Hypermodal AI Converges Predictive, Causal And Generative AI

AI is diverse. While we live through the seemingly almost instantaneous evolution of the modern approach to Artificial Intelligence (AI) and its application into the algorithmic logic that runs our applications, AI is diverse and divisive in terms of our collective ability to agree on when and where it should be used.

In software application development environments, the consensus is gravitating towards the use of AI as a helping and testing mechanism, rather than it being wholly offered the chance to create software code in and of itself. The concept here is that if so-called citizen developer business laypeople start creating code with software robots, they will never be able to wield the customization power (and ability to cover security risks) that hard-core software developers have.

As we now grow with AI and start to become more assured in terms of where its impact should be felt, we may now logically look to the whole spectrum of automation that it offers. This involves the concept of so-called hypermodal AI i.e. intelligence capable of working in different ‘modes’, some of which will predict, some of which will help determine and some of which will generate.

Predictive, causal, generative

Today describing itself as unified observability and security platform company (IT vendors are fond of changing their opening ‘elevator sell’ line every few years), Dynatrace has now expanded its Davis AI engine to create hypermodal AI that converges fact-based predictive AI, with causal AI insights with new generative AI capabilities.

“Generative AI is a transformative technology with seemingly limitless possibilities for delivering productivity gains,” said Bernd Greifeneder, CTO at Dynatrace. “As organizations look to tap into this potential, the key to success is hypermodal AI that combines generative AI with powerful causal and predictive AI techniques. This is because only predictive AI can see into the future reliably, only causal AI can deterministically know the root cause of an issue… and only generative AI can tailor recommendations and solutions to specific problems using advanced probabilistic algorithms. With the release of the expanded Davis AI, we address this need and redefine how observability and security solutions work.”

Traversing vertical topologies

Greifeneder and team explain this technology’s ability to ‘traverse a vertical topology map’ of every entity in an enterprise’s hybrid multi-cloud would i.e. a map that takes us from on-premises computing power in the company headquarters, onwards to public cloud datacenter resources and back through the ‘edge’ computing estate made up of mobile devices and remote units (machine sensors, cameras etc.) that form the Internet of Things.

The expanded functions in this AI offering produce generative AI recommendations fueled by context from causal and predictive AI techniques that reflect the individual attributes of an organization’s cloud stack. It will also simplify tasks such as creating automations (such as software bots designed to take repetitive but accurately measurable manual tasks away from human users) and dashboards used to track IT system health.

As it is probabilistic by nature, Dynatrace’s Greifeneder says that generative AI’s value depends on the quality of its training data and user prompts. Because of this, he asserts, the power of generative AI can be greatly amplified by converging it with causal and predictive intelligence to create a single hypermodal AI with each type of AI within it having specific capabilities. So let’s look at each of those modal modes to see how they work and try to understand how they can work together.

• Predictive AI

Predictive AI is (unsurprisingly perhaps) best used for forecasting, it’s not too tough to work this one out. By ingesting data from past events covering everything from customer demand trends, seasonality, popularity spikes and so on, we can build up a picture of historical behavior types that exist in any given market or scenario. When we apply that same approach to a cloud estate and look at corresponding levels of performance and application health through the same cycle period, then we can make predictive forecasts to recommend future actions as market dynamics continue to play out.

• Causal AI

This is AI used to determine why things have happened. Causal AI in cloud management terms is built to deliver fact-based, deterministic precise answers based on analyzing dependencies (vital links between applications or services) across large sets of observability and security data while retaining an accurate context that reflects each data point’s source. Used diligently, this type of AI can enable us to provide intelligent automation to control systems for better performance.

• Generative AI

As it sounds, generative AI is used to generate. While we may use it in line with the text-based approach characterized by Large Language Models (LLMs) to generate answers based upon next-most-likely words or terms, it is applied in cloud management to recommend how to solve specific tasks in the context of the customer’s own environment and situation.

“Davis predictive AI models and dynamic machine learning is used to anticipate future behavior based on past data and observed patterns. This allows customers to anticipate and remediate future needs and issues related to the performance and security of their applications and the underlying infrastructure before problems occur,” notes the Dynatrace team, in a technical product statement.

The company further notes that Davis causal AI analyzes real-time context-rich observability and security data within its own-branded Grail data lakehouse and causal dependencies from its Smartscape topology to provide the precise answers and intelligent automation that are necessary for issue prevention, deterministic root-cause analysis and automated risk remediation.

Finally, for now, Dynatrace also makes note of its Davis CoPilot generative AI technology. This works with Dynatrace causal and predictive AI to automatically provide recommendations, create suggested workflows and dashboards, or let people use natural language to explore, solve or complete tasks.

“Generative AI is already proving to be useful in broadening the accessibility of operations insights to new personas and speeding workflows for users of observability solutions,” said Nancy Gohring, IDC research director for enterprise system management, observability and AIOps. “However, when combined with other forms of AI, generative AI has the potential for additional notable impact. For instance, leveraging other forms of AI to feed generative AI with more than just user inputs can deliver more value for customers and help maximize the value of generative AI for business, development, security and operations use cases.”

The AI mixing pot

Do we need more than one core type of AI engine to be at work in order for the total automation package to be smart in all the right ways – and does the number of modes need to stop at three? What about theory of mind AI – an approach where AI starts to be able to appreciate human thoughts and emotions – does that need to be in the mix too? We can probably skip over self-awareness, machine sentience and synthetic consciousness for now because this is ‘only’ cloud system observability and security we’re concerned with, but for how long?

When hypermodal AI becomes super-hypermodal AI, then we’ll know we’ve moved to another level. For now, let’s just observe.