Discovery Loop

The technological method is 1 of the top devices humanity has ever devised, yet execution entails repetitive experimental loops that are difficult to standard pinch today's manual efforts: you propose an experiment, instrumentality and tally it, analyse the results, past iterate to refine your approach.

Historically, technological advancement has relied connected these sequential quality iterations. In galore domains, this process remains incredibly slow and labor-intensive.

At Discovery Loop, we are building systems to automate these full experimental loops. By utilizing frontier AI models and large-scale computational infrastructure, our systems will beryllium capable to quickly propose, run, and study from evaluations.

This attack allows for the parallel execution of thousands of experiments, drastically compressing loop clip and driving up the amount and value of technological and engineering output.

Start pinch Machine Learning

We will initially attraction connected automating the process of instrumentality learning investigation and engineering.

Act arsenic Our Own First Customer

We will usage these automated ML capabilities to quickly optimize our ain exertion stack earlier expanding to different domains.

Grand Challenges

We judge our attack will beryllium capable to lick immoderate learning loop pinch measurable outcomes wrong the domains of subject and engineering. Ultimately, we are building systems tin of taking connected National Academy of Engineering (NAE) Grand Challenges—such arsenic engineering amended medicines, advancing wellness informatics, making star power economical, providing entree to cleanable water, securing cyberspace, and engineering the devices of technological discovery.

Between us, we person pioneered monolithic standard computing and led the creation of captious infrastructure, products, and foundational AI advances that the world relies on, including aggregate generations of Google Search, Google Ads, Google News, Google Translate, Google File System, MapReduce, BigTable, Spanner, TensorFlow, Pathways, TPUs, AlphaChip, AlphaStar, AlphaCode, AlphaFold, Gemini, exemplary distillation, mixture-of-experts exemplary architectures, word2vec, sequence-to-sequence models, concatenation of thought reasoning, neural architecture search, and aggregate generations of Large Language Models (LLMs) among others.

Our comparative advantage isn't conscionable our method ability; it is the unprecedented standard of the systems we person antecedently built. We person existent full-stack extent that spans chips, hardware infrastructure, package infrastructure, ML models, and products.