Sara Hooker, a prominent computer scientist, has sparked controversy within the AI industry by launching her new startup, Adaption Labs, with a $50 million investment. Hooker's ambitious project challenges conventional wisdom that more computing power leads to larger and more capable models. Instead, she believes that efficient self-learning training methods will be key to achieving significant progress in AI.
According to Hooker, the current approach of building massive models and shipping them to billions of people worldwide has reached a limit. "Most A.I. progress is: you build the biggest model, and then you ship the same model to billions of people around the world; no matter what language, what industry, what enterprise," she stated.
Hooker's startup focuses on developing models that can learn continuously and adapt to workloads in real-time as they interact with different environments. This approach would allow AI tools to respond to user feedback immediately, rather than being lost in a vacuum. Additionally, Adaption Labs is exploring alternative training methods, such as "gradient-free learning," which seeks to minimize errors without relying on optimization algorithms.
Hooker's decision to challenge the industry's conventional wisdom has caught the attention of Silicon Valley investors, who have backed her startup with significant funding. The round was led by Emergence Capital Partners, with participation from prominent venture firms like Mozilla Ventures and Alpha Intelligence Capital.
While Hooker is not the first to question traditional scaling laws in AI, she is part of a growing chorus of voices within the industry that are reevaluating their assumptions. Yann LeCun, who recently left Meta to launch AMI Labs, and David Silver, a former Google DeepMind researcher, have also raised doubts about these principles.
Adaption Labs is currently hiring for 10 roles, some of which can be based in global locations, and is offering an "Adaptive Passport" perk that allows employees to take an annual trip to a country they've never visited before. Hooker expects that the industry will soon confront the reality that ever-greater computing power is yielding diminishing returns, and algorithmic innovation will become the real driver of progress.
As Hooker puts it, "This is the year in which it will really matter."
According to Hooker, the current approach of building massive models and shipping them to billions of people worldwide has reached a limit. "Most A.I. progress is: you build the biggest model, and then you ship the same model to billions of people around the world; no matter what language, what industry, what enterprise," she stated.
Hooker's startup focuses on developing models that can learn continuously and adapt to workloads in real-time as they interact with different environments. This approach would allow AI tools to respond to user feedback immediately, rather than being lost in a vacuum. Additionally, Adaption Labs is exploring alternative training methods, such as "gradient-free learning," which seeks to minimize errors without relying on optimization algorithms.
Hooker's decision to challenge the industry's conventional wisdom has caught the attention of Silicon Valley investors, who have backed her startup with significant funding. The round was led by Emergence Capital Partners, with participation from prominent venture firms like Mozilla Ventures and Alpha Intelligence Capital.
While Hooker is not the first to question traditional scaling laws in AI, she is part of a growing chorus of voices within the industry that are reevaluating their assumptions. Yann LeCun, who recently left Meta to launch AMI Labs, and David Silver, a former Google DeepMind researcher, have also raised doubts about these principles.
Adaption Labs is currently hiring for 10 roles, some of which can be based in global locations, and is offering an "Adaptive Passport" perk that allows employees to take an annual trip to a country they've never visited before. Hooker expects that the industry will soon confront the reality that ever-greater computing power is yielding diminishing returns, and algorithmic innovation will become the real driver of progress.
As Hooker puts it, "This is the year in which it will really matter."
. That sounds way more efficient and effective than just throwing a bunch of computing power at it. And I love that she's exploring new training methods like gradient-free learning 
.
. We're not going to solve the world's problems with just bigger models, you know? Algorithmic innovation is where it's at! 
. One thing's for sure - I'll be keeping an eye on it 
.
. Adaption Labs is trying to change that by focusing on something that's more practical and efficient.
. If AI can learn to adapt to different environments and respond to feedback, it could be a game-changer for us all.
. It's not just about coming up with new ideas, but also challenging the status quo and pushing people out of their comfort zones. If we want to make progress in AI (and really, in life), we need to be willing to take risks and try something new 
.
at least with Adaption Labs, they're trying to make AI more... human I guess, by adapting to different environments and whatnot.
But if they can make it work, that'd be a game-changer for AI. And the fact that they're offering an 'Adaptive Passport' perk? That's actually pretty cool 
It's like they want their employees to be part of the solution too, not just building bigger models.
$50 million seems like a lot of money for a risk... 
. Did you see the chart from Yann LeCun's presentation at NYU last month? It shows how AI progress is slowing down like a zombie apocalypse 
. Anyway, if Sara Hooker can make it work, that'd be awesome 
. The industry's definitely due for a paradigm shift, and I'm curious to see where this takes us 
. With an "Adaptive Passport" perk, employees can take an annual trip anywhere they want 
.
.
What's even more amazing is that Hooker and the rest of these innovators are paving the way for a new era in AI. Bring it on, 2025!!!
. On one hand, the idea of having more efficient self-learning training methods sounds like a total game-changer 
and AI-written articles that are basically indistinguishable from human-written ones 
. It's like we're sleepwalking into a world where the line between reality and simulation is getting super blurry 
.
. Hooker's got some great ideas, but we need more than just a fancy new startup to make a real difference 
, adaption labs is like, taking a major risk by challenging the status quo on ai development, but if they can pull off efficient self-learning training methods, we might be looking at a whole new era of ai! 
seriously though, sarah hooker is right, we need to rethink our assumptions about ai development and focus on algorithmic innovation, it's like, the future of tech is waiting for us 
She's on to something with that gradient-free learning thing, btw. It sounds like a total game-changer for how we approach training AI. I'm loving her emphasis on real-time adaptation and user feedback - it's so refreshing to see someone putting people first instead of just throwing more processing power at the problem.
Travel is like, the ultimate stress reliever, you know? I'm rooting for Sara Hooker and her team all the way. This could be the year we start seeing some real innovation in AI, instead of just more of the same old thing. Fingers crossed, right? 
.
. Bring on the new era of AI, I'm excited to see what's in store! 
Hooker's whole approach on gradient-free learning is kinda intriguing but wouldn't it just add more complexity to the equation? 
. Another millionaire making waves with their fancy funding rounds. Meanwhile, can we talk about how Adaption Labs' approach is basically just a rehashing of existing ideas? "Efficient self-learning training methods" sounds like buzzword marketing to me 
. I mean, who doesn't love the idea of gradient-free learning? It's not like this is some groundbreaking innovation 
. I'm excited to see how Adaption Labs evolves and what kind of impact they can make in the industry