According to the Presidents Council of Advisors on Science and Technology, U.S. companies are expected to spend more than $100 billion annually on AI R&D by 2025. According to a Georgetown University study, a majority of today's spending is made by six tech companies, including Apple, Google, Amazon and IBM. What if you are a startup that relies on AI as its core product?Are early-stage companies able to support a research-based workflow. Scaleups and startups often place more emphasis on product development than research. Companies want to create products that are valuable to investors, customers and other stakeholders. There is a way to do them both.This advice will help you decide if youre ready for the challenge of starting an AI research laboratory before you invest in staffing it.Collect the right research teamIf your organization is committed to innovative AI research, then the first step should be to hire one or more researchers. Unbabel did this by quickly hiring Ph.D.s to get started with research on a product that was not yet developed. While some researchers will create a product from scratch, others will use your data to identify a model that suits your needs.Google's X division may be able to invest in moonshots but most startups cannot.You will need to hire researchers engineers or machine-learning operations professionals. The research phase is just one part of AI production. Your research will be released into production by researchers. They will monitor the results of your model and make improvements if necessary. They will often use automation to automate monitoring and deployment, as opposed to manually doing everything.This is not something that a researcher scientist would be able to do because they are most comfortable working with data sets and models in training. Engineers and researchers will have to collaborate in a continuous feedback loop in order to improve and retrain models based upon actual inference performance.Select the problems that you are most interested in solvingCSET research shows that around 85% of AI labs across North America and Europe conduct basic AI research. Less than 15% of these AI labs focus on developing. The rest of the world is a different story: Most labs in India and Israel focus on developing.
