Empowering pilots
Additionally within the Flexential survey, 43% of corporations are seeing bandwidth shortages, and 34% are having issues scaling knowledge middle house and energy to satisfy AI workload necessities. Different reported issues embody unreliable connections and extreme latency. Solely 18% of corporations report no points with their AI functions or workloads over the previous 12 months. So it is sensible that 2023 was a yr of AI pilots and proofs of idea, says Bharath Thota, accomplice within the digital and analytics observe at enterprise consultancy, Kearney. And this yr has been the yr when corporations have tried to scale these pilots up.
“That’s the place the problem is available in,” he says. “This isn’t new to AI. But it surely’s amplified as a result of the quantity of information it’s essential entry is considerably bigger.” Not solely does gen AI eat dramatically extra knowledge, nevertheless it additionally produces extra knowledge, which is one thing that corporations usually don’t anticipate.
As well as, when corporations create a mannequin, it’s outlined by its coaching knowledge and weights, so retaining observe of various variations of an AI mannequin would possibly require retaining copies of each particular person coaching knowledge set. It relies on the particular use case, says Thota. “No one has discovered what the easiest way is,” he says. “All people is studying as they’re iterating.” And all of the infrastructure issues — the storage, connectivity, compute, and latency — will solely enhance subsequent yr.
