Good morning.
OpenAI’s ChatGPT had its one-year anniversary on Nov. 30. And generative AI remains a hot topic. But a lot of companies are still just testing the waters.
Cnvrg.io, an Intel company, released its annual 2023 ML Insider global survey and the findings indicate the majority of organizations are still in the research or testing phase of incorporating generative AI in production. Just 10% have launched generative AI solutions, according to the AI and large language model (LLM) platform provider. Meanwhile, 25% of respondents said they’re building pilot projects for selected use cases. (In this report, a “solution” refers to an application that uses generative AI for a purpose.)
The findings are based on a global survey of 430 AI and ML practitioners, such as data scientists, engineering/DevOps, and software developers across industries including IT, financial services, banking, defense, insurance, education, automotive, and telecommunications. Half of the respondents came from companies with 600 employees or more.
The research also found that 75% of respondents reported that their organization has yet to deploy generative AI models to production in the past year, compared to 25% that have done so. (A “model” refers to generative AI separate from a solution—it’s essentially at an earlier stage of maturity and not yet a part of a solution that has a specific purpose.) The survey shows that U.S.-based respondents (40%) are significantly more likely than those outside the U.S. (22%) to deploy generative AI models.
“While still in early development, generative AI has been one of the most talked-about technologies of 2023,” Markus Flierl, corporate VP and general manager of Intel Cloud Services, said in a statement. “The survey suggests organizations may be hesitant to adopt generative AI due to the barriers they face when implementing LLMs.”
Intel is among the major companies that joined the “AI Alliance” co-launched by IBM and Meta, according to a Tuesday announcement. The alliance challenges the perceived dominance of OpenAI, Microsoft, Google, and recently Amazon, Fortune reported.
Although generative AI applications like ChatGPT are recent, LLMs are nothing new. So the companies surveyed, for the most part, are approaching generative AI by building their own LLM solutions and customizing it to their use cases, according to the report. But 46% see infrastructure as the greatest barrier to developing LLMs into products. Also, lack of knowledge, cost, and compliance are also seen as barriers.
But those who have deployed generative AI indicated better customer experience (58%), improved efficiency (53%), and cost savings (47%), according to the report.
Experts warn that you shouldn't just incorporate generative AI without really assessing how it can add value to your company. Alex Singla, a senior partner at McKinsey and co-leader of QuantumBlack, the AI arm of the firm, said his team first looks across an institution and asks the question: “What are the core business problems you’re trying to solve?” And from there they assess whether or not generative AI can play a role in delivering against that business opportunity, he recently told me.
Singla said companies also need to assess their existing technology, asking the questions: Do I have good infrastructure in place? Do I have my cloud platforms and cloud relationships already established? Do I have the right data architecture in place?
Regarding the survey findings, Flierl anticipates increased adoption of generative AI in the coming year as organizations gain greater access to cost-effective infrastructure and services.
Have a good weekend.
Sheryl Estrada
sheryl.estrada@fortune.com
