Did you miss a Data Summit session? Watch On-Demand Here.
Accelerated by comprehensive measures to digitize operations, the enterprise enthusiastically embraces AI. According to IDC’s 2022 AI InfrastructureView survey, 31% of companies say they already have AI in production while most are actively piloting AI technologies. Increasingly, the use of AI is leading to higher profitability, with 27% of businesses responding to a McKinsey survey in December 2021 saying at least 5% of their earnings before interest and taxes (EBIT) is now associated with AI.
But many barriers remain to the successful implementation of AI. Of the companies participating in the AI InfrastructureView poll, only one-third claim to have reached a “mature” state of adoption where their entire organization benefits from an enterprise-wide AI approach. Moreover, while nearly two-thirds of companies in the McKinsey survey say they will continue to increase their investments in AI over the next three years, half admit to experiencing higher-than-expected AI project costs. .
Disconnect data science
Why are AI projects so hard to do? The reasons vary, according to Jeff Boudier, head of product and growth at AI language startup Hugging Face. But typically, companies fail to establish systems that will allow their data science teams – the teams responsible for deploying AI technologies – to properly version and share models, code, and AI datasets, he says. This creates more work for AI project managers, who need to keep track of all the models and datasets created by teams so they don’t have to recreate the wheel for every business request.
“Today, data science is mostly done in‘ single player ’mode, where code lives in notebooks on local machines,” Boudier told VentureBeat via email. “This is how business software was created 15 years ago, before the days of modern versions of control systems and… collaborative workflows changed.”
The emerging discipline of MLOps, which means “machine learning operations” (a term coined by Gartner in 2017), aims to address the diverse and quiet nature of AI development by establishing skills for in collaboration between data scientists. By simplifying AI management processes, the goal of MLOps is to automate the deployment of AI models in an organization’s core software systems.
For example, startups like ZenML allow data scientists to express their workflows as pipelines that, with configuration changes, can accommodate a variety of infrastructure and dev tools. . These can be built into a framework to solve reproduction and versioning problems, which reduces the need for interaction between DevOps teams and data scientists.
Increasing size – and data requirements
But collaboration isn’t the only barrier faced by companies using AI. Others are consequences of machine learning models that continue to grow significantly, according to Boudier. Large models often do not fit into commodity hardware and can be slow and expensive to run. Or they are locked into proprietary APIs and services and doubtfully treated as general problem solvers.
“[Proprietary models hamper] The adoption of AI as… teams cannot review code and properly evaluate or improve models, and continues to create confusion over how to approach AI problems pragmatically, ”Boudier said.“ The deploying large production models to be applied to large amounts of data requires diving the model graph down to hardware, which requires skills not available to most companies. ”
Sean Hughes, ecosystem director at ServiceNow, said companies often rely too much from AI models without doing the work necessary to tailor them for their business. But that can lead to other problems, including the lack of data available to adjust the models in each context in which they will be used. In a survey by Dun & Bradstreet in 2019, companies rated a lack of data equal to a lack of internal expertise as the leading setbacks in further implementation of AI in their organizations.
“Hype and sensationalism are generated when AI scientific research opens up a source of work that achieves new innovative benchmark results that can be misinterpreted by the general public as the same as ‘problem solved.’ But the reality is that the state-of-the-art for a particular AI solution can only achieve 78% accuracy for a well-defined and controlled configuration, ”Hughes told VentureBeat via email . “[A major challenge is] the enterprise user expects that [an off-the-shelf] the model will understand the nuances of the business environment to be useful for decision making … [Without the required data,] even if there is the potential for AI to suggest a direction to correct the next best action, it cannot, because it does not understand the context of the user’s intent in that enterprise. “
On the same page
Feiyu Xu, SVP and global head of AI agreed with SAP, adding that AI projects have the best chance of success when there is alignment between business lines and AI technology teams. This alignment can promote “focused” and “measurable” solutions for delivering AI services, he asserted, and handle ethical issues that may arise during thinking, developing, or deployment.
“The best use cases of AI-powered applications ensure that AI technologies are fully embedded and automated for end users. Also, AI systems work best when experts securely use real business data to train, test, and deploy AI services, ”Xu said.“ Companies need to clearly define the guidelines and guardrails to ensure that ethics issues are carefully considered in developing new AI services from scratch. In addition, it is important to include external and independent experts to regularly review cases and topics in question. ”
On the topic of challenges associated with data AI deployment, Xu points to the emergence of platform-as-a-service solutions designed to help developers and non-developers link resources. data in different backend systems. Torch.AI, for example, connects apps, systems, services, and databases to enable the reconciliation and processing of both unstructured and structured data for AI applications.
“AI plays an important role in empowering companies and industries to be smart businesses,” Xu said. “Most AI users have little experience developing software to design, modify, and improve their own workflows and business applications. Here an intuitive, code -free development environment for functions such as intelligent process automation, workflow management, and robotic process automation. “
The mission of VentureBeat has become a digital town square for technical decision makers to gain knowledge about transformative enterprise technology and transactions. Learn more