Generative artificial intelligence is transforming every industry, and the intersection of generative AI, cloud computing and data transformation is reshaping industries and user experiences. The open question, however, is how businesses will leverage and profit from AI as they grapple with the rapid evolutions in the industry.

That was a central focus of SiliconANGLE’s and theCUBE’s Supercloud 4 event last week, which examined how every single industry is affected by these new advances.

“This is changing how organizations are really looking at their data, entering into conversations with their customers, using that data and having their customers actually engage with them through the AI,” said theCUBE industry analyst Rob Strechay (pictured, right). “It’s really exciting stuff.”

Analysts for theCUBE, SiliconANGLE Media’s livestreaming studio, spoke with artificial intelligence executives, experts, technologists, investors and thought leaders during the event. They explored key trends surrounding generative AI, supercloud, as well as the future impact on the enterprise world.

Here are three key insights you may have missed from the Supercloud 4 event:

1. GenAI is the wave people thought it was.

With Supercloud 4 in the rear-view mirror, it’s clear to theCUBE industry analyst John Furrier (left) that GenAI represents the biggest wave we’ve ever seen.

“Probably bigger than the combination of all the waves combined — PC, web, mobile and now this,” he said during a final wrap-up analysis from Supercloud 4. “If you remember our re:Invent scoop with Adam Selipsky, we were kind of focused on this next-gen cloud with the ecosystem, introducing the supercloud concept, understanding it. It’s happening.”

Super-structures of data and super apps will have a big impact, according to Furrier. That’s because they’ll be able to scale labor, reduce costs and increase the creative intellect. “This is a generational movement,” he said. “It’s a revolution, in my opinion.”

It’s similar to what happened with the internet, according to theCUBE industry analyst Dave Vellante (middle). Everyone was able to take advantage of it, which is what supercloud is all about.

“We’ve now got this cloud infrastructure that’s out there that’s proven, it’s resilient, it’s secure, and you can build value on top of that,” Vellante said. “And the whole world is doing that. It’s digital transformation meets cloud meets supercloud meets AI.”

Of course, as SiliconANGLE Editor in Chief Robert Hof noted this week, the excitement and activity around AI has yet to translate into earnings for big cloud and tech companies. The pace has been slow but has been moving past sporadic exploration, according to Ori Goshen, co-founder and co-chief executive officer at AI21 Labs Ltd.

“There’s almost no company, no enterprise on earth, that is not looking at how this impacts the business,” Goshen told Furrier. “Thinking at it at the board level, I think it really reached that kind of attention, and visibility, and that drove a lot of interest.”

Here’s the complete video analysis with Dave Vellante and John Furrier, part of SiliconANGLE’s and theCUBE’s coverage of Supercloud 4:

2. Democratization provides benefits and dangers.

There has been a tsunami of announcements from companies working with generative AI in recent months, including Amazon Web Services Inc. rolling out general availability of Amazon Bedrock earlier this year. There are, of course, curated state-of-the-art models, but there are also open-source models, among other kinds of models, noted Bratin Saha, vice president and general manager of AI and machine learning at AWS.

“We think that a single model isn’t going to work, necessarily, for all use cases,” Saha said. “Customers want this choice.”

Of course, the democratization of AI and the massive demands on data that go with that shift have also posed challenges for enterprises. Companies must ensure they protect against what’s known as “model collapse,” according to Andy Pernsteiner, chief technology officer of VAST Data Inc.

“The premise is basically that, based on the fact that much of the data being generated now is being generated by generative AI technologies like GPT-4 and others, that the data being proliferated into the world is now data that isn’t really useful for training on,” Pernsteiner said. “Even OpenAI and other organizations are having to purchase and acquire real data because they don’t want to be left in a place where they’re only left with training against data that is synthetic.”

There are ways that people developing models can work around it, but it is extra effort as opposed to training against real data, according to Pernsteiner. That’s because if everything coming in is synthetic, it’s challenging to know if it’s a real signal or a fake signal.

“It’s very easy to get bias into the models,” he said. “It’s very easy for the models to become polluted over time. That’s one of the challenges, to prevent model hallucinations especially, is that the input has to be quality.”

Here’s theCUBE’s complete video interview with Andy Pernsteiner, part of SiliconANGLE’s and theCUBE’s coverage of Supercloud 4:

3. It’s all about the data.

One thing to keep an eye on moving forward is the fact that the consumerization expectations from AI users will force every company to have an experience for those users, noted Furrier. That requires a new way of managing the data that’s need to customize those experiences.

“The change of how data will be managed is going to be a radical shift,” Furrier said. “Everything we know about data, data warehouses, data management, all that stuff, in the next three to five years will be completely transformed.”

There are various new approaches at play when it comes to these matters, such as ChaosSearch Inc. focusing on log analytics at large scale, facilitating data integration and analysis. There’s been a recent rise of vector databases, and they are a great tool to do a similarity search, according to Thomas Hazel, founder and chief technology officer of ChaosSearch.

“However, it’s a little complicated to put all your data streamed into it, particularly at scale,” he said. “What’s interesting about these LLMs, these multimodal LLMs, is maybe the future is to transcribe in high fidelity, all these images, all this logging data, all these audio, videos, into text, and index that text as semantic search.”

As a part of theCUBE’s research team, Strechay has been tracking the role of data in this new world over recent months. There’s still a lot unknown about what comes next in that regard.

“Everybody is trying to figure out, where is all their data,” Strechay said. “So there’s a big infrastructure push to, do I consolidate it all into one data platform? Do I have multiple data platforms? How do I have a mesh over them? Where do I do transformation? There’s a lot still going on at that data layer as people talk about building data projects.”

Moreover, that data needs to be carefully managed as more companies focus on domain-specific segmented language models or smaller language models. For instance, one may be using it for human resources, while another may be using it for finance, Strechay noted.

“Bringing that data together, because they own that, and that’s their IP,” he said. “But there’s big security concerns, and how do I have a moat between that and make sure that it’s not going back into the models? There’s a lot going on with that data infrastructure.”

Here’s theCUBE’s complete video interview with Thomas Hazel, part of SiliconANGLE’s and theCUBE’s coverage of Supercloud 4:

To watch more of theCUBE’s Supercloud 4 event, here’s our complete event video playlist:

Photo: SiliconANGLE