![](\"https:\/\/fortune.com\/img-assets\/wp-content\/uploads\/2026\/02\/DSC01071-1-e1770045331321.jpg?w=2048\")
<\/p>\n<\/p>\n
Sara Hooker, <\/strong>an AI researcher and advocate for cheaper AI systems that use less computing power, is hanging her own shingle.<\/p>\n The former vice president of research at AI company Cohere and a veteran of Google DeepMind, has raised $50 million in seed funding for her new startup, Adaption Labs<\/p>\n Hooker and cofounder Sudip Roy, who was previously director of inference computing at Cohere, are trying to create AI systems that use less computing power and cost less to run than most of the current leading AI models. They are also targeting models that use a variety of techniques to be more \u201cadaptive\u201d than most existing models to the individual tasks they are being asked to tackle. (Hence the name of the startup.)<\/p>\n The funding round is being led by Emergence Capital Partners, with participation from Mozilla Ventures, venture capital firm Fifty Years, Threshold Ventures, Alpha Intelligence Capital, e14 Fund, and Neo. Adaption Labs, which is based in San Francisco, declined to provide any information about its valuation following the fundraise.<\/p>\n Hooker told Fortune <\/em>she wants to create models that could learn continuously without the expensive retraining or fine-tuning and without the extensive prompt and context engineering that most enterprises currently use to adapt AI models to their specific use cases.<\/p>\n Creating models that can learn continuously is considered one of the big outstanding challenges in AI.\u00a0\u201cThis is probably the most important problem that I\u2019ve worked on,\u201d Hooker said.\u00a0<\/p>\n Adaption Labs represents a significant bet against the prevailing AI industry wisdom that the best way to create more capable AI models is to make the underlying LLMs bigger and train them on more data. While tech giants pour billions into ever-larger training runs, Hooker argues the approach is seeing diminishing returns. \u201cMost labs won\u2019t quadruple the size of their model each year, mainly because we\u2019re seeing saturation in the architecture,\u201d she said.<\/p>\n Hooker said the AI industry was at a \u201creckoning point\u201d where improvements would no longer come from simply building larger models, but rather by building systems that can more readily and cheaply adapt to the task at hand.<\/p>\n Adaption Labs is not the only \u201cneolab\u201d (so-called because they are a new generation of frontier AI labs following the success that more established companies like OpenAI, Anthropic, and Google DeepMind have had) pursuing new AI architectures aimed at cracking continuous learning. Jerry Tworek, a senior OpenAI researcher, left that company in recent weeks to found his own startup, called Core Automation, and has said he is also interested in using new AI methods to create systems that can learn continually. David Silver, a former Google DeepMind top researcher, left the tech giant last month to launch a startup called Ineffable Intelligence that will focus on using reinforcement learning\u2014where an AI system learns from actions it takes rather than from static data. This could, in some configurations, also lead to AI models that can learn continuously. <\/p>\n Hooker\u2019s startup is organizing its work around three \u201cpillars\u201d she said: adaptive data (in which AI systems generate and manipulate the data they need to answer a problem on the fly, rather than having to be trained from a large static dataset); adaptive intelligence (automatically adjusting how much compute to spend based on problem difficulty); and adaptive interfaces (learning from how users interact with the system).<\/p>\n Since her days at Google, Hooker has established a reputation within AI circles as an opponent of the \u201cscale is all you need\u201d dogma of many of her fellow AI researchers. In a widely-cited 2020 paper called \u201cThe Hardware Lottery,\u201d she argued that ideas in AI often succeed or fail based on whether they happen to fit existing hardware, rather than their inherent merit. More recently, she authored a research paper called \u201cOn the Slow Death of Scaling,\u201d that argued smaller models with better training techniques can outperform much larger ones.<\/p>\n At Cohere, she championed the Aya project, a collaboration with 3,000 computer scientists from 119 countries that brought state-of-the-art AI capabilities to dozens of languages for which leading frontier models did not perform well\u2014and did so using relatively compact models. The work demonstrated that creative approaches to data curation and training could compensate for raw scale.<\/p>\n One of the ideas Adaption Labs is investigating is what is called \u201cgradient-free learning.\u201d All of today\u2019s AI models are extremely large neural networks encompassing billions of digital neurons. Traditional neural network training uses a technique called gradient descent, which works a bit like a blindfolded hiker trying to find the lowest point in a valley by taking baby steps and trying to feel whether they are descending a slope. The model makes small adjustments to billions of internal settings called \u201cweights\u201d\u2014which determine how much a given neuron emphasizes the input from any other neuron it is connected to in its own output\u2014checking after each step whether it got closer to the right answer. This process requires enormous computing power and can take weeks or months. And once the model has been trained, these weights are locked in place.<\/p>\n To hone the model for a particular task, users sometimes rely on fine-tuning. This involves further training the model on a smaller, curated data set\u2014usually still consisting of thousands or tens of thousands of examples\u2014and making further adjustments to the models\u2019 weights. Again, it can be expensive, sometimes running into millions of dollars.<\/p>\n Alternatively, users simply try to give the model highly specific instructions, or prompts, about how it should accomplish the task the user wants the model to undertake. Hooker dismisses this as \u201cprompt acrobatics\u201d and notes that the prompts often stop working and need to be rewritten whenever a new version of the model is released.<\/p>\n She said her goal is \u201cto eliminate prompt engineering.\u201d<\/p>\n Gradient-free learning sidesteps many of the issues with fine-tuning and prompt engineering. Instead of adjusting all of the model\u2019s internal weights through expensive training, Adaption Labs\u2019 approach changes how the model behaves at the moment it responds to a query\u2014what researchers call \u201cinference time.\u201d The model\u2019s core weights remain untouched, but the system can still adapt its behavior based on the task at hand.<\/p>\n \u201cHow do you update a model without touching the weights?\u201d Hooker said. \u201cThere\u2019s really interesting innovation in the architecture space, and it\u2019s leveraging compute in a much more efficient way.\u201d<\/p>\n She mentioned several different methods for doing this. One is \u201con-the-fly merging,\u201d in which a system selects from what is essentially a repertoire of adapters\u2014often small models that are separately trained on small datasets. These adapters then\u00a0 shape the large, primary model\u2019s response. The model decides which adapter to use depending on what question the user asks.<\/p>\n \u00a0Another method is \u201cdynamic decoding.\u201d Decoding refers to how a model selects its output from a range of probable answers. Dynamic decoding changes the probabilities based on the task at hand, without changing the model\u2019s underlying weights.<\/p>\n \u201cWe\u2019re moving away from it just being a model,\u201d Hooker said. \u201cThis is part of the profound notion\u2014it\u2019s based on the interaction, and a model should change [in] real time based on what the task is.\u201d<\/p>\n Hooker argues that shifting to these methods radically changes AI\u2019s economics. \u201cThe most costly compute is pre-training compute, largely because it is a massive amount of compute, a massive amount of time. With inference compute, you get way more bang for [each unit of computing power],\u201d she said.<\/p>\n Roy, Adaption\u2019s CTO, brings deep expertise in making AI systems run efficiently. \u201cMy co-founder makes GPUs go extremely fast, which is important for us because of the real-time component,\u201d Hooker said.<\/p>\n Hooker said Adaption will use the funding from its seed round to hire more AI researchers and engineers and also to hire designers to work on different user interfaces for AI beyond just the standard \u201cchat bar\u201d that most AI models use.\u00a0<\/p>\n<\/div>\n #Adaption #Labs #secures #million #seed #build #models #change #fly<\/p>\n","protected":false},"excerpt":{"rendered":" Sara Hooker, an AI researcher … <\/p>\n","protected":false},"author":1,"featured_media":18705,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[2],"tags":[11718,82,544,5341,4170,781,4076,913,814,2267,9064,11717,4172,1571],"_links":{"self":[{"href":"https:\/\/microvibenews.com\/index.php?rest_route=\/wp\/v2\/posts\/18704"}],"collection":[{"href":"https:\/\/microvibenews.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/microvibenews.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/microvibenews.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/microvibenews.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=18704"}],"version-history":[{"count":0,"href":"https:\/\/microvibenews.com\/index.php?rest_route=\/wp\/v2\/posts\/18704\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/microvibenews.com\/index.php?rest_route=\/wp\/v2\/media\/18705"}],"wp:attachment":[{"href":"https:\/\/microvibenews.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=18704"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/microvibenews.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=18704"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/microvibenews.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=18704"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}\n