Intel announced today that it will acquire SigOpt, a company based in San Francisco that is building a platform to optimize models of AI applications. The terms of the deal, expected to close this quarter, were not announced, but Intel says it plans to use SigOpt’s technology to accelerate, amplify, and scale AI software tools for developers through its hardware products.
The acquisition comes as Intel reacts to setbacks by doubling the AI silicon market by 2024, which is estimated to be worth more than $25 billion. Intel announced that it would push back its 7-nanometer development process following a delay in its 10-nanometer processor hardware. In February, after spending $2 billion purchasing the chipmaker Habana Laboratories, the company finished designing its Nervana AI model training technology.
The software side of AI and machine learning focuses on SigOpt. CEO Scott Clark created the system that would become SigOpt ‘s product while working at Yelp after completing his Ph.D. at Cornell University, where he used it to refine models and A / B tests. Scott founded SigOpt in 2014 with CTO Patrick Hayes, who had worked as a software engineer at Research in Motion, FourSquare, Bloomberg, and Facebook, a BlackBerry developer. With the concept of taking technology to other sectors, the duo launched the company.
The platform and API of SigOpt optimizes hyperparameters-parameters whose values are used to monitor the training phase of the AI model-to generate high-performance production models. It supports 100 hyperparameters and 100 times parallelism, integrating a number of features such as multimetric optimization, optimization of multitasks, and conditional parameters, and utilizing algorithms of Bayesian and global optimization.
Teams can integrate SigOpt into workflows with a few lines of code, irrespective of the kind of AI framework, model management solution, cloud storage, client library, or model sort. SigOpt helps to parametize models and run optimization loops, monitor metadata from experiments and provide visualizations in a web insights dashboard by tapping a library that can generate up to tens of thousands of concurrent experiments.
SigOpt claims to have collaborated with MIT, the University of California, Berkeley, Carnegie Mellon, and Stanford with’ leading’ algorithmic traders and government agencies, as well as academic users. Investors, including Andreesen Horowitz, SV Angel, U.S. intelligence agency partner In-Q-Tel, and Blumberg Capital, raised a total of $8.7 million in venture capital.
The team from SigOpt, including Clark and Hayes, will join Intel’s post-acquisition machine learning output division. Intel says the software innovations of SigOpt will complement its current portfolio of AI software, which includes OpenVINO, a toolkit launched in 2018 to help developers speed up AI inference on edge and cloud devices.
Intel recently announced that “meaningful income” is generated by its AI solutions, totaling over $3.8 billion in 2019. In May, the company reported that Intel Capital had committed a total of $132 million to 11 startups focused on AI translation services , health care automation, big data analytics, and chipset design to its global investment organization. Intel Capital says that it is on track to invest between $300 million and $500 million in AI-specialized startups, with a specific emphasis on smart edge devices and transformation of networks.
No Comments
Leave a comment Cancel