adaptive Passport" perk - is that just a clever marketing ploy or a genuine attempt to foster global collaboration and understanding?
, but also kinda worried at the same time... like, what if we're all gonna be dependent on these adaptive models for everything? It's like, how much control do we really want to give up? 
. Its about time we moved away from just relying on brute force computing power and energy 
. I love how Adaption Labs is thinkin globally with their "Adaptive Passport" perk - its not just about buildin models, its about creatin a global community that's all about collaboration & cultural exchange 
. Bring it on!
.
who wouldn't wanna travel the world and experience different cultures while still working for a company that's changing the game? 
. But at the same time, it makes total sense that learning continuously and adapting to environments would be key to creating super smart AI models.
. Their thoughts on scaling laws are spot on, in my humble opinion 
That's some next-level employee benefits right there!
... the idea of self-learning, adapting models in real-time is SO smart 
An "Adaptive Passport" that lets them travel to new countries every year? That's like, the ultimate perk 
also, gradient-free learning? sounds like some hippie nonsense to me... 
but at the same time, i've been reading about how massive data isn't always the best way to train AI and that's kinda cool. 